00:00:00
Jason where are you that is that a virtual background or oh oh right that's place I I did look
00:00:07
architecturally familiar to me it is architecturally significant we bleep out the beep's house but yes I am you know
00:00:13
this is like top three or four of places I like to be a house guest you know like
00:00:18
I'm in rotation right now my house guest you're K kining through our friend group
00:00:23
basically you know I'm just a great house guest people like to have breakfast with me people like having me
00:00:27
around so I just find myself in I around the world hey Nick do you have an updated picture of K kin is he still
00:00:34
alive and kicking he's alive for sure he's got to be 70 or something right he's gotta be he was on one of those
00:00:40
reality TV shows with Dr Drew I think a couple years ago what does k o kin do he's a good hang he's yeah there's a lot
00:00:46
of these people in La you know like just they just hang they kind of hang they set things up oh my
00:00:54
God is that current k k man he's 64 now wow God Almighty that's incredible how did he survive he's got an Instagram
00:01:04
account I'll tell you how he survived see no evil you're no evil exactly I didn't see nothing all I know is Hey
00:01:13
listen man you gave you a pool house snitches get stitches I didn't see anything what did you see
00:01:19
nothing let your winners ride Rainman David and in said we open sources it to the fans and they've just gone crazy
00:01:31
with love [Music] youen all right everybody welcome to your favorite podcast the Allin podcast
00:01:39
episode 168 David saaks can you believe it we've made it to 168 episodes with me
00:01:46
again the Rainman David Sachs how you doing buddy good yeah I heard you got a big talk coming up another speech yeah
00:01:57
yes a talk yes I'm giving a talk all right get ready for that all of the GOP fans out there all your saxs fans you
00:02:04
got a big keynote coming from saaks next week also with me of course the Sultan of science formerly known as the queen
00:02:11
of quinoa he's got another crop he's uh growing David freeberg how are you doing how's the crops how's the fields
00:02:21
how's life in the fields worth cold open not a cold do these are intros oh intro
00:02:27
I'm glad you're not producing there's a reason why I'm the executive producer go
00:02:30
ahead free how's your um how's your crops how's the how's life in the fields it's great got a great team making
00:02:37
progress Tech is awesome having a lot of fun it's great back in the CEO seat huh and of course the chairman dictator
00:02:45
who's becoming completely insufferable because he invested in a company seven years ago that is absolutely crushing
00:02:50
right now called grock we talked about it last week you're back you're back chth everybody's talking about Gro
00:02:57
you're you're on uh Cloud 9 seem gr gr gr Bitcoin Bitcoin Bitcoin groc gr I mean it is interesting how your
00:03:10
book just literally whatever chat's book is that's his mood it's just Bitcoin 60
00:03:17
grock well the honesty level is going to be through the roof now oh right he going to be running for governor and
00:03:25
buying the Hamptons again the truth bombs that jamath is going to drop now there is nothing like Peak Zer jth I
00:03:32
mean are we going back is are we back right now I on paper what is your stake and grock already worth let we're not
00:03:39
talking we're not we're not we're not talking that's so un it's so UNC but he's buying the Hamptons and he's
00:03:44
probably gonna buy stop what has it been uncouth on this podcast to talk about about wins or
00:03:51
wins or assets these are not I agree with that it's not a win you can never count you
00:03:56
can never count your chickens before they hatch that's absolutely right I am not counting any chickens don't book The
00:04:01
win yes there's a lot of hard work to do there's a lot more people to sell products to to build products by the way
00:04:09
developers by the way on Gro just this past week in the queue the weight list tripled now there's almost 10,000
00:04:19
developers that's a big deal well that's crazy that's a great sign I think the most important Northstar metric for
00:04:25
these developer platforms is basically that as goes the developers so goes the platform you can kind of count on some
00:04:33
amount of pull through because some of those developers just statistically will land really important products they'll
00:04:37
consume more apis they'll just consume more stuff it's just all goodness because if you have 9,000 or 10,000
00:04:44
people just again waiting in line once those guys get in somebody's going to create something
00:04:51
magical that's the great part of having a platform people just take it and they build on it while you're sleeping
00:04:58
something amazing can just get built and you get some amount of the uh credit for
00:05:03
that the thing that's really interesting is the number of developers is increasing I've seen two or three
00:05:08
founders recently just the last couple of weeks who were previously idea Founders who are now have taught
00:05:14
themselves to code so you know this there's 1% of the you whatever number of millions of delers in the world I think
00:05:20
it's going to like double or triple so the the pool of people who are writing code I think is about to grow very
00:05:26
meaningfully I mean you guys you guys saw the release from the White House so they were like we we don't want you to
00:05:32
code and C at C++ anymore yeah that was very interesting I mean they were talking about memory
00:05:37
leaks and like this is I guess the source of most memory leaks therefore the White House wants you to learn to
00:05:42
develop in memory safe languages oh great awesome yeah so why are they involving themselves like they got
00:05:48
nothing better to do yeah I mean they have an opinion on uh you know preventing security leads all right
00:05:55
listen the top issue this week is the same issue as last week Google's Gemini I Dei black eye continues we covered
00:06:03
this woke AI disaster last week was kind of funny I was watching CNBC and they had a hard time describing the problem
00:06:11
the woman just I guess didn't want to call it what it is it's a racist AI you type in text and it gives you the
00:06:18
opposite or just culturally insane responses so if you put in you want a picture of George Washington from
00:06:24
Google's gemini or Sergey Bren you might get back like a beniton style diversity
00:06:29
with like George Washington being black or Sergey Bren being Asian Etc and so this has caused um a bit of a gr fluffle
00:06:39
here in the industry to say the least the stock is down 5% since we talked about it last week and uh Sundar sent a
00:06:45
memo to the Gemini team of course when they write these memos to a team it's written to the entire world because you
00:06:51
know it's going to get leaked and so you're writing as a such it might as well be a press release I'll give uh two
00:06:56
quick quotes here and then I'll throw it to the besties because there's so many questions that we have to address quote
00:07:00
number one I know that some of its responses referring to Gemini have offended our users and shown bias to be
00:07:07
clear that's completely unacceptable and we got it wrong and to be clear it wouldn't show white people especially
00:07:13
ones like George Washington or just somebody who is obviously a Caucasian and so the next quote will be driving a
00:07:21
clear set of actions including structural changes to me structural changes means we're going to lay off a
00:07:28
bunch of people and we're going to get rid of the Dei group that's that's a new ma that's a new found motivational riff
00:07:34
quote in my mind freeberg will Sundar survive and is Google too broken to fix I'm just going to ask you since you work
00:07:43
there I don't want mean to make it uncomfortable but what are the chances Sundar survives this and what are the
00:07:49
chances that Google can be fixed and produce great products quickly that to light
00:07:57
users I don't know how answer the sunart Will Survive because it's kind of an idiosyncratic organization there
00:08:04
are a couple of Founders who have super voting shares um and ultimately comes down to their decision and the direction
00:08:14
they want to take the company and I have no insights into what they individually
00:08:19
think so um frankly I've spoken to a lot of folks who are investors in Google over the last week and a lot of folks
00:08:27
are just deeply frustrated and angry um uh on a number of fronts the business really there's you know three
00:08:38
businesses inside of Google there's you know search and ads there's YouTube and there's cloud and the rest of it is kind
00:08:43
of noise and to give you a sense of how big these businesses are right YouTube did 10 billion a quarter roughly Cloud
00:08:49
did 10 billion a quarter they have this devices business and subscriptions business does about 10 billion a quarter
00:08:56
and then search does about 50 billion a quarter and the margin on search is much
00:08:59
higher than any of those other businesses and so the search margin the ad revenue on
00:09:04
search is you know probably 100% of the true operating profit of the business so
00:09:10
the real threat to Google is more are they in a position to maintain their search Monopoly or maintain the chunk of
00:09:19
profits that that drive the business under the threat of AI are they adapting and less so about the anger around woke
00:09:26
and Dei because most of the investors I spoke with aren't angry about the woke Dei search engine they're angry about
00:09:32
the fact that such a blunder happened and that it indicates that Google may not be able to compete effectively and
00:09:38
isn't organized to compete effectively in AI just from a consumer competitiveness perspective so you know
00:09:44
investors are banging the table and in the past we saw this with meta I think it was towards the end of 22 if you guys
00:09:53
remember and it was a similar situation investors were like why are you investing in VR and AR this is crazy why
00:09:58
do you have all the these people that are getting overpaid and everyone started to write off the stock and the
00:10:03
stock took a big nose dive for a period of time just like Google's is right now and then the changes came and much like
00:10:09
Google there's an individual with super voting shares who basically said you know what I am going to step in and
00:10:15
we're going to make these changes and we're going to fix this organization and we're going to right siiz and we're
00:10:19
going to focus on the product winning and since then meta stock is up a tremendous amount 5x since then Google
00:10:26
Shares are down about 10% over the past two weeks by the way I was one of the stupid people to sell meta around that
00:10:31
time and your thinking was that they just can't get out of their own way and the god king is it's it's another yeah
00:10:37
it's exactly it's another one of these idiosyncratic problems you don't know what this individual is thinking and
00:10:42
what he is individually going to do the point I'm making is that at Google Now something has to give because the noise
00:10:48
is so loud the board is hearing this left and right investors are banging the table analysts are banging the table and
00:10:56
I I'll tell you a couple anecdotes internally employees are now banging the table a story I a story I heard this
00:11:02
week was that someone stood up in a meeting and said a couple of weeks ago if I had stood up in this meeting and
00:11:09
said we can't show black people in the image generation at the rate that we're showing I would have been cast a racist
00:11:18
and I didn't have permission to do that inside of the organization but today and
00:11:22
everyone's like you're right you would not have been able to stand up in the organization and say that but the tenor
00:11:26
has changed inside of Google with a lot of the employ that I've spoken with who are now saying I can stand up and I can
00:11:33
say that this group called responsible AI has too much power and it's a one-sided asynchronous problem where
00:11:40
they get to come in and say we need to change this and if you step up and disagree with them you are deemed a
00:11:44
racist you were deemed you know culturally inappropriate easier to keep your head down it's easier to keep head
00:11:52
circum understand and and keep collecting your RS all of a sudden you wake up one day and you see the blunder
00:11:56
that happened last week yeah and so now internally people are waking up and saying we need to change this and I
00:12:01
heard that that's made its way up to the higher ranks at Google and they're very
00:12:04
actively you know so there may be a moment here where Google stock which currently is trading at just 17 times
00:12:12
2025 consensus earnings which is cheaper than all the other big tech companies by
00:12:17
far and it's still growing Core Business is growing cloud is growing 20% search is growing 15% all these other
00:12:23
businesses YouTube's growing 20% a year it's a growth business that's very profitable
00:12:30
and um it's trading at a very cheap discount so there's you know the bull case is now's a great time to buy
00:12:35
because it's so cheap and there could be this moment where you know you see some
00:12:38
of the changes that are needed internally to get the AI products to where they need to be to maintain the
00:12:44
lead that is inherent because of C so that would mean Sachs it would require because we're using Zuckerberg as an
00:12:50
example we could also bring Twitter and X into this it would require founder authority to come in there and make
00:12:56
these changes it would require larryan Sur who have the super voting chairs to come in there and say hey this is all
00:13:01
changing enough of this do you think there's any chance of that happening or is Google just too broken to fix and
00:13:10
they're going to just lose this opportunity and it's Microsoft under Steve bomber missing mobile and
00:13:15
Cloud well to quote Jefferson the tree of Liberty must be refreshed from time to time with the blood of patriots and
00:13:23
Tyrant and I think there's something analogous here I mean you have to if you're going to refresh this company you
00:13:28
have to go in and you got to go in and make major Cuts not just to rank and file but to leadership who doesn't get
00:13:34
it and that's the only way it's going to get fixed do I think that Larry and Sergey are going to come in and pull an
00:13:41
Elon and go deep and figure out which 50% or 20% of the company's actually good and doing their jobs probably not
00:13:51
but is it possible that they could make a leadership change yeah it's possible probable I don't know I mean I've heard
00:13:58
that the company is the way it is because they like the way it is I mean that they're part of the problem in effect
00:14:04
they don't see they don't see the problem or at least they haven't until now maybe they'll get the wakeup call
00:14:10
what do you think shimar watching while this happen I think it's basically a small Cadre of a thousand people that
00:14:17
have built literally the most singular best business model and Monopoly ever to be created on the internet and a whole
00:14:26
bunch of other people that have totally transformed this organization as sack said
00:14:33
into ability and a platform to reflect their views and so I don't not a shareholder
00:14:40
of Google and outside of the tools I use I don't think I really have much voting
00:14:47
power so I don't and and I have so many Alternatives now so I actually think like the I don't really care that much I
00:14:55
guess is the point I think that the employees should care and the shareholders should care and they should
00:15:00
come together and vote and I think saak is right I think the company is the way it is because they've chosen to be that
00:15:07
way and I think freeberg is right which is that there's a small group of people who have been protecting and breathing
00:15:14
life into the single greatest business ever built ever in the history of business but now we need to have a
00:15:22
confrontation amongst all of these three different groups of people and they need
00:15:25
to make some decisions let me put a few data points in in play here jcal and this all this
00:15:32
all speaks to the problems being let's say deliberate as opposed to a glitch or an accident so first you have sundar's
00:15:40
letter to the company which there was a very interesting tweet by Lulu Chang misseri who does coms at Activision and
00:15:51
she writes a Blog called Flack she said quote the OB she a com's expert just for
00:15:56
the audience she's a comms expert yeah so she kind of red sundar's letter and it was scorching she said the ausc lack
00:16:03
of clarity and fundamental failure to grasp the problem are due to a failure of leadership a poorly written email is
00:16:09
just the means through which that failure is revealed so that was one reaction Mark andrees had a series of
00:16:17
posts indicating that the AI was programmed to be this way again it's not like a bug it's more of a feature and
00:16:25
that it's not an accident this is this is happening because because companies wanted to be that they chose to be this
00:16:32
way and in fact he goes further and says that these companies are lobbying as a group with great intensity to establish
00:16:38
a government protected cartel to lock in their shared agenda and corrupt products
00:16:43
for decades to come wow I mean again that's a really scorching critique here and then you know the the question I
00:16:51
would ask is who's been fired for this I mean imagine if I don't know one of elon's products had a launch that went
00:16:58
this badly do you think no heads would roll no heads have rolled and so you have to kind of Wonder well what if it's
00:17:06
a structural issue sex I mean I I I the point does resonate with me that there is a group internally and I think it's
00:17:12
called the responsible AI team at Google and this team's job is to enforce those
00:17:18
principles remember that we brought up last week that they've defined and so they go in and they're like well you
00:17:23
know if you just render an image of a software engineer and all the images are just white guys in hoods that's
00:17:29
inappropriate because there are plenty of non-white people you need to introduce diversity so then the
00:17:33
programmers say okay we'll go ahead and overweight the model and make sure that there's diversity and you can't say no
00:17:38
because otherwise you are deemed a racist so who is the individual that's responsible given that structural
00:17:44
circumstance that exists within the organization it's more of a cultural and structural problem to me than you know
00:17:50
one I guess ultimately there's leadership that's lacking but actually I I I agree with that to some degree let
00:17:55
me describe how I think it it works mhm so what what I've heard about Google is that every meeting above a certain size
00:18:02
has a Dei person in it I mean literally so it's kind of like in the days of the Soviet Union their military the Red Army
00:18:10
would have in every division or unit there would be you know a commander or lieutenant and there'd be a commissar
00:18:18
okay and the commander reported up the chain and the commissar reported to the party and the commissar would just
00:18:24
quietly take notes in all the meetings of the unit and if the commissar didn't like what the lieutenant was doing
00:18:29
Lieutenant would be taken out and shot okay now that's kind of a dramatic example but the point is that in every
00:18:36
large meeting at Google you've got this Dei commissar who's like quietly taking notes I'll point out that at Google the
00:18:44
only person we can ever remember to get fired was James Dore who was an engineer
00:18:49
who complained about the political bias at Google in other words he was a whistleblower about the very problem
00:18:55
that's now manifested he's the only person you can think think of to get fired at Google Google was mocked on the
00:19:01
show Silicon Valley it was called hulie but remember this yeah they the the nobody ever got fired because it's a
00:19:08
company where it's impossible to get fired unless you blow the whistle on the political problem and I think that if
00:19:13
you're sitting in those meetings W with a Dei commissar present and you know or you have suspicions that the AI product
00:19:22
is not working right are you really going to speak up and risk the fate of a James Dore of course you're not
00:19:28
right so you're right freeberg that it's a structural problem that there's probably a lowgrade fear that's
00:19:35
pervasive through the organization and no one's willing to say the emperor wears no clothes the woke Emperor knows
00:19:40
wears no clothes because they don't want to why stick your neck out you may not know you're going to get fired but you
00:19:45
know you're a chance three years ago on Twitter nobody would talk about certain topics related to Dei because it was in
00:19:52
the aftermath of George Floyd and people just did not feel comfortable even calling out something minor that was
00:19:58
unfair no I would say they didn't feel comfortable until Bill Amman broke the seal on this just uh like a month ago
00:20:04
where he really went after Dei I mean because remember the part of the reaction to bill akman was like wow he's
00:20:11
really going there even though everyone's saying that like basically agreed with him and knew that he was
00:20:16
making a correct point but people were still afraid to like again call out this this woke
00:20:22
Emperor yeah and the way this works for people who are not super aware you have a language model and you know you write
00:20:29
a bunch of code but then there are guard rails put in and then there are red teams and people who test it and so at
00:20:34
the very least even if you were trying to do something with great intent as you pointed out Friedberg hey you know if we
00:20:39
pull up a doctor it doesn't shouldn't necessarily always be a white guy there are other people who are doctors in the
00:20:45
world somebody should have caught it in testing they probably did Jason and they
00:20:50
didn't report it who wants to report that problem that's an interesting rub yeah how guts they didn't miss it they
00:20:57
didn't miss it didn't have the guts to you have to build a somebody built a reward model or people a reward model
00:21:04
was built for the reinforcement learning from all the humans and their feedback so these decisions were explicit the
00:21:11
question that is that you guys are framing is was this though a case where it was
00:21:17
explicitly imposed on people and people felt a fear of pushing back or did they just agree and say this is a great
00:21:26
decision and these rewards make ton of sense and the whole point is that I think what this is highlighted is that's
00:21:33
the truth that Google needs to figure out and they need to figure it out quickly because what is going to happen
00:21:39
now is it's we I know I think we talked about this a year ago all Google needs to see is 300 500 basis points of change
00:21:49
and the market cap of this company is going to get cut in half okay because there is only one way to go when you
00:21:54
have 92% share of a market and that is down and so the setup for the stock is now that people are looking at this
00:22:02
saying okay if I see 92% go to 91 or 90 or 89 that's all that has to happen and people will say the trend is to 50% and
00:22:13
you will price this company at a fraction of what it's worth today so I think it's really critical this is the
00:22:19
moment that the senior leadership that really understands business can separate it from politics and decide is this a
00:22:26
thing of fear where there's one Rog group that's run a muck or is this what we believe because if it's the ladder
00:22:34
you just got to go with it there's nothing you can do because you're not going to replace 300,000 employees or
00:22:38
however many Google has but if it's the former sax is right you're going to have
00:22:42
to some heads need to roll and they need to tell the marketplace that this was a mistake
00:22:47
where a group of Rogue employees got way out way too much power we were asleep at
00:22:52
the job but now we're awake they're fired and this is so they have two choices but in in all of these choices
00:22:59
what I'm telling you on the the dispassionate Market side is if you see perplexity or anybody else clip off 50
00:23:07
basis points or 100 basis points of share and search this thing is going straight down by 50% by the way unless
00:23:14
Cloud takes off right because the other hedge that Google has is gcp and the tooling that they've built
00:23:20
in gcp can enable and support and be integral to a lot of Alternatives and competitors to what ultimately might be
00:23:28
searched so I think there's also a play here in thinking about some of the hedges that Google has implicitly built
00:23:34
into the business I'll just say one more thing on this structural point I kind of
00:23:39
thought about this as when I started speaking to people internally about how this happened from a product perspective
00:23:44
it felt a lot like when you talk to a lawyer as you guys know you're making a business decision and there's some risk
00:23:50
I mean think about Travis building Uber and the lawyer will say it is illegal you need a medallion license to do what
00:23:57
you're doing in this City and Travis is like well you know what I'm going to take that risk Brian chesky and Airbnb
00:24:02
he said you know what I'm going to take that risk and the problem is that like a
00:24:05
lawyer's job is to tell you what you can't do to identify all the Peril of your of Your Action and then you as an
00:24:13
executive or a business leader or a manager your job is supposed to be to take that as one piece of input one
00:24:20
piece of data that you then make the informed business decision about what's the right way to build this product
00:24:25
what's the right way to build this company and I'll take on some risk and I think one of the challenges structuring
00:24:30
side of Google is that product leaders and other folks were never enabled to make the decision that there were these
00:24:37
as saaks pointed out kind of like policing type organizations that were allowed to come in and veto things and
00:24:43
the vetoing or make unilateral decisions and those vetoing on whether it's waiting or or some you know training
00:24:49
data set or output that ends up killing the opportunity for the smart business leader to say that doesn't make sense we
00:24:55
can't do this is where this is where having founders in the organization every day changes everything because the
00:25:01
founders would say hey we we're here to index the world's information we're here
00:25:04
to present the world's information we're not here to interpret it we're not here
00:25:07
to win hearts and Minds we're not here for a political agenda but there's a group of people there who it's apparent
00:25:13
think that they're there to change the world that Google is a vehicle for them to make social changes in the world and
00:25:21
you know that's art and there's other ways to do that and just paradoxically like Hamilton making the founding
00:25:27
fathers diverse and doing art you know that can win Tony's in that context you could win tons of awards and a claim and
00:25:35
and make incredible beautiful art but on the other side like people are not looking for Google to do that they're
00:25:41
looking for Google to give them the answer and the data and then these people are thinking it's our job to
00:25:45
actually interpret the data for you as we kind of touched on last week so uh this I think look I think you guys are
00:25:53
really close to the bullseye here but I would just refine slightly what you're saying
00:25:59
so here's how I think it happens I I don't think this is a rogue group I actually think this is a highly
00:26:03
empowered group within Google I do agree with you yeah yeah I think what happens
00:26:08
is Sundar says from the top that we're going to be on the Forefront of diversity and inclusion because he
00:26:15
personally believes that and that's the way the social winds are blowing and they think it's good for the company on
00:26:21
some level okay to implement that Mantra or that platitude really HR hires a bunch of Dei experts okay lots of them I
00:26:30
think the company has like this huge HR department and a lot of those people are
00:26:36
basically Fanatics I mean they're their chosen career yeah yeah they're trained marxists basically and so they're the
00:26:42
commissars and they're sitting in all these meetings and again they're the ones taking notes and they're the ones
00:26:48
who push the company in a certain direction but you have to then go back to senior leadership and say it's their
00:26:53
fault for letting this happen because they should have made a course correction they should have realized
00:26:58
what was happening they should have realized that through a combination of bias and through this sort of like
00:27:07
overly empowered HR team who are pulling a legal card I mean freeberg is right about that they're saying that our point
00:27:14
of view is the law which isn't true but they're basically pulling the legal card
00:27:18
and pushing the whole organization a certain direction it was up to leadership to realize what was going on
00:27:24
and make a correction and the the thing that you're seeing now is that in the face of what's happened the statement
00:27:33
that we got really doesn't cut it funny they use the word problematic yeah exactly they're describing it as a
00:27:39
glitch or a bug it's not it's a much deeper problem and so therefore you give it gives no confidence that the solution
00:27:47
is going to be pursued in as comprehensive manners as necessary yeah I think it's just a that that memo is a
00:27:53
tip off that there is a 20,000 30,000 person riff and a reorganized and they're just going to keep cutting the
00:27:59
de group and they've already meta and Google have cut the DI groups a bit I don't know if it's about as much about
00:28:05
cutting as much as it is about empowering if you said to the product leaders you can make the decision Dei
00:28:14
and responsible Ai and whatever other groups they're going to inform you on their point of view but they're not
00:28:20
going to tell you what your point that diag is pulling the legal card they're saying this is the law they're saying
00:28:25
that say that's what needs to change right if you don't do things the way we tell you Google's going to be hit with a
00:28:31
civil rights lawsuit that's what they're saying oh well I mean let's see if leadership can overcome that threat but
00:28:37
I think that's exactly the threat that is keeping the organization from resolving this problem or could keep the
00:28:42
organization that but I also wouldn't underestimate the bias so you know everybody there not I shouldn't say
00:28:48
everybody but it's a very liberal culture right it's a monoculture so when you're swimming in that much bias it's
00:28:54
hard to see right yeah yeah when everybody is left to Center I mean just look at something like political
00:29:00
contributions right it's 90s something per are democratic I mean like high 90s so it's just a a liberal bubble
00:29:09
basically and so right when everybody's liberal it's very hard to see when you know the results are Way offc Center as
00:29:17
well also if you if you give people this job like how are they what are they going to do every day freedberg if
00:29:22
they've been given the job Dei and they've been given the job to do trust and safety they they're looking to fill
00:29:26
their time in make an impact you got to I think maybe have less of these people I honestly think you guys saw you know
00:29:33
Shan Maguire is a partner at seoa and he used to be a member of the team at Google Ventures when it was called
00:29:39
Google Ventures and he did a Twitter post did you guys see this yeah where he said when he was at Google he was told
00:29:45
by his manager I'm not really supposed to tell you this it could get me fired but you're one of the highest performing
00:29:49
people here but I can't promote you right now because I have a quota my hands are tied you'll get the next slot
00:29:55
please be patient I'm really sorry It ultimately LED sea to leave Google and uh you know the rest is history he's a
00:30:02
partner at seoa now you're saying because he's a white male because he's a white male yeah and so the you knowon
00:30:11
yeah no I had the same thing happened to me on AOL when I was at AOL and shabbath
00:30:14
was there at the same time the whole organization was white men from Virginia whatever and they gave me an SVP title
00:30:24
and I said well I want the EVP one and they're like you know what we can't have any more white males in that position
00:30:29
right now we have to like get some more women and people of color in that position and they told me you'll have
00:30:34
the same comp and bonus but we just can't give you the title jamat do you think these
00:30:40
like race and gender driven quotas make sense for HR departments to try and enforce upon
00:30:51
managers to M to increase diversity I mean is that a good objective for an organization to have this is like a
00:30:57
topic a lot of people I've heard kind of flip back and forth on I there is no company where I have majority control
00:31:03
where I have an HR department you don't have an HR department no say more why I think that it's very important you
00:31:09
should go to a very respected lawyer at a thirdparty firm someone very visible an Eric CER type person and you should
00:31:18
work with that law firm and retain them so that you have an escape valve if there
00:31:24
are any kind of serious issues that need to be escalated so that you can get them
00:31:29
into the hands of a dispassionate third party person who can then appropriately inform the board the CEO and investigate
00:31:39
so that that covers sort of all the bad things that can happen then there are buckets of I think good things one one
00:31:46
important set of things is around benefits my perspective is that the team should build their own benefits package
00:31:54
that they want they should understand the p&l of the company that they work for they should be given a budget and in
00:32:01
in my companies again what I do is I allow committees to form and I ask those committees to be diverse but what I mean
00:32:09
by diverse is I want somebody who has a sick partner I want somebody who has a family I want somebody who's young and
00:32:18
single so that the diversity of benefits reflects what all these people need they
00:32:22
go and talk to folks they come back and they choose on behalf of the whole company and there's a voting mechanism
00:32:28
then when it comes to hiring I think what has to happen is that the person that for whom that person will end up
00:32:34
working for it's the head of engineering it's the head of sales those are the people that should be running the hiring
00:32:42
processes I don't like to Outsource it to recruiters I don't like to Outsource it to HR so when you strip all of these
00:32:48
jobs away HR doesn't have a role and what is left over is the very dark part of HR in most organizations which is the
00:32:56
police person the police policeman right the what what is Saks call the commiss that is why everybody hates HR I've
00:33:03
never met a company where that is a successful role over long periods of time they are this conflict creating
00:33:11
entity inside of an organization that slows organizations down so that allows me to empower individuals
00:33:19
to actually design the benefits that they want to hire the team that they want and I let them understand the p&l
00:33:25
in a very clear transparent way and the results are what the results are you want more bonuses hire better people and
00:33:31
then what I do is at the end of every year I talk about this distribution of talent and I make sure we are
00:33:37
identifying the bottom five or 10% they need to be managed up or they must be fired every year and it does not matter
00:33:47
how big the company is you must manage up or out the bottom 5 to 10% and in some cases I'm talking about one person
00:33:54
because it's a small company and in other cases I'm talking about 30 or 40 people so just hire the best person for
00:34:00
the job no you eliminate HR yeah you empower the team make the make allow them to make their own
00:34:07
decisions measure it hold them accountable right so if you have a salesperson who just hires their I don't
00:34:15
know 15 sorority sisters or fraternity Brothers whatever it is and it just becomes not a diverse group well hold on
00:34:22
it is what it is that still may be diverse this is my point like the the thing is that like there's different
00:34:27
ways to sell there's different sales motions there's the sort of like elephant hunting kind of sales model
00:34:33
there's the Dialing for Dollars thing so even those 15 sorority sisters the way you describe it could actually be
00:34:38
diverse my point is that kind of superficial marking based on the mutable traits will not yield a great
00:34:45
organization instead it's you're empowered to hire whomever you want just know that at the end of the year we're
00:34:50
going to measure them your bonus their bonus the company's performance all performance yeah performance I mean
00:34:56
Frank slutman said this and he he got barbecued at some point he said I don't have time to do this diversity stuff the
00:35:03
best person by the way in in my companies like for example like when you do sock 2 compliance you have to
00:35:07
generate these reports okay especially for some of our customers some of our companies that actually show the
00:35:13
diversity of our team and when we measure them on the immutable traits that whatever they represent as their
00:35:21
gender whatever they represent as other dimensions we are incredibly incredibly diverse
00:35:28
anyways but the way I've seen a startup that isn't diverse you know Silicon Valley attracts such a diverse mix of
00:35:37
people you know what do we know about success factors for startups number one the ones that are successful have a
00:35:43
cultural meritocracy number two they're non-bureaucratic and they don't have too
00:35:48
much GNA basically overhead in the company now all those things are cont contradicted by having a large HR team
00:35:59
or especially a Dei organization right they add bureaucracy they add overhead and what's that Pips performance
00:36:09
Improvement they cut into well Pips are great I pip people all the time does it work I mean I I know some people are
00:36:15
just like these Pips don't work if you're an unper performing member of the team and you've been identified in the
00:36:19
bottom five or 10 percent we have a responsibility as management to coach you up or to get you to an organization
00:36:26
where you're not in the bottom 5 or 10% that's the right thing to do for people you do that by being very transparent
00:36:32
and writing it down you are not good at these things you're underperforming in these things fix them or you will not be
00:36:38
here that's a very fair thing to tell somebody yeah look how you want to implement your own meritocracy I think
00:36:44
there's different ways for Founders to do that the point is you want these companies to be a meritocracy you don't
00:36:50
want advancement in the company to be based on factors other than skill Merit hard work things like that we know
00:36:57
performance we know that's a bad bad path to go on I think a lot of Founders don't understand that Dei is not
00:37:04
something they have to do they don't have to have a Dei organization this has somehow become a thing it's it's not
00:37:11
required and I think people are realizing like why would you do that why would you create this large bureaucracy
00:37:16
in the company that undercuts the meritocracy that adds a lot of costs and that slows you down none of those things
00:37:24
will help your company what you need to have I think J laid out some really good
00:37:28
best practices you should have an outside Law Firm that I would say is you could call it HR law but I would just
00:37:35
call it employment law I would say like a non-ideological partner e an expert in employment law
00:37:41
who sets up your company correctly and to whom you can take a complaint if if an HR complaint gets raised through the
00:37:48
chain of your company you do have to take it very seriously and that's be a proper investigation and that's probably
00:37:53
best handled by an outside lawyer so get that out lawyer and then by the way always there's never a case where your
00:38:00
coworker should know the intimate details of any of those and that's what also creates this horribly rotten
00:38:07
culture in HR where these people act like Gatekeepers of secret information salary information bonus information and
00:38:15
then all of the other things oh you know did you know this person did this with it's terrible to have that inside of a
00:38:21
company it should go to a dispassionate third party person whose job it is to maintain confidential it and discretion
00:38:27
yeah while investigating the truth yeah when HR is too powerful in a company that's a red flag at the end of the day
00:38:33
HR should be an administrative function their job should be to get people on boarded sign their offer letter and
00:38:40
their confidentiality and uh invention an assignment agreement and set them up in payroll and get them benefits and
00:38:47
that kind of stuff it should fundamentally be an administrative function and if it starts getting more
00:38:53
powerful than that it means there's been a usurpation you don't even need people
00:38:57
for that like you have software that does that now and it's all automated you know you send them a link you go to your
00:39:02
favorite you know HR site boom it's done yeah I don't think you need a lot of people doing this anything you want to
00:39:07
add to this discussion freeberg as we as we move on to the next topic I think that unfortunately the term
00:39:13
diversity Equity inclusion has been captured along as chth points out a single Vector which is this immutable
00:39:21
trait of your racial Identity or gender and I think the more important aspect for the success of a team for the
00:39:29
success of an organization is to find diversity in the people that comes from different backgrounds different
00:39:36
experiences different ways of thinking and so I I'm not a huge fan of race-based
00:39:44
metrics or gender based metrics driving I'm generally more oriented around be blind
00:39:50
to those those those variables and focus much more on the variables that can actually influence the outcome of your
00:39:56
organization yeah one of the great paradoxes of this as well is we are moving to a much more Multicultural
00:40:04
mixed race Society anyway people filling out fors a lot of our kids you know are
00:40:09
are could pick two or three of the different boxes on a Dei form it's it's not going to make much of a difference
00:40:14
in the coming decades all right issue two Google go splashy cach AES licensing deals for training data and it's now
00:40:21
becoming a bit of a pattern Google we talked about I think just last week had to deal with Reddit for $60 million
00:40:29
that's reportedly per year today stack Overflow is now using its overflow API to train Gemini no word on the contract
00:40:38
value I did get some back channel that it's a multi-year non-exclusive deal according to R's S1 they have already
00:40:45
closed 200 million worth of AI licensing deals over the next two to three years so you maybe it's going to be 75 million
00:40:51
a year 100 million a year who knows how big that business can get we're going to
00:40:54
talk about the S1 from Reddit in just a moment and uh this is on top of all the other licensing deals that have occurred
00:41:02
AEL Springer in open AI you remember that one and opening eyes and talks reportedly with CNN Fox and time to
00:41:09
license their content that comes on the hills of that Blockbuster New York Times
00:41:13
open AI lawsuit that we talked about I don't know 10 episodes ago and open AI lawyers are trying to get that one just
00:41:18
to give you a little update on it they're trying to get that case dismissed saying that the New York Times
00:41:22
hacked chat GPT to get certain results and that the York Times took tens of thousands of tries to generate the
00:41:29
results yada yada and they said here's the quote from the filing from open AI the allegations in the times complaint
00:41:35
do not meet its famously rigorous journalistic standards both open Ai and Google and Gemini have been fly Guard
00:41:42
railing their systems as we talked about as well to stop copyright infringement like trying to make pictures of Darth
00:41:48
Vader and that kind of stuff Jam you've talked a little bit about your Tac 2.0 framework maybe you could talk about
00:41:54
what you see Happening Here with all these licensing deals and what it means for startups in the AI
00:42:01
space well just to maybe catch everybody up Tac is this thing called traffic acquisition cost and you can see it most
00:42:09
importantly in Google's quarterly releases which is that what they realized very early on at
00:42:16
the beginning of the search wars in the early 2000s is that they could pay people to offer Google search people
00:42:25
would use it and then it would generate so much money that they could give them a huge ref share and it would still make
00:42:30
money so I remember Jason when you and I were at AOL this is the first time I met
00:42:34
Omid we were flying back to California we were both in Dallas at the same time in like
00:42:40
2003 or four and that's when Omid did the first big search deal between AOL and Google and it was it was I want to
00:42:48
say hundreds of millions of dollars back then where Google pays you up front you
00:42:53
have to Syndicate Google search and then they clean it up on the back end with some kind of rure so what's incredible
00:43:00
is that that process has escalated to a point now where for example on the iPhone it's somewhere between 18 and 20
00:43:08
odd billion dollars a year is what now Google pays Apple so that's the traffic acquisition cost 1.0 R Tac
00:43:16
1.0 and I just said that we should call this Tac 2.0 except now what Google is doing is instead of paying for search
00:43:24
they're actually paying for your data and saying give it to me so that I can train my models and make it better and I
00:43:29
think that that's an incredible thing both it's very smart for Google but also it's great for these businesses because
00:43:36
it's an extremely high margin thing to do when you have a really good Corpus of of data that is
00:43:43
very unique so in the case of Reddit that $60 million deal I didn't I I looked through the S1 to try to figure
00:43:49
out whether it was a multi- your deal or not it wasn't totally clear but the point is that you know Google's paying
00:43:56
Reddit 60 million bucks and Jason you just said that they're they've done a couple more of these things that's
00:44:01
incredible this Tac 2.0 thing is amazing so if you're an entrepreneur building a
00:44:06
website or building an app that has really unique training data or really unique data you'll be able to license
00:44:12
and sell that and that'll be an incremental Revenue stream to everything you do in the near future that's amazing
00:44:17
that's what T 2.0 is it's going to be incredible for the entire uh content and community- based Industries you think
00:44:23
this could sustain content creation where advertising has become very difficult sax I guess my question to
00:44:29
toal would be do you think this is going to be available to small websites they'll somehow be some sort of program
00:44:36
or because I mean Reddit is one of the biggest sources of content on the entire web right it's like a top five traffic
00:44:41
site with tons and tons of user generated content yes do you think like a small publication would be able to
00:44:48
make these types of deals yeah and in fact I think like if you go back to search 1.0 that's exactly what these
00:44:53
small companies were able to do which was in a more automated way they were able to basically partner and in that
00:44:59
example what they would say is here Google why don't you just run your ads on our page right and that was sort of
00:45:05
in that web 1.0 world so Google had solutions for the largest companies on the internet all the way to the smallest
00:45:12
and in this Tac 2.0 world I do think that that it works in that way as well the problem in the in that world if it's
00:45:19
a small website that says here's my training data the question is how do you attribute how much incremental value a
00:45:27
model derived from it versus something else and so I think that that part has to get figured
00:45:33
out and so you know what Google will be able to pay you will probably be pretty di
00:45:41
minimouse bound is 60 million for Reddit then the average website's going to get
00:45:45
a few hundred bucks but that still may be a good start and when Google figures out how to
00:45:50
monetize stuff or somebody else where then they can give you back some way to make money I think I think that there's
00:45:57
a real monetization here I really do you took the other side of this but now that you see the
00:46:06
market-based solution starting to emerge what do you think of this market-based solution think it's a got
00:46:12
legs well I'm not convinced that this looks like Tac in the traditional sense where you're basically buying a
00:46:21
continuous stream of traffic and then you're helping to monetize that traffic that's a effectively what Google did in
00:46:27
the ad syndication business and does today that business makes about 10 billion a quarter at
00:46:32
Google and they're paying call it 70 to 80 cents on every dollar back out to the
00:46:38
owners of that traffic the folks where that traffic is derived from I would say that this looks a lot
00:46:44
more like the content licensing deals to build a proprietary audience which is effectively what
00:46:52
Netflix did they they paid Studios for Content Apple does this they have proprietary content that they pay
00:47:01
producers to make and they put on Apple TV Amazon does this and so on this is a lot more like that where there are
00:47:08
content creators out there whether that content is proprietary like the New York
00:47:12
Times or user generated like Reddit and what they're trying to do is acquire that content to build a better product
00:47:17
on Google Search and I'm not sure how you get paid a continuous licensing stream for that content once you've
00:47:24
trained the model the content gets old it gets stale at some point in a lot of cases like news and then
00:47:30
eventually if you don't have a highquality continuous stream of content it's not worth as much anymore to give
00:47:37
you guys a sense humans generate in total well let me just give you some stats there's about a million pedabytes
00:47:42
of data on the internet today and humans are generating about 2500 pedabytes of data new data uh per day right now
00:47:52
remember I shared a couple weeks ago YouTube's generating about two pedabytes of data uh per
00:47:57
day half of all data generated is never used so this is like records and files and stuff that gets put on log files log
00:48:05
files gets stored somewhere never accessed never used the majority of the rest of that data is not in the public
00:48:11
domain it's not on the internet so there is a lot of data out there what some people might call dark data to train on
00:48:19
and I think that as the identification of better sources of training data and the value of training data right now
00:48:27
we're in this kind of shotgun approach we're trying to blast out and you know Source lots of content lots of data just
00:48:33
like over time Netflix got better at figuring out what content to buy and what to pay for it and they're the best
00:48:39
at it I think so too will Google and others figure out what data is actually particularly useful what it's worth and
00:48:47
what to pay for it and so there's a lot of data out there to go and identify to mine to pay license fees to get access
00:48:54
to whether those are continuous license fees or one time is still TBD that's a key issue yeah so I think we're still a
00:49:00
little bit early to know if this is like you know a continuous model like a tack
00:49:04
type business or if these are sort of chunky type deals and we don't really know what the real value is yet and that
00:49:09
all changes over time and remember the the the rate of data generation is increasing so while we're generating 200
00:49:16
pedabytes of data per day as a species on the internet that number is going up every day and so every year all the old
00:49:23
data becomes worth even less so this is all changing fairly dynamically and I think there's a lot still to be figured
00:49:29
out on what the monetization model will be for Content creators and and how that's going to change over time yeah I
00:49:35
think it's a really good point some things will be like The Sopranos or Seinfeld or Simpsons where that library
00:49:40
is worth fortune and people will pay a billion dollars a half billion dollars some no one no one's no one's paying a
00:49:45
lot for old NFL games you know Nick founded by the way Reddit licensing deal was 203 million over it says 2 to three
00:49:53
years so let's assume it's call it three just to be so it's you know6 million a year it doesn't say whether the deal
00:50:00
with Google is exclusive wow so they could do they could do that same licing deal multiple times that's interesting
00:50:05
yeah none of these are exclusive it seems this is what I don't understand why why don't you do head deals like
00:50:10
with folks like Reddit where you actually do it exclusively like it seems like it's more valuable to spend a
00:50:16
multiple of this number for for one of the big seven who have tens of billions of dollars of cash anyways and block the
00:50:23
other players and block everybody else that just seems Reddit multiple gets capped if I'm Reddit I don't want to do
00:50:27
that deal yeah Reddit me I put it on the table Yeah because then my multiple is Capp like the the way I can monetize my
00:50:33
content is now set and I'm done inv get bought investors are like oh you're worth five times eai you know that's it
00:50:40
Reddit cor stack Overflow they're going to just get taken out I this is I think this is going to be the new mod did
00:50:45
around actually didn't cor raise uh 50 500 recently I think they're going to get taken out I I I I think these
00:50:51
businesses will become too valuable because they do have ongoing content that used to run you started web blogs
00:50:59
right and but think about the value of that content today it's negligible like it was very valuable at the time and as
00:51:06
time went on more content was being created a 100 times a thousand times 10,000 times more content that started
00:51:12
to overshadow the value of that content at the time the acquisition was done it made a ton of sense but all of a sudden
00:51:17
two years later particularly with the rate at which data is growing on the Internet it's like does it make sense to
00:51:22
buy any content anymore so you well on the other side of that is historical content could be worth a lot of money
00:51:28
especially some could be so if you had the Charlie Rose archive as an example what is you know he's probably
00:51:33
interviewed Kissinger 10 times and he's interviewed Kissinger for you know 10 hours I've got two almost 2,000 episodes
00:51:39
of podcasts I've done with startups over 13 years like yeah this weekend startups
00:51:45
archive is going to be worth something at some point right I don't think it's worth what are old baseball games Worth
00:51:49
right like I mean yeah they're rewatchable and I don't think the data from them is particularly important so I
00:51:56
I agree on that one but historical yeah we don't know but and I just question how much of reddit's content is actually
00:52:02
like long-term valuable versus like they're covering a topic and they're talking about interesting stuff and
00:52:08
then later it's a really really excellent point yeah and we'll figure that out okay and that's why I think
00:52:14
it's like it's the early days of knowing how to Value all this content particularly for llms and so we don't
00:52:18
really know yet over the next year this will all start to become clearer but it's it's again same thing happened in
00:52:24
music licensing yeah right but if if all the content creators could kind of unionize then it might increase that's
00:52:32
kind of like the yeah like music music industry has ASAP socialist I love it well I'm not I'm not advocating for this
00:52:39
but the point I'm making is if they can't unionize then there's a lot there's just a huge number of vendors of
00:52:46
content and so models will need to buy some but as long as they can get some they don't need to have all and
00:52:52
therefore it's basically highly competitive among suppliers and there's a very limited number of buyers so that
00:52:59
tends this is why the news industry should have always had a federation because they could have just said to
00:53:03
Google hey we're going to de-index ourselves from the Google search engine and so you won't have the New York Times
00:53:08
Washington Post LA Times you're just not going to have any of us unless you give
00:53:11
us X Y and Z and they were just too stupid and not coordinated to do it yeah music industry the exact opposite you
00:53:19
you try to do anything with the music industry they're going to come down on you like a ton of bricks to this day I
00:53:24
see what that's actually really interesting yeah if all the um old school Legacy
00:53:30
newspapers and magazines magazin so on basically formed their own whatever Cel trade Association yeah that would have
00:53:39
been powerful yeah I mean that's what Murdoch wanted Murdoch saw it clearly he was like you know Google's the enemy
00:53:45
here they're going to take all of our revenue and they're going to get all our customer names and we're not going to
00:53:48
even know the names of our customers all right issue three clowner crushes customer queries with AI you may have
00:53:53
seen this trending on X and Twitter and in the Press if you don't know Clara they're a Swedish ftech company they do
00:53:59
that buy now pay later stuff I think they were The Originators of that online and they put out a press release with
00:54:04
some really eye popping claims AI assistants are now doing the work of 700 full-time agents at CL they moved issue
00:54:14
resolving times from 11 minutes with humans to 2 minutes with AI and customer satisfaction is on par
00:54:22
with human agents and it said it's resol ions are more accurate than humans creating a 25%
00:54:27
drop in repeat inquiries that tracks and so far their AI which they built with open AI has had 2.3 million
00:54:35
conversations accounting for 2third of cler's customer support service chats cler estimates its AI agent will drive
00:54:43
wait for it a $40 million increase in profits this year we could talk a little bit more about cloner and their
00:54:50
valuations but freeberg what what do you think this means if we're in year this is the start of year two of chat GPT as
00:54:58
like a phenomenon let's say what do we think year three looks like I think the tech the techn pessimist point of view
00:55:05
is oh my God look at all these jobs that are getting lost I think the optimistic
00:55:08
point of view is that that company has all of that excess Capital now to reinvest in doing other things that
00:55:15
Capital doesn't just become profits that flow out the door and everyone's done with that money and that money just gets
00:55:21
put away in a sock that money gets reinvested and that money gets reinvested invested in higher order
00:55:26
functioning work and that's really where there's an opportunity to move the workforce overall forward which is what
00:55:32
I think is super exciting and um I'm super positive about we've seen this in every technological Evolution that's
00:55:39
happened in human history from you know the plow in agriculture to automobiles to Computing and to now ai that you know
00:55:47
humans moved from manual labor to knowledge work to now ideally and hopefully more creative work and so I do
00:55:55
think that it isn't just about eliminating jobs and making more money but it's about enabling the creation of
00:56:01
entirely new class of work whether that's prompt engineering or you know building
00:56:06
entirely new businesses that simply can't exist today or perhaps even downscaling businesses where you no
00:56:11
longer need to have a 10,000 person organization smaller organizations can be stood up as startups to start to
00:56:17
replace large functioning organizations so I don't know I think it's a time of great opportunity I know that some
00:56:21
people would view it as being highly shocking I think it's inevitable that knowledge labor where the job of the
00:56:26
human is simply the ingestion of data and then communicate an output of data seems like it will eventually be
00:56:32
replaced by Computing somehow and this is happening now in an accelerated way with these llms so I think that what we
00:56:38
should focus on and think about is what are all the new businesses all the new jobs all the new opportunities that are
00:56:44
just couldn't have existed 10 years ago that are now emerging that are very exciting as the workforce transitions
00:56:50
sacked You by this techno utopian view of this all these jobs that are going to obviously be retired are going to open
00:56:58
up the opportunity for these humans to do even better work at clana or do you think it's just going to go straight to
00:57:04
the bottom line well it sounds like they're able to eliminate a lot of Frontline customer
00:57:10
support roles by using AI which is what I would expect yeah I think this is a very natural application for AI you know
00:57:19
it was already the case that you could pretty much find answers to questions by searching the FAQ things like this this
00:57:25
this is an even better way of doing that so so look I I believe that this will be
00:57:30
a big area for AI is saving on again I use the word Frontline customer support because the way that customer support is
00:57:37
typically organized is there's level one level two level three the more difficult
00:57:43
queries or cases get escalated up the chain depending on how hard they are and I think the AI will do a really good job
00:57:49
eliminating level one it'll start to eating to level two but you're probably going to
00:57:53
need humans to deal with the more uh complex cases now the question is where do those displaced humans go I think
00:58:01
there's going to be new jobs new work that's always been the history of technological progress and one of the
00:58:07
things you're already seeing is there's a whole bunch of new AI companies that are exploiting this technology and they
00:58:12
need to hire people so I basically agree with freeberg that you will Elevate people's work by automating away the
00:58:19
less interesting parts of people's jobs and then creating more productivity and more opportunity by the just to use your
00:58:25
just to use your example sacks so imagine if all the level one support people some chunk of them can now do
00:58:31
level two support and so the customers are going to get greater Hands-On care more customers will get access to a
00:58:37
higher level of service the organization can afford to do that they'll be more competitive in the marketplace because
00:58:43
customers feel better taken care of I just think that's how the organizations get leveled up as new technology kind of
00:58:49
shows up like this it's a great point and then those folks can have a much deeper level of interaction with with
00:58:54
their customers than they are todaying question gets more complex and then people might get better at the
00:59:00
software and they might discover new features that you might be able to redeploy those people if you look at
00:59:05
coffee like I don't know was it 40 years ago you went to order a cup of coffee it
00:59:09
was decaf or regular coffee milk and sugar those are like those are your four choices and and now you go order coffee
00:59:16
I don't know if you guys have used the Starbucks app or the you know I just had the sweet green CEO on the Pod and man
00:59:22
you can the Fidelity and and the Nuance of what you want to order is absurd Jam where do you think this is all heading
00:59:30
because there is the issue of displacement how people how quickly people can re be redeployed and if we're
00:59:35
seeing in year two customer support and developers getting 10x what other categories do you think we're going to
00:59:42
see fall next I think the truth is that as you said the real world applicability
00:59:47
to AI was not last year so I think we're really in the first five or six weeks of
00:59:53
the first year so you consider that year zero that's year zero that was sort of like the you
00:59:59
know where everybody was running around building toy apps Pro of concept this is
01:00:03
one of the first few times where you're seeing something in production where there's measurable economic value and
01:00:09
the important thing to note about that is that it's not just what it means for Clara but what it means for everybody
01:00:13
else so if you look at everybody else for example here's Teleperformance which is a a French company that runs call
01:00:20
centers they lost $1.7 billion of market cap when that tweak went out about 20% of their market cap so this is this is
01:00:28
the real practical implication Yes Clara replaced 700 people and they saved 40 million of Opex but teleport 4 months
01:00:38
while they were just doing their everyday work lost 1.8 billion of their market cap at the exact same moment and
01:00:45
so what does it mean I think that what Clara should do is open source what they've built and
01:00:54
the reason is that you want to give companies like Teleperformance a chance to retool themselves with the best
01:01:03
possible technology so they they can actually preserve as many of the jobs as possible because at the limit if every
01:01:10
single company is able to implement something that is as economically efficient as what Clara did
01:01:15
Teleperformance doesn't exist and there's 10 billion and 335,000 employees that will not have a job Y and so for
01:01:24
Clara the reason to open source it is twofold one is they don't lose anything because you will still need to train it
01:01:31
on your own data and so there's no disadvantage that Clara will have right they're just saying look I built this on
01:01:37
top of GPT here's what it looks like and that production code can be used by anybody else go for it but it has to run
01:01:44
on your own data that's a very reasonable thing so I think it has the benefit of both a setting a technical
01:01:50
Pace that can help them attract better employees and more highly qualified people who find the scope of work even
01:01:56
more interesting and B I think it's on the right side of history with all this AI stuff where it's allowing everybody
01:02:03
to sort of benefit in a way that is the least destructive but I just wanted to show you that the destruction was quite
01:02:09
quick and it was pretty severe and if two or three other big companies launch these kinds of tweets after real
01:02:15
measurable results Teleperformance will be a 1 billion doll company short in short order there's a third reason I I
01:02:21
think it's a brilliant idea for clar to open source this this tool because it's not their business right this is just
01:02:27
something they did as a as a productivity Improvement they get the benefit back to them of the community
01:02:32
working to advance that technology so they don't have to put more Engineers like advancing the ball on their
01:02:38
customer support AI totally they can just re remerge in the changes that the open source Community comes up with and
01:02:44
since they're not in the business of selling AI directly there's no reason not to do it like jamas said so I think
01:02:50
it's kind of brilliant yeah this is meta strategy by the way I mean z should be building this on meta's Open Source
01:02:56
Products and Apple's Open Source Products right SX yeah well so what what meta said what Zuck said on the last
01:03:02
meta call is the reason we open source everything is because we don't directly sell AI we create products that AI makes
01:03:09
better so by open sourcing this we allow the community to advance the ball and we
01:03:14
get to reincorporate those changes so it's a very smart strategy for companies that aren't directly selling the AI now
01:03:21
you know if you're if you're like Brett Taylor's new company Sierra obviously you're not going to open source it
01:03:26
because your whole business model is to create a proprietary solution yeah but then to chat's you know 8090 whatever he
01:03:34
was talking about the with his you know incubator concept and you know these things it's a company it's a company
01:03:40
okay sorry um is it 8090 or 980 I'm sorry 8090 yeah 8090 is there a third word that comes off of it are just G to
01:03:46
call it 890 just 8090 got it okay 80% of the features at a 90% discount so back to that like what does this mean if
01:03:53
these things are getting to freeberg your point about the pace man the pace of these things is it going to improve
01:03:59
10% a year or 10% a month if it's improving 10% a month we're going to get to 98% of queries done this year if it's
01:04:07
doing 10% a year okay we're going to get to 99 or 98% of queries in four years in
01:04:11
other words this is happening folks and it's happening at a blistering Pace by the way think of it's so not just
01:04:18
Teleperformance which is a it was A10 billion now a $ 7.58 billion USD company but think about zenes right zenes was a
01:04:27
I think A8 to1 billion yeah it was A10 billion take private in the hands of PE the entire
01:04:35
zenes workflow could be replaced by a handful of these open source agents where all of a sudden people can
01:04:41
eliminate a lot of opacs I think the I think the thing to keep in mind here is where the world is going has always been
01:04:48
to try to lower cost and the original foundational principle of SAS was that there's these line items in on-prem
01:04:56
software that are just extremely expensive over time very hard to justify right and so when people moved to SAS
01:05:03
from on Prem they were looking for cost savings that was the initial thing now yes it's actually not cheaper anymore
01:05:09
but it's much more featur so you get a lot more value in SAS etc etc but the point of these AI agents and Bots and
01:05:16
workflows is that it'll reintroduce the concept of cash of of cost Savings of this idea that you can have cheaper
01:05:24
faster and better and the more that that stuff is open source my gosh I think it
01:05:29
just makes it very hard for companies that have Point products to survive yeah we I was talking to a friend of mine
01:05:35
Josh MOA who was the city head of uber in New York and he launched his own note-taking app he's writing it
01:05:43
himself and he's obsessed with this concept of building a billion dollar a unicorn company with one employee and
01:05:48
this is something that a lot of people have been talking about my friend Phil Kaplan from DH kid built a very large
01:05:54
business in District a unicorn with a very like I think low single digit number of people this could be the
01:05:59
future of you know efficiency you you could build if you catch fire with a really hot company that gets a million
01:06:05
custom absolutely the future it's absolutely the future and then if you think about our jobs in terms of capital
01:06:09
allocation well how much Capital does that founder need do they need to dilute 10% 20% 30% they're not going to need to
01:06:15
delute 60 70 80% a oneperson company should be able to spend less than a few hundred grand to get to product Market
01:06:21
fit in the next few years yeah I mean that's that's kind of what we're seeing just to go back to your question J like
01:06:28
where does this go next yeah please I think it's it's it's really interesting to speculate about that so what Clara
01:06:34
seems to be talking about are email-based customer support cases I think we're this going to go next is to
01:06:41
phone 100% and these call centers use what they're called ivrs these interactive voice response systems but
01:06:48
they're very rigid it's a lot of pre-recorded messages and it says push one if your problem is this push two if
01:06:53
your problem is this everyone hates those things yeah I think where it goes next is you'll call up the call center
01:06:59
and you'll get a a voice that sounds like a human just talk to you yeah and you won't even necess realize that
01:07:05
you're talking to an AI because they're already these AI companies that can do generative voices any language any
01:07:12
accent oh and they're fast now and they're fast and multi language think about that just localizing them across
01:07:18
the globe you know you want to launch your product in Japan by the way did you guys see there was a meta demo which I
01:07:23
thought was really cool which was it was run on Lama 70b but it was a real-time translation tool where the person was
01:07:30
speaking in hen Chinese and the and the other person was speaking in English and
01:07:34
they were able to understand each other but saxs to your point when that person calls for example B OFA now Spanish is
01:07:42
not a language it's actually many dialects and many many many different accents right depending on which country
01:07:48
you're from like the accent that you hear in Spain is totally different than the accent in El Salvador the accent in
01:07:54
Chile or Argentina and so wouldn't it be amazing where you you call your bva app
01:07:59
it picks up your accent and your tonality and it responds with the person of that exact same accent and tonality
01:08:06
that is incredible well I'll take it to another level you call it recognizes your number it knows what you're doing
01:08:12
in the software it knows the problem you've had it knows the last three times you called and how long you've been
01:08:17
using the software and it anticipates like okay I know this person has a Windows machine and it's you know still
01:08:23
5 years old and it's like are you still using that same 5-year-old Windows machine yeah we know it's a bug with you
01:08:28
know Windows whatever uh you should probably upgrade I mean it's going to know the entire context of this and so
01:08:33
it's going to just get it could be more efficient than a human could ever be the
01:08:38
best customer support interaction I had was I called JP Morgan Chase because I had an old credit card that I had had
01:08:44
for 20 years and I think that they had outsourced it to somewhere in the Caribbean and this woman picked up the
01:08:50
phone it was so cool jamat what is the problem man and I was like oh this is the best and I had this whole
01:08:56
conversation with her for 15 20 minutes nothing to do with the phone uh nothing to do with my credit card rather it was
01:09:02
great she's like I man you want to cancel your card I was like yes please no but you could do celebrities best
01:09:08
yeah celebrity voices yeah J Cal do the djt okay let's let's role play uh you're
01:09:13
Donald Trump okay and I and I am JP Morgan hi Mr President chath you're huge you've got big spending I every time I
01:09:22
go see chath go see yourth I go in amazing Mr Mr President I would like to cancel my credit card sir okay you don't
01:09:29
want to cancel it we've got a great a 3.9% but for jamas you know what he's my Sri Lankan friend how much do we love
01:09:37
Sri Lankans okay huge and not jcal nasty nasty man jcal very nasty he acts like a dog a dog
01:09:48
okay he's a dog some people say TDS how much do we love our sex Saxy poop great Mar Lago I go to Sax's house
01:09:58
his house huge huge house almost as big as Maraga not quite I gotta work on it because I haven't
01:10:07
done it for years that's really good it's really good have you seen that guy there's a there's a new like Impressions
01:10:13
guy who's amazing he does seen him he's on Howard Stern all the time right he does the Howard no not Shan Gillis this
01:10:20
is another kid Matt freed friend yeah yeah yeah does Howard back to Howard it's wa that video is good too hold on
01:10:31
Donald on other Trump Impressions so like I'll tell you what Zack there are so many people that try to do it you see
01:10:38
the failed Alec baldwi Baldwin comes out on SNL which used to be a lot funnier back in the day with chff but Baldwin
01:10:45
goes like this he goes we've got a great show I don't do that I never say boy boo
01:10:53
this St coar he comes out like I know a do do dot a DOT do a dot dot dot a China
01:11:00
I don't say the D and the last one is Jimmy failing fell and he touched my hair like a dog and Jimmy fa and he goes
01:11:09
okay look like dog we're rolling okay I don't do these moves they're all wrong so that's
01:11:16
a incredible can let's see him doing uh how got see him do Howard Stern it's Next Level the other one he does that's
01:11:25
incredible but subtle is Stanley Tucci how are you by the way Robin you look beautiful right thank you is this still
01:11:32
you up Howard it's me up right yeah right because when I start talking to Robin after you leave I get crazy you
01:11:39
know right can we just talk about this can we talk right or wrong right right right
01:11:47
right Howard left or right right uh right right right right right right right on red right right right right
01:11:57
it's so great he is killing it I think we have our halftime act for the all in Summit 24 oh for sure oh my God yeah
01:12:03
he'd be amazing I love Stanley Tucci by the way he's great oh you got to see this kid Stanley
01:12:09
Tui have have you guys watched the Stanley Tui on HBO where he cooking yeah where like he goes to Italy it's so
01:12:17
great it's so great he does it on Tik Tok too he just randomly cooks something and he's like I'm going to make a
01:12:22
frittata I have some over Yi and I'm just going to I do watch him on Tik Tok Thanksgiving is a special time I'm in
01:12:29
London celebrating so today I'm making St stuffing and also Sugo de car or as it's known gravy now the thing about
01:12:40
stuffing is you might use a traditional white bread but in my household I use a homemade fatea because I'm Italian on
01:12:49
both sides and nothing tastes better than bread with a little M SAU it's a time for gratitude and giving
01:12:56
thanks enjoy your holidays and thank you very much Felicity good that is really good he's so good he's so good all right
01:13:03
listen I don't know how I keep the show going here but three two okay Issue four
01:13:07
reddits S1 is kind of fun let's break it down everybody 2023 Revenue 84 million up 21% year-over-year they're still
01:13:15
losing money uh net loss 91 million in 2023 they lost 159 million in 2022 so they're cutting the loss free cash flows
01:13:22
negative here's a chart of their quarterly revenue and cash flow so they're kind of
01:13:27
bouncing along the break even Mark as you can see there in the chart they got a wonderful gross margin because they
01:13:33
don't pay to produce the content unlike the New York Times or Netflix and so an 86% gross margin up 2% year-over-year
01:13:42
their daily active unique 76 million up 27% year-over year average revenue per user is incredibly low three bucks and
01:13:50
42 and their daily active unique users uh here you can see another chart growing nicely quarter over quarter they
01:13:59
got a billion two in cash most interesting probably uh and could be challenging to execute on is their
01:14:07
direct share program they're going to carve out a bunch of shares in the IPO to sell to their most active mods those
01:14:14
are the moderators the people who run the different channels or subreddits as they're called what could go wrong what
01:14:20
could go wrong wrong we've seen this movie before and uh but they're going to invite you
01:14:24
since to participate on it on a rolling basis more unqualified retail buyer pool does not
01:14:32
exist in the Reddit mods except maybe the Reddit participants W Street bat knows what they're doing yeah I think
01:14:37
they'll buy the uh I think they'll buy after the lock up comes up but let's just get started here I think you looked
01:14:45
at the S1 a little bit freeberg and you had asked us to you asked me to put on the docket because you were you were
01:14:52
digging into it anything stick out to you or thoughts on the business overall no I mean I wasn't pulling I just asked
01:15:00
if you guys had read it umy um I hey I think I made a funny joke read it accient get it I I'll let
01:15:12
you guys go on for a minute go ahead go ahead tell me when you're ready it was a good pun too bad it was
01:15:21
accidental no it was it was intend I'll give you credit it was we'll give it to you so I think the the thing about
01:15:26
Reddit if you could pull up the chart with the quarterly average daily active user data this was a business that the
01:15:33
last couple of years everyone was like had flatlined because it was only growing 5% a year in terms of usage and
01:15:40
then all of a sudden in the last two quarters so starting in late summer early fall of 23 so just six months ago
01:15:48
the usage started to climb pretty significantly growing 15 and most in the most recent quarter 27% year-over-year
01:15:55
absent that growth story it's a really challenged business because a business without much growth gets value typically
01:16:02
on a multiple of the cash flow that they're generating and you know there's less upside and all this kind of
01:16:08
optionality goes away that's kind of a key point I I don't know I think like for you to make an investment at a $5
01:16:12
billion valuation here you've really got to believe that the growth continues at
01:16:17
this rate and it doesn't revert back to the mean growth rate of the last couple of years of basically 5 % which is
01:16:23
roughly flatlined the other challenge they have is that their AR only in that kind of $3 range which is like less than
01:16:30
10% of where Facebook is at and the data that Facebook collects on their users gives them the ability to do much better
01:16:37
targeting on ads and therefore monetize Their audience much better than Reddit has been able to do to the order of over
01:16:43
10x and then if you look at the arpo number how much they've been able to grow that metric it's also you know been
01:16:49
a little bit flatlined so this business I think is a real question mark I mean you could argue it's probably worth in
01:16:55
the best best case in the2 to3 billion kind of valuation range and then you have to believe the bull case that the
01:17:02
growth continues or accelerates from here and they have a plan to address the arpo problem they have other paths for
01:17:09
monetizing Their audience than what they're kind of doing today what do you guys think this thing's worth do you buy
01:17:15
it if it goes out at three billion or five billion I think the first question which you nailed that of buysight
01:17:24
investor will ask is what happened in the last two quarters that was different than the last 15
01:17:31
quarters that's going to be a very important question and I think they're going to have to have a very buttoned up
01:17:35
answer for that right and if they can point to very specific repeatable things I think that'll be good the thing that
01:17:44
they this IPO if it goes off in the next four weeks they won't have to wait but if it doesn't get off in the next four
01:17:53
weeks they'll have to update the S1 probably with q1 and so you'll see whether this thing
01:17:59
is a trend or whether it was a one-time thing do you know what it is well the growth in the logged out is probably
01:18:06
largely because typically if you use it on a [Music] phone it tries to force you to use the
01:18:14
app right so that you can be in this logged in experience and if you just turn that off you can get a lot more
01:18:20
logged out because Reddit gets tremendous rank Authority from Google so if you just if you just turn
01:18:27
that off I think that you'll have a lot of logged out customers and that will grow very quickly and so maybe it's a
01:18:32
decision that they'd rather have the Topline number grow than have logged in users grow but the logged in user growth
01:18:37
has still been pretty healthy it's basically doubled in the last three years so but to your point freeberg if
01:18:43
they said or our business is really only these 30 odd million logged in users it
01:18:49
would be worth a lot less than saying 75 million I think it's I think you're right it's kind of like in the mid you
01:18:57
know kind of two three four billion dollish range the the big problem is the arpu because these are not
01:19:04
users that represent sort of Facebook's bread and butter kind of a $40 arpu lives in a good suburb in the United
01:19:14
States and is monetized like crazy I just don't think that's what these users are but you could see that
01:19:21
as a challenge or opportunity because if their RP is only 10% of Facebooks there's a lot of Headroom there to grow
01:19:27
it if those users become economically more valuable RP is actually down 2% year over your sex the issue with this
01:19:35
user base is they're incredibly sophisticated internet users who don't click on ads and are kind of anti- ads
01:19:42
as opposed to you know the the general population on a Facebook or a generic service and it's Anonymous so you don't
01:19:48
know who the user is which is how Facebook has such incredible demographic targeting capabilities there's been a
01:19:53
lot of conspiracy theories longc con theories that came up Sachs that we were talking about on group chat maybe you
01:20:01
could summarize this long game that was played by Sam wman and the allegedly founders of Reddit to
01:20:10
wrestle control of Reddit back from their previous corporate owner kest well this was a post by Yan who is
01:20:18
a former CEO of Reddit that was published back in I think 2015 and he kind of lays out what I think
01:20:26
happened or I mean he says at the end just kidding but if he's the former CEO describing these events he must be
01:20:33
describing something he knows about I would just think but in any event what happened is
01:20:39
that Reddit was sold for only about $10 million a year after it launched so like
01:20:43
really really small and I think that it kept growing and the founders realized maybe they made a mistake or that this
01:20:50
was actually a bigger property and so they started scheming on how to get Connie Nas to spin it back out and so
01:20:57
Yan lays out the steps they went through they recruited a CEO who they kind of pre-agreed on then they had that CEO
01:21:07
demand options specifically in Reddit from Ki Nast which meant that Ki Nast had to create a separate cap table for
01:21:15
it and then once they had a separate cap table as a subsidiary of Cony Nast then
01:21:21
they could sort of pressure to have like an outside investor bought in for the expertise that just happened to be Sam
01:21:26
Alman and his fund and you know eventually like step by step they worked it to the point
01:21:33
where they got Connie asked to spin off the company and I guess this plan worked now
01:21:39
it should be said that the largest shareholder in Reddit according to the S1 is Ki Nas or Ki Nas parent company so
01:21:46
no one's going to benefit more from this plan if you want to or scheme if you want to call it that than Connie n it
01:21:52
was a smart thing for them to do to spin out Reddit to allow the employees to have options and then I would say to
01:22:00
bring back the founder Steve Huffman as CEO several years ago so great it worked
01:22:05
out for everybody you know who knows if it was all premeditated so if they own 30% and it goes out for five billion
01:22:12
they got 1.5 billion and they paid 10 million for it so that's a Web 2.0 1.0 2.0 no it's it
01:22:23
launched in 2005 2.0 the same time as like web logs Inc and delicious and Flicker and all that that whole cohort
01:22:31
of little web apps that used Ajax and other things to you know the web was just getting faster and there were a lot
01:22:39
of users well anyway congratulations it was it was smart for Cony Nas to do the spin out and give up 70% in order for to
01:22:47
have 30% of what's going to be a multi-billion dollar IPO great outcome for everybody all right issue five five
01:22:53
Apple doesn't have a fast car project Titan is DBA dead before arrival apple as you know has been working on an
01:23:00
electric vehicle for a decade self-driving as well they've invested billions in the project according to
01:23:06
Bloomberg and apple was targeting 100K price point basically going after the model S plaid with FSD company had 2,000
01:23:15
employees working on this project it was called Titan and they had designers from
01:23:19
Austin Martin Lamborghini Porsche according to the report most of the team will be transferred to Apple's
01:23:24
generative AI division we talked a little bit about Maggie their um generative AI image language model
01:23:32
there's going to be some layoffs it's unclear how many what does building a car have to do with building an llm well
01:23:37
they were yeah so they weren't just building a car they were all in on self-driving and not having a steering
01:23:42
wheel so I understand I'm just saying it doesn't make a lot of sense that 2,000 employees that were specialized in
01:23:49
building a car all of a sudden now become the AI team it would be more like the full
01:23:55
self-driving team is probably getting those AI jobs and the rest are probably getting laid off with Incredible
01:24:00
packages but what do you think sax I was always skeptical that Apple was even working on a car it's just a very
01:24:05
different kind of product than anything else they make and so I never really treated it that seriously that they were
01:24:11
going to make a car so I'm not I the surprise to me is not that they canceled this but that the that it was even true
01:24:18
that they're working on it in the first place yeah no it was it there were reports of them having test tracks and
01:24:24
everything it was pretty well established and you could see that people from Tesla and other places had
01:24:29
does it say why they actually killed it no this is just a report and I the speculation is that
01:24:36
they're going all in on AI they just see that as a much better future well I think it's more core to what they do I
01:24:43
mean the car never seemed that core no and uh there were reports and elon's talked about it publicly so we're
01:24:50
not speaking out of school here or anything but that Elon and apple had and Tim Cook
01:24:57
reportedly had talked or there have been overtures that maybe Tesla was going to
01:25:01
get bought and Elon it's been pretty clear that during the model 3 roll out he was considering selling it to Apple
01:25:07
that would have been a smart acquisition if they had done that can you imagine all the Apple showrooms having a Model S
01:25:13
in it or something or a model y I mean boom they would have just sold so many of them and having the Apple operating
01:25:18
system on that display well wait was it before the Model S came out or the model
01:25:23
3 it was model 3 was the time I think they were talking seriously but remember like no one really believed in the model
01:25:28
3 till it came out and started selling like hot cakes the opposite they thought it was going to kill the company right
01:25:35
all right everybody for the Rainman David Sachs chairman dictator chth p hoaa and consultant of science David
01:25:44
freeberg I am the world's greatest moderator and we will see you next time byebye gr
01:25:51
GR we let your winners ride Rainman David and instead we open sourced it to the fans and they've just gone crazy
01:26:03
with it love queen [Music] of Besties my dog taking driveways man oh man meet me we should all just get a room
01:26:24
and just have one big huge orgy cuz they're all this useless it's like this like sexual tension that they just need
01:26:29
to [Music] release we need to get [Music] merch