Speaker 2
hello everybody it is saturday february the 1st 2025 happy february everyone
Words and timings
helloeverybodyitissaturdayfebruarythe1st2025happyfebruaryeveryone
Transcript Viewer
AI's Sputnik Moment
Feb 1, 2025 ยท 2025 #4. Read the transcript grouped by speaker, inspect word-level timecodes, and optionally turn subtitles on for direct video playback
Speaker Labels
Name the speakers
Edit labels for this show, save them in this browser, or download a JSON override for the production folder.
Transcript Playback
AI's Sputnik Moment
Human Transcript
Timed transcript
Blocks are grouped by speaker for readability. Expand a block to inspect word-level timing.
Speaker 2
January already seems a long time ago. A week is a long time in politics, Harold Wilson famously said, although apparently he didn't actually say it like most well-attributed quotations. The person supposed to have said it didn't, but he certainly, we know what Wilson meant, that things change dramatically. And certainly a week is a long time in tech, especially on That Was The Week podcast that I do with Keith. Keith was asking whether Trump's America is now an oligarchy controlling the world. But since then, we've had what Marc Andreessen calls AI's Sputnik moment. And Keith has borrowed the term from Andreessen to define this week's newsletter. So, Keith, a week is a long time in tech, isn't it?
Words and timings
Januaryalreadyseemsalongtimeago.Aweekisalongtimeinpolitics,HaroldWilsonfamouslysaid,althoughapparentlyhedidn'tactuallysayitlikemostwell-attributedquotations.Thepersonsupposedtohavesaiditdidn't,buthecertainly,weknowwhatWilsonmeant,thatthingschangedramatically.Andcertainlyaweekisalongtimeintech,especiallyonThatWasTheWeekpodcastthatIdowithKeith.KeithwasaskingwhetherTrump'sAmericaisnowanoligarchycontrollingtheworld.Butsincethen,we'vehadwhatMarcAndreessencallsAI'sSputnikmoment.AndKeithhasborrowedthetermfromAndreessentodefinethisweek'snewsletter.So,Keith,aweekisalongtimeintech,isn'tit?
Speaker 4
It certainly is. And there are very few moments like this, actually. But they do reoccur through the history of tech. There are always weeks like this. This week probably was inevitable. No one knew why or where it would come from. But basically, the entire business model of AI has now been put into a state where it has to be re-understood because of DeepSeek.
Words and timings
Itcertainlyis.Andthereareveryfewmomentslikethis,actually.Buttheydoreoccurthroughthehistoryoftech.Therearealwaysweekslikethis.Thisweekprobablywasinevitable.Nooneknewwhyorwhereitwouldcomefrom.Butbasically,theentirebusinessmodelofAIhasnowbeenputintoastatewhereithastobere-understoodbecauseofDeepSeek.
Speaker 2
Yeah, one of your... One of your teachers, Lenin, famously said that sometimes a whole century happens in a week. Have we had a century of events in technology in a week, Keith? Is this a paradigmatic moment?
Words and timings
Yeah,oneofyour...Oneofyourteachers,Lenin,famouslysaidthatsometimesawholecenturyhappensinaweek.Havewehadacenturyofeventsintechnologyinaweek,Keith?Isthisaparadigmaticmoment?
Speaker 4
Well, we've certainly handed their ass to them for those who have predicted that AI has peaked. AI clearly has... Who said AI has peaked?
Words and timings
Well,we'vecertainlyhandedtheirasstothemforthosewhohavepredictedthatAIhaspeaked.AIclearlyhas...WhosaidAIhaspeaked?
Speaker 2
Only Gary Marcus, who's made a career at him.
Words and timings
OnlyGaryMarcus,who'smadeacareerathim.
Speaker 4
Yeah, there's a few followers of his that kind of agree with him. And it is true that squeezing more juice out of training large language models, throwing money at GPUs is producing diminishing returns. That's true, although I don't think it's peaked. It's just slowing down.
Words and timings
Yeah,there'safewfollowersofhisthatkindofagreewithhim.Anditistruethatsqueezingmorejuiceoutoftraininglargelanguagemodels,throwingmoneyatGPUsisproducingdiminishingreturns.That'strue,althoughIdon'tthinkit'speaked.It'sjustslowingdown.
Speaker 2
Well, let's be clear, though, on what's happening, because not everyone watching, especially my podcast followers are not as tech savvy as you. What exactly is DeepSea, Keith? We talked about it last week. And actually, you were kind enough to know that I suggested last week that... It could mean a sharp correction for the markets. For the first time in my life, I've ever been right about the markets. But what does it mean? What is DeepSeek and why does it change everything? Yeah.
Words and timings
Well,let'sbeclear,though,onwhat'shappening,becausenoteveryonewatching,especiallymypodcastfollowersarenotastechsavvyasyou.WhatexactlyisDeepSea,Keith?Wetalkedaboutitlastweek.Andactually,youwerekindenoughtoknowthatIsuggestedlastweekthat...Itcouldmeanasharpcorrectionforthemarkets.Forthefirsttimeinmylife,I'veeverbeenrightaboutthemarkets.Butwhatdoesitmean?WhatisDeepSeekandwhydoesitchangeeverything?Yeah.
Speaker 4
Well, so DeepSeek is more than a single thing. Firstly, it's a model called R3, which is...
Words and timings
Well,soDeepSeekismorethanasinglething.Firstly,it'samodelcalledR3,whichis...
Speaker 4
some time ago was made. And R3 is the equivalent of OpenAI's large language model. And it's basically huge and is very good at answering questions and helping you with problems that you want to solve. It's as good as OpenAI. And it was trained on, we now know, roughly one and a half billion dollars worth of
Words and timings
sometimeagowasmade.AndR3istheequivalentofOpenAI'slargelanguagemodel.Andit'sbasicallyhugeandisverygoodatansweringquestionsandhelpingyouwithproblemsthatyouwanttosolve.It'sasgoodasOpenAI.Anditwastrainedon,wenowknow,roughlyoneandahalfbilliondollarsworthof
Speaker 2
of investment which which sounds to most people like a lot but it actually isn't a
Words and timings
ofinvestmentwhichwhichsoundstomostpeoplelikealotbutitactuallyisn'ta
Speaker 4
lot in the in the ai arms racism it's roughly 100 of of what open ai says it needs and and and it's even less than that uh compared to the 500 billion dollar stargate project open ai is now which we talked about last week which is the
Words and timings
lotintheintheaiarmsracismit'sroughly100ofofwhatopenaisaysitneedsandandandit'sevenlessthanthatuhcomparedtothe500billiondollarstargateprojectopenaiisnowwhichwetalkedaboutlastweekwhichisthe
Speaker 2
these strange bedfellows of Donald Trump, Elon, not Elon Musk, he's out of it. OpenAI, Sam Altman, SoftBank, and Larry Ellison, of all people.
Words and timings
thesestrangebedfellowsofDonaldTrump,Elon,notElonMusk,he'soutofit.OpenAI,SamAltman,SoftBank,andLarryEllison,ofallpeople.
Speaker 4
Right. So R3 has been known about for a while, you would have thought that would have, you know, triggered some thinking, but it didn't. And so last week, R1 was released by DeepSeek. And what R1 is, is a reasoning model. That is to say, it doesn't just answer questions, it can reason. And even better, it tells you it's reasoning, it shows you it's reasoning as it's doing it, and then it gives you the answer. And it compares to the 01 and 03 models. We'll talk about 03 in a minute because OpenAI released 03 yesterday well ahead of schedule, partly in reaction to all of this. Probably more than partly, Keith. Yeah. So it's a reasoning model, and it outperforms pretty much every other model by most of the test criteria. And it was trained for about $6 million. as an additional investment by using something called...
Words and timings
Right.SoR3hasbeenknownaboutforawhile,youwouldhavethoughtthatwouldhave,youknow,triggeredsomethinking,butitdidn't.Andsolastweek,R1wasreleasedbyDeepSeek.AndwhatR1is,isareasoningmodel.Thatistosay,itdoesn'tjustanswerquestions,itcanreason.Andevenbetter,ittellsyouit'sreasoning,itshowsyouit'sreasoningasit'sdoingit,andthenitgivesyoutheanswer.Anditcomparestothe01and03models.We'lltalkabout03inaminutebecauseOpenAIreleased03yesterdaywellaheadofschedule,partlyinreactiontoallofthis.Probablymorethanpartly,Keith.Yeah.Soit'sareasoningmodel,anditoutperformsprettymucheveryothermodelbymostofthetestcriteria.Anditwastrainedforabout$6million.asanadditionalinvestmentbyusingsomethingcalled...
Speaker 2
reinforcement learning. It was an interesting piece, Keith. You didn't have it in the newsletter. I forgot to send it to you by Marina Hyde, who's not a tech writer for The Guardian, suggesting that if indeed, as some people suggest, DeepSeek, quote unquote, borrowed a lot of their data, it's appropriate for OpenAI because they borrowed it from all of us. So is there an element of piracy here? Where did they get all this data?
Words and timings
reinforcementlearning.Itwasaninterestingpiece,Keith.Youdidn'thaveitinthenewsletter.IforgottosendittoyoubyMarinaHyde,who'snotatechwriterforTheGuardian,suggestingthatifindeed,assomepeoplesuggest,DeepSeek,quoteunquote,borrowedalotoftheirdata,it'sappropriateforOpenAIbecausetheyborroweditfromallofus.Soisthereanelementofpiracyhere?Wheredidtheygetallthisdata?
Speaker 4
Actually, they don't need the data for R1. What they do is they send questions to OpenAI using its public API and pay. They look at the answers and they learn from the answers and train their own model on OpenAI's answers.
Words and timings
Actually,theydon'tneedthedataforR1.WhattheydoistheysendquestionstoOpenAIusingitspublicAPIandpay.TheylookattheanswersandtheylearnfromtheanswersandtraintheirownmodelonOpenAI'sanswers.
Speaker 2
So OpenAI was, and Musk has always been very critical of this, OpenAI really was open. Some people suggested it's actually more of a closed AI, but in this sense...
Words and timings
SoOpenAIwas,andMuskhasalwaysbeenverycriticalofthis,OpenAIreallywasopen.Somepeoplesuggestedit'sactuallymoreofaclosedAI,butinthissense...
Speaker 2
deep-seek used open ai to build a better model or certainly a better model in economic terms
Words and timings
deep-seekusedopenaitobuildabettermodelorcertainlyabettermodelineconomicterms
Speaker 4
yeah less money so andrew azim azar's listening or watching and just noted that he
Words and timings
yeahlessmoneysoandrewazimazar'slisteningorwatchingandjustnotedthathe
Speaker 2
yeah he uh he he's a smart guy we need to get him on say hi from me hi azim was i didn't see you at the tecra at dld this year but he's he makes the point that as
Words and timings
yeahheuhhehe'sasmartguyweneedtogethimonsayhifrommehiazimwasididn'tseeyouatthetecraatdldthisyearbuthe'shemakesthepointthatas
Speaker 4
far back as the 31st of december he He wrote Exponential View. Yeah, which is excellent.
Words and timings
farbackasthe31stofdecemberheHewroteExponentialView.Yeah,whichisexcellent.
Speaker 2
We need to include him more. We should get him on the show.
Words and timings
Weneedtoincludehimmore.Weshouldgethimontheshow.
Speaker 4
He's from Golders Green, Keith. He's too clever. He'll make us look stupid. Well, that's not hard.
Words and timings
He'sfromGoldersGreen,Keith.He'stooclever.He'llmakeuslookstupid.Well,that'snothard.
Speaker 4
I'm only joking. You are clever, though. Anyway, he's been ahead of the curve and he's been talking about this stuff a lot. But I even think you, Azeem, didn't quite connect the dots between our three. I mean, I'm going to guess you were a little bit taken about last week as well, that you didn't have 20-20 vision. If you put a comment... You're talking to Azeem or me? Azeem.
Words and timings
I'monlyjoking.Youareclever,though.Anyway,he'sbeenaheadofthecurveandhe'sbeentalkingaboutthisstuffalot.ButIeventhinkyou,Azeem,didn'tquiteconnectthedotsbetweenourthree.Imean,I'mgoingtoguessyouwerealittlebittakenaboutlastweekaswell,thatyoudidn'thave20-20vision.Ifyouputacomment...You'retalkingtoAzeemorme?Azeem.
Speaker 2
uh but you andrew speaks back yeah well i you know i'm famous my brand is very much associated with stock market uh prescience uh you you connect also with an interesting piece from steven sinofsky very smart guy who argues that um maybe you don't put it in but you connect it with me deep seek was inevitable is that i mean everything always looks inevitable in retrospect but was there a degree of inevitability about this and now that the Flood gates, Keith, have opened. Are we going to get more and more deep seeks? Is it going to become, as some people suggest, totally commodified, this market?
Words and timings
uhbutyouandrewspeaksbackyeahwelliyouknowi'mfamousmybrandisverymuchassociatedwithstockmarketuhprescienceuhyouyouconnectalsowithaninterestingpiecefromstevensinofskyverysmartguywhoarguesthatummaybeyoudon'tputitinbutyouconnectitwithmedeepseekwasinevitableisthatimeaneverythingalwayslooksinevitableinretrospectbutwasthereadegreeofinevitabilityaboutthisandnowthattheFloodgates,Keith,haveopened.Arewegoingtogetmoreandmoredeepseeks?Isitgoingtobecome,assomepeoplesuggest,totallycommodified,thismarket?
Speaker 4
Yeah. Well, look, Stephen, for those who don't know, was extremely senior at Microsoft running, I think, Windows and the operating system. And this piece is a broad view of history of tech where he makes the point that... putting money behind scaling up and innovation is always the first phase, but there, there is never in tech. There's never been a, um, a timeline that didn't include reaching the end of scaling up and the start of scaling out. And what it means by scaling out is that, um, An innovation becomes commoditized where it can now spread almost for free everywhere.
Words and timings
Yeah.Well,look,Stephen,forthosewhodon'tknow,wasextremelysenioratMicrosoftrunning,Ithink,Windowsandtheoperatingsystem.Andthispieceisabroadviewofhistoryoftechwherehemakesthepointthat...puttingmoneybehindscalingupandinnovationisalwaysthefirstphase,butthere,thereisneverintech.There'sneverbeena,um,atimelinethatdidn'tincludereachingtheendofscalingupandthestartofscalingout.Andwhatitmeansbyscalingoutisthat,um,Aninnovationbecomescommoditizedwhereitcannowspreadalmostforfreeeverywhere.
Speaker 2
So this is more than Moore's law. Moore's law just means exponential, more and more power. But there's a certain moment where the technology becomes essentially productized. Is that fair?
Words and timings
SothisismorethanMoore'slaw.Moore'slawjustmeansexponential,moreandmorepower.Butthere'sacertainmomentwherethetechnologybecomesessentiallyproductized.Isthatfair?
Speaker 4
Productized or applied are both. And I think applied is more appropriate in the context of AI. where every developer can now use the deep C KPI almost for free to build whatever logic or back and forth.
Words and timings
Productizedorappliedareboth.AndIthinkappliedismoreappropriateinthecontextofAI.whereeverydevelopercannowusethedeepCKPIalmostforfreetobuildwhateverlogicorbackandforth.
Speaker 2
I mean, you and I talked about you.com and I had an interview with a guy from you. I don't know what that will mean. You do a good job, though, in your editorial, Keith, noting who it isn't, what this Sputnik moment, what it means for all the players. You say that for Meta, it's a very concerning moment. Why is it such bad news for Meta?
Words and timings
Imean,youandItalkedaboutyou.comandIhadaninterviewwithaguyfromyou.Idon'tknowwhatthatwillmean.Youdoagoodjob,though,inyoureditorial,Keith,notingwhoitisn't,whatthisSputnikmoment,whatitmeansforalltheplayers.YousaythatforMeta,it'saveryconcerningmoment.WhyisitsuchbadnewsforMeta?
Speaker 4
Well, Mark Zuckerberg has, for the longest time, struggled to find a canvas that people use for tech that he owns. He really doesn't like Google owning Android and iOS being owned by Apple.
Words and timings
Well,MarkZuckerberghas,forthelongesttime,struggledtofindacanvasthatpeopleusefortechthatheowns.Hereallydoesn'tlikeGoogleowningAndroidandiOSbeingownedbyApple.
Speaker 4
uh he is you know he's an application on smartphones and he's a website on the internet what he really wants is to be the the first place people go before they access each other and information he wants to be the plumbing he wants to be the plumbing and his play has been to build an open source ai that can be used
Words and timings
uhheisyouknowhe'sanapplicationonsmartphonesandhe'sawebsiteontheinternetwhathereallywantsistobethethefirstplacepeoplegobeforetheyaccesseachotherandinformationhewantstobetheplumbinghewantstobetheplumbingandhisplayhasbeentobuildanopensourceaithatcanbeused
Speaker 2
everywhere for free essentially giving away everything to hurt the other players, OpenAI in particular.
Words and timings
everywhereforfreeessentiallygivingawayeverythingtohurttheotherplayers,OpenAIinparticular.
Speaker 4
Correct. To slow the progress of OpenAI owning everything. And we've talked about this a lot, but it's still a prospect, OpenAI, because it is a consumer-facing wrapper in a chat format. It is now... Making, I think it's something like six to eight billion a year.
Words and timings
Correct.ToslowtheprogressofOpenAIowningeverything.Andwe'vetalkedaboutthisalot,butit'sstillaprospect,OpenAI,becauseitisaconsumer-facingwrapperinachatformat.Itisnow...Making,Ithinkit'ssomethinglikesixtoeightbillionayear.
Speaker 2
I don't know if you've got it in the newsletter. They made an announcement about the numbers. And again, it sounds to me like panic stations at OpenAI are trying to calm their investors and partners down.
Words and timings
Idon'tknowifyou'vegotitinthenewsletter.Theymadeanannouncementaboutthenumbers.Andagain,itsoundstomelikepanicstationsatOpenAIaretryingtocalmtheirinvestorsandpartnersdown.
Speaker 4
Correct. Now, so an open source model like DeepSeek that's better than Llama, obviously is a huge threat. And, you know, DeepSeek have spent maybe a billion and a half dollars, all told, all time. And Metra is spending, this year is going to spend, I think it said, something like 30 to 40 billion this year on CapEx. So, The world has shifted massively, and I think Lama starts to have to question everything about its strategy and its spending.
Words and timings
Correct.Now,soanopensourcemodellikeDeepSeekthat'sbetterthanLlama,obviouslyisahugethreat.And,youknow,DeepSeekhavespentmaybeabillionandahalfdollars,alltold,alltime.AndMetraisspending,thisyearisgoingtospend,Ithinkitsaid,somethinglike30to40billionthisyearonCapEx.So,Theworldhasshiftedmassively,andIthinkLamastartstohavetoquestioneverythingaboutitsstrategyanditsspending.
Speaker 2
So what does it mean for OpenAI? You say it means there is competition. I mean, I always understood OpenAI had competition, Google, Anthropic. You have to assume this isn't good news. I mean, Altman's put a good face on it, suggesting that they need to improve their game and it's a good way to have competition. But I assume it's not good news for OpenAI, is it?
Words and timings
SowhatdoesitmeanforOpenAI?Yousayitmeansthereiscompetition.Imean,IalwaysunderstoodOpenAIhadcompetition,Google,Anthropic.Youhavetoassumethisisn'tgoodnews.Imean,Altman'sputagoodfaceonit,suggestingthattheyneedtoimprovetheirgameandit'sagoodwaytohavecompetition.ButIassumeit'snotgoodnewsforOpenAI,isit?
Speaker 4
You know, Reid Hoffman was asked that question this week. He's got a new book, Super Agency.
Words and timings
Youknow,ReidHoffmanwasaskedthatquestionthisweek.He'sgotanewbook,SuperAgency.
Speaker 2
Yeah, and I've got his co-author on the show next week. So I'm looking forward to having that. Maybe we can talk about that more in next week's issue.
Words and timings
Yeah,andI'vegothisco-authorontheshownextweek.SoI'mlookingforwardtohavingthat.Maybewecantalkaboutthatmoreinnextweek'sissue.
Speaker 4
But Reid made the point, and I think this is true, but it would take someone a bit more technical than me to affirm it for sure, that the whole value chain down to DeepSeq relies on innovation in large language models. The better the large language model, the better the reinforcement learning will be. And the better the distilled versions, which is the smaller versions, will be, which most people can run on their smartphone or on their computer at home. So Reid's point is OpenAI still is the leader, although I think Anthropic, honestly, is better than it in some ways.
Words and timings
ButReidmadethepoint,andIthinkthisistrue,butitwouldtakesomeoneabitmoretechnicalthanmetoaffirmitforsure,thatthewholevaluechaindowntoDeepSeqreliesoninnovationinlargelanguagemodels.Thebetterthelargelanguagemodel,thebetterthereinforcementlearningwillbe.Andthebetterthedistilledversions,whichisthesmallerversions,willbe,whichmostpeoplecanrunontheirsmartphoneorontheircomputerathome.SoReid'spointisOpenAIstillistheleader,althoughIthinkAnthropic,honestly,isbetterthanitinsomeways.
Speaker 2
Yeah, and on your advice, I use Anthropic.
Words and timings
Yeah,andonyouradvice,IuseAnthropic.
Speaker 4
But they're both good. I think if they keep innovating, they will get back the money they're investing. DeepSeek will feed off it because unless they don't have a paid interface, they can't really block DeepSeek and others. Only yesterday, by the way, the Allen Institute released a new model that's better than DeepSeek. They also open sourced it. And more than open sourcing it, they open sourced the data, the weightings, and all of the elements that went into building it.
Words and timings
Butthey'rebothgood.Ithinkiftheykeepinnovating,theywillgetbackthemoneythey'reinvesting.DeepSeekwillfeedoffitbecauseunlesstheydon'thaveapaidinterface,theycan'treallyblockDeepSeekandothers.Onlyyesterday,bytheway,theAllenInstitutereleasedanewmodelthat'sbetterthanDeepSeek.Theyalsoopensourcedit.Andmorethanopensourcingit,theyopensourcedthedata,theweightings,andalloftheelementsthatwentintobuildingit.
Speaker 2
I don't mean to rub your nose in this, Keith, A few months ago, a year ago, you were talking on it, you may not have used the I word, inevitability, but you were suggesting that OpenAI was almost unavoidably going to become a multi-trillion dollar company, and it was way ahead, and there was no way it was going to be stopped. Have you begun to be a little more nervous about that kind of prediction?
Words and timings
Idon'tmeantorubyournoseinthis,Keith,Afewmonthsago,ayearago,youweretalkingonit,youmaynothaveusedtheIword,inevitability,butyouweresuggestingthatOpenAIwasalmostunavoidablygoingtobecomeamulti-trilliondollarcompany,anditwaswayahead,andtherewasnowayitwasgoingtobestopped.Haveyoubeguntobealittlemorenervousaboutthatkindofprediction?
Speaker 4
I actually haven't, because you have to use these tools to know why, but Open AI, broadly speaking, is far ahead of everybody else in building large language models and rolling out reasoning models. And it's just begun to roll out agentic models. It's schedule. It's very primitive, but it does have it. And as you look at the likely trajectory of AI to become, you know, to get to an end point, possibly this year, where we can all create armies of agents that do various tasks for us. OpenAI is in the lead and is owning the revenue. The real winner here will own the revenue. I think the large language model is table stakes. Owning the revenue is about delivering things that people use.
Words and timings
Iactuallyhaven't,becauseyouhavetousethesetoolstoknowwhy,butOpenAI,broadlyspeaking,isfaraheadofeverybodyelseinbuildinglargelanguagemodelsandrollingoutreasoningmodels.Andit'sjustbeguntorolloutagenticmodels.It'sschedule.It'sveryprimitive,butitdoeshaveit.AndasyoulookatthelikelytrajectoryofAItobecome,youknow,togettoanendpoint,possiblythisyear,wherewecanallcreatearmiesofagentsthatdovarioustasksforus.OpenAIisintheleadandisowningtherevenue.Therealwinnerherewillowntherevenue.Ithinkthelargelanguagemodelistablestakes.Owningtherevenueisaboutdeliveringthingsthatpeopleuse.
Speaker 2
That's right. And so two companies that are very good at owning the revenue are Microsoft and Google. Microsoft own the revenue for The pre-internet, Google have earned it for the internet. You're ambivalent on Microsoft. You say for Google, it's more bad news. I mean, you've always been rather bearish on Google. If you look at the stock price, I don't think most people agree with you. Why isn't this good news for Google? I mean, Google is so invested on so many fronts. Surely slowing open AI down is actually good news for Google.
Words and timings
That'sright.AndsotwocompaniesthatareverygoodatowningtherevenueareMicrosoftandGoogle.MicrosoftowntherevenueforThepre-internet,Googlehaveearneditfortheinternet.You'reambivalentonMicrosoft.YousayforGoogle,it'smorebadnews.Imean,you'vealwaysbeenratherbearishonGoogle.Ifyoulookatthestockprice,Idon'tthinkmostpeopleagreewithyou.Whyisn'tthisgoodnewsforGoogle?Imean,Googleissoinvestedonsomanyfronts.SurelyslowingopenAIdownisactuallygoodnewsforGoogle.
Speaker 4
Well, I think if you look at the world through Google's AI team, They're probably in third or fourth place, quality-wise. They're first by some criteria, but... But it's the Champions League, Keith, the model.
Words and timings
Well,IthinkifyoulookattheworldthroughGoogle'sAIteam,They'reprobablyinthirdorfourthplace,quality-wise.They'refirstbysomecriteria,but...Butit'stheChampionsLeague,Keith,themodel.
Speaker 2
If you're in third, you still get in the Champions League. They're probably happy with that. They don't need to win the league because they've got lots of other games going on.
Words and timings
Ifyou'reinthird,youstillgetintheChampionsLeague.They'reprobablyhappywiththat.Theydon'tneedtowintheleaguebecausethey'vegotlotsofothergamesgoingon.
Speaker 4
Yeah, but wait, I said by quality. So you're right on quality. They've done a great job. They're in last place on revenue. Because the only place you use Gemini is if you own an Android phone, it's built in, but it doesn't get a lot of use there.
Words and timings
Yeah,butwait,Isaidbyquality.Soyou'rerightonquality.They'vedoneagreatjob.They'reinlastplaceonrevenue.BecausetheonlyplaceyouuseGeminiisifyouownanAndroidphone,it'sbuiltin,butitdoesn'tgetalotofusethere.
Speaker 2
Last place after who? Microsoft and OpenAI. Who else is even in this league at the moment?
Words and timings
Lastplaceafterwho?MicrosoftandOpenAI.Whoelseiseveninthisleagueatthemoment?
Speaker 4
I would guess that even Amazon's making more money from AI than Google.
Words and timings
IwouldguessthatevenAmazon'smakingmoremoneyfromAIthanGoogle.
Speaker 2
Wow, Keith. I'll have to tell Ruth Porat to Google that. You may have to go in and get them to pull their socks up. So they're making less money from there. Google's making less money from AI. Anyone from Google listen to Keith Teer on this one than Amazon?
Words and timings
Wow,Keith.I'llhavetotellRuthPorattoGooglethat.Youmayhavetogoinandgetthemtopulltheirsocksup.Sothey'remakinglessmoneyfromthere.Google'smakinglessmoneyfromAI.AnyonefromGooglelistentoKeithTeeronthisonethanAmazon?
Speaker 4
Well, Andrew, that's the interesting thing. AI revenue has become a consumer-facing proposition. It isn't a business-facing proposition. Businesses with engineers can build stuff almost for free now using AI without paying anyone. But that isn't true of consumers. And so Google is really, really bad at distributing to consumers. You would think it'd be awesome because of search, but it isn't. And even Microsoft isn't that good.
Words and timings
Well,Andrew,that'stheinterestingthing.AIrevenuehasbecomeaconsumer-facingproposition.Itisn'tabusiness-facingproposition.BusinesseswithengineerscanbuildstuffalmostforfreenowusingAIwithoutpayinganyone.Butthatisn'ttrueofconsumers.AndsoGoogleisreally,reallybadatdistributingtoconsumers.Youwouldthinkit'dbeawesomebecauseofsearch,butitisn't.AndevenMicrosoftisn'tthatgood.
Speaker 2
Microsoft, yeah, you're going to get banned from Google, Keith, for this. What about Microsoft? Is this good or bad news?
Words and timings
Microsoft,yeah,you'regoingtogetbannedfromGoogle,Keith,forthis.WhataboutMicrosoft?Isthisgoodorbadnews?
Speaker 4
A bit of both. Microsoft was the first to put DeepSeek into its cloud service this week. Why? Well, because it's good enough, number one. And number two, because they don't have to pay for it and they do pay OpenAI.
Words and timings
Abitofboth.MicrosoftwasthefirsttoputDeepSeekintoitscloudservicethisweek.Why?Well,becauseit'sgoodenough,numberone.Andnumbertwo,becausetheydon'thavetopayforitandtheydopayOpenAI.
Speaker 2
But it's not real money, it's investment money, isn't it? It's tokens.
Words and timings
Butit'snotrealmoney,it'sinvestmentmoney,isn'tit?It'stokens.
Speaker 4
Well, it's real money for Microsoft because they have to build the cloud servers that service the whole thing without getting paid by OpenAI. So it's real money for Microsoft. So Microsoft basically has the option of detaching its dependence on open AI. It probably won't for some time to come, but by putting it in so fast, it's a signal that Microsoft is agnostic to who owns and runs the model.
Words and timings
Well,it'srealmoneyforMicrosoftbecausetheyhavetobuildthecloudserversthatservicethewholethingwithoutgettingpaidbyOpenAI.Soit'srealmoneyforMicrosoft.SoMicrosoftbasicallyhastheoptionofdetachingitsdependenceonopenAI.Itprobablywon'tforsometimetocome,butbyputtingitinsofast,it'sasignalthatMicrosoftisagnostictowhoownsandrunsthemodel.
Speaker 2
Right, so Microsoft is moving farther and farther away from AI. What about some of the other players? You suggested that Amazon's making money. What about NVIDIA? and Grok Elon Musk, what does it mean?
Words and timings
Right,soMicrosoftismovingfartherandfartherawayfromAI.Whataboutsomeoftheotherplayers?YousuggestedthatAmazon'smakingmoney.WhataboutNVIDIA?andGrokElonMusk,whatdoesitmean?
Speaker 4
Well, I argue that the impact on NVIDIA is massively overstated, that everyone's still going to have to innovate large language models. Everyone's still going to have to have reinforcement learning and RAG, which is basically when you put your own data into an AI. And so I don't see NVIDIA's sales slowing down at all because of this. In fact, they may speed up as the law is saying buy nvidia is that you know we don't give investment advice and i'm even worse than you at public market investing because it's all about timing and sentiment but you know you're not good on timing or
Words and timings
Well,IarguethattheimpactonNVIDIAismassivelyoverstated,thateveryone'sstillgoingtohavetoinnovatelargelanguagemodels.Everyone'sstillgoingtohavetohavereinforcementlearningandRAG,whichisbasicallywhenyouputyourowndataintoanAI.AndsoIdon'tseeNVIDIA'ssalesslowingdownatallbecauseofthis.Infact,theymayspeedupasthelawissayingbuynvidiaisthatyouknowwedon'tgiveinvestmentadviceandi'mevenworsethanyouatpublicmarketinvestingbecauseit'sallabouttimingandsentimentbutyouknowyou'renotgoodontimingor
Speaker 2
sentiment tell janet that terrible grok is grok is interesting i thought grok was
Words and timings
sentimenttelljanetthatterriblegrokisgrokisinterestingithoughtgrokwas
Speaker 4
going to be the deep seek because A, because of Elon being so smart, but B, because of the investment. They've probably been caught with their pants down by DeepSeq. And Grok 3, which is their next version, built on 100,000 NVIDIA GPUs,
Words and timings
goingtobethedeepseekbecauseA,becauseofElonbeingsosmart,butB,becauseoftheinvestment.They'veprobablybeencaughtwiththeirpantsdownbyDeepSeq.AndGrok3,whichistheirnextversion,builton100,000NVIDIAGPUs,
Speaker 4
Where, by the way, one of the learnings from DeepSeek is the communications between GPUs on the network of GPUs is the key to being able to do it. That's where DeepSeek innovated. That's also where Grok has innovated.
Words and timings
Where,bytheway,oneofthelearningsfromDeepSeekisthecommunicationsbetweenGPUsonthenetworkofGPUsisthekeytobeingabletodoit.That'swhereDeepSeekinnovated.That'salsowhereGrokhasinnovated.
Speaker 2
Well, I think some people, Keith, are going to be thrilled with the fact that Elon Musk got caught with his pants down. We will see so much changes in a week. Let's move on to some other stuff. This wasn't the only news of the week. You've got some really interesting essays, one that really caught my attention from The Washington Post, from Vlad Tenev from Robinhood. An investment revolution is coming. The US isn't ready for it. You've been predicting this for years, Keith. It's the foundation of your startup. What does Tenev say here? Why is it so important? And do you agree with him? I'm not sure you're even on the same page quite as him.
Words and timings
Well,Ithinksomepeople,Keith,aregoingtobethrilledwiththefactthatElonMuskgotcaughtwithhispantsdown.Wewillseesomuchchangesinaweek.Let'smoveontosomeotherstuff.Thiswasn'ttheonlynewsoftheweek.You'vegotsomereallyinterestingessays,onethatreallycaughtmyattentionfromTheWashingtonPost,fromVladTenevfromRobinhood.Aninvestmentrevolutioniscoming.TheUSisn'treadyforit.You'vebeenpredictingthisforyears,Keith.It'sthefoundationofyourstartup.WhatdoesTenevsayhere?Whyisitsoimportant?Anddoyouagreewithhim?I'mnotsureyou'reevenonthesamepagequiteashim.
Speaker 4
I don't think I'm exactly on the same page as him. He's really focused on where crypto meets shares. And he's talking about tokenizing assets and being able to trade them. I am a believer in that. For example, it astounds me that we still have mortgages. rather than tokenizing fractional ownership of residential property and having co-owners who can make money as the value goes up. It seems obvious to me that at some point, tokenizing assets and being able to trade them is going to be a thing. And that's really the core of what he writes about. Tokenizing private company stock is his focus. Now, at SignalRank, we securitize private company stock. So we buy private companies. We've now done 24 Series Bs in the time we've been investing since May 23, which we're number two in the world at doing B rounds, which is shocking even to me. You get into the Champions League if you're number two. So basically we're bundling together B rounds and people buy a signal rank index share.
Words and timings
Idon'tthinkI'mexactlyonthesamepageashim.He'sreallyfocusedonwherecryptomeetsshares.Andhe'stalkingabouttokenizingassetsandbeingabletotradethem.Iamabelieverinthat.Forexample,itastoundsmethatwestillhavemortgages.ratherthantokenizingfractionalownershipofresidentialpropertyandhavingco-ownerswhocanmakemoneyasthevaluegoesup.Itseemsobvioustomethatatsomepoint,tokenizingassetsandbeingabletotradethemisgoingtobeathing.Andthat'sreallythecoreofwhathewritesabout.Tokenizingprivatecompanystockishisfocus.Now,atSignalRank,wesecuritizeprivatecompanystock.Sowebuyprivatecompanies.We'venowdone24SeriesBsinthetimewe'vebeeninvestingsinceMay23,whichwe'renumbertwointheworldatdoingBrounds,whichisshockingeventome.YougetintotheChampionsLeagueifyou'renumbertwo.Sobasicallywe'rebundlingtogetherBroundsandpeoplebuyasignalrankindexshare.
Speaker 2
Yeah, but enough signal ranking.
Words and timings
Yeah,butenoughsignalranking.
Speaker 4
You asked the question and you associated it and you're right to do so because he's talking about tokenizing private company stock, meaning normal people who are not accredited investors can own high growth private companies.
Words and timings
Youaskedthequestionandyouassociateditandyou'rerighttodosobecausehe'stalkingabouttokenizingprivatecompanystock,meaningnormalpeoplewhoarenotaccreditedinvestorscanownhighgrowthprivatecompanies.
Speaker 2
Yeah. And I think It's the D word, democratization, always the promise, often disappointed when it comes to tech. So basically, Tenev is saying that you can do it through crypto. But for the guy on the street, the cab driver who fancies a little flutter on some AI startup, they're not going to be able to do it through you. They can do it through tokens.
Words and timings
Yeah.AndIthinkIt'stheDword,democratization,alwaysthepromise,oftendisappointedwhenitcomestotech.Sobasically,Tenevissayingthatyoucandoitthroughcrypto.Butfortheguyonthestreet,thecabdriverwhofanciesalittleflutteronsomeAIstartup,they'renotgoingtobeabletodoitthroughyou.Theycandoitthroughtokens.
Speaker 4
Well, that's where Tenev gets it wrong. Private companies grow according to a law called the power law, which means that almost none of them make money. And individual retail investors, even if they use tokens, are going to be very bad at picking private companies to invest in. And private investors, private companies, by the way, won't want them to because they're subject to all kinds of regulations. So the real trick with private companies is being able to de-risk the investments, which you can do through an index approach, but not through a single company picking approach. And that's the key to it.
Words and timings
Well,that'swhereTenevgetsitwrong.Privatecompaniesgrowaccordingtoalawcalledthepowerlaw,whichmeansthatalmostnoneofthemmakemoney.Andindividualretailinvestors,eveniftheyusetokens,aregoingtobeverybadatpickingprivatecompaniestoinvestin.Andprivateinvestors,privatecompanies,bytheway,won'twantthemtobecausethey'resubjecttoallkindsofregulations.Sotherealtrickwithprivatecompaniesisbeingabletode-risktheinvestments,whichyoucandothroughanindexapproach,butnotthroughasinglecompanypickingapproach.Andthat'sthekeytoit.
Speaker 2
Can't you combine your startup with what Robinhood is doing?
Words and timings
Can'tyoucombineyourstartupwithwhatRobinhoodisdoing?
Speaker 4
Yes, you could.
Words and timings
Yes,youcould.
Speaker 2
We could certainly... Why don't you open up your platform to tokens?
Words and timings
Wecouldcertainly...Whydon'tyouopenupyourplatformtotokens?
Speaker 4
It's a constant conversation whether we should. It is a non-proposition in the US due to SEC rules. You could do it overseas.
Words and timings
It'saconstantconversationwhetherweshould.Itisanon-propositionintheUSduetoSECrules.Youcoulddoitoverseas.
Speaker 2
Interesting. Well, that will be we'll come back to another very interesting piece you link to something quite different from Lydia Paul Green. Something extraordinary is happening all over the world. Immigration is big, but of course, the political force is against immigration. I was particularly struck with Musk, who seems to be have one foot very much on in each camp on this. On the one hand, of course, he's an example of an immigrant and an innovator. On the other hand, he's funding all these anti-immigration parties. Tell us about the Paul Green piece. It was from the New York Times.
Words and timings
Interesting.Well,thatwillbewe'llcomebacktoanotherveryinterestingpieceyoulinktosomethingquitedifferentfromLydiaPaulGreen.Somethingextraordinaryishappeningallovertheworld.Immigrationisbig,butofcourse,thepoliticalforceisagainstimmigration.IwasparticularlystruckwithMusk,whoseemstobehaveonefootverymuchonineachcamponthis.Ontheonehand,ofcourse,he'sanexampleofanimmigrantandaninnovator.Ontheotherhand,he'sfundingalltheseanti-immigrationparties.TellusaboutthePaulGreenpiece.ItwasfromtheNewYorkTimes.
Speaker 4
Yeah, and to be honest, if it hadn't just come out this morning, I would have added Frank Ferreira's piece on his substat where he talks about the need to end immigration to preserve cultures.
Words and timings
Yeah,andtobehonest,ifithadn'tjustcomeoutthismorning,IwouldhaveaddedFrankFerreira'spieceonhissubstatwherehetalksabouttheneedtoendimmigrationtopreservecultures.
Speaker 2
And, you know... Frank Ferreira... An ex-Marxist, now an extreme reactionary. Well, you know, he's... And he's an old teacher of yours, so he's a nice guy, a Spurs fan, but I think his politics are appalling. But anyway, go on.
Words and timings
And,youknow...FrankFerreira...Anex-Marxist,nowanextremereactionary.Well,youknow,he's...Andhe'sanoldteacherofyours,sohe'saniceguy,aSpursfan,butIthinkhispoliticsareappalling.Butanyway,goon.
Speaker 4
Well, so I think this article points to a very important debate. This article is really saying that the populations of the leading economic nations are shrinking and immigration is required just for economic growth. Ferredi makes the point that economic growth comes from innovation and technology and productivity, and that immigration is basically a BS, and that most immigration hasn't been motivated by that.
Words and timings
Well,soIthinkthisarticlepointstoaveryimportantdebate.Thisarticleisreallysayingthatthepopulationsoftheleadingeconomicnationsareshrinkingandimmigrationisrequiredjustforeconomicgrowth.Ferredimakesthepointthateconomicgrowthcomesfrominnovationandtechnologyandproductivity,andthatimmigrationisbasicallyaBS,andthatmostimmigrationhasn'tbeenmotivatedbythat.
Speaker 2
Let's leave Frank out of it. What about Musk? I mean, how can he be on both sides of the debate? On the one hand, he is the model of the global innovator coming from South Africa to the US, traveling around the world. And on the other hand, he's funding these anti-immigration parties. How can you do both?
Words and timings
Let'sleaveFrankoutofit.WhataboutMusk?Imean,howcanhebeonbothsidesofthedebate?Ontheonehand,heisthemodeloftheglobalinnovatorcomingfromSouthAfricatotheUS,travelingaroundtheworld.Andontheotherhand,he'sfundingtheseanti-immigrationparties.Howcanyoudoboth?
Speaker 4
I think what he's against, I think you have to get to the next level of detail below the word immigration. He's against open door borders that are uncontrolled. Which probably everyone, yeah, I was gonna say, probably everyone is. He is for, strongly for, immigration of using H1B visas in the US and equivalents elsewhere of skilled workers coming in to do jobs that they're needed for. And, you know, for example, semiconductors are going to grow over the next few years and there's a million shortage of engineers who can do semiconductor development. So very high skills needed. I think he's wrong on both counts. I think if you look at the US as an example, unskilled immigration, especially of young people, and this is where I disagree with Frank as well, and I agree with the Times article, unskilled immigration to do, for example, agricultural jobs is completely necessary and needed in the short term. And so some of the anti-immigrant sentiment is irrational. To put it mildly, to use a nice word. And I think Trump understands that. Now, Germany is a totally different case. Germany more or less has had open borders. And its culture has been completely transformed by those open borders. And normal Germans are pissed off.
Words and timings
Ithinkwhathe'sagainst,Ithinkyouhavetogettothenextlevelofdetailbelowthewordimmigration.He'sagainstopendoorbordersthatareuncontrolled.Whichprobablyeveryone,yeah,Iwasgonnasay,probablyeveryoneis.Heisfor,stronglyfor,immigrationofusingH1BvisasintheUSandequivalentselsewhereofskilledworkerscomingintodojobsthatthey'reneededfor.And,youknow,forexample,semiconductorsaregoingtogrowoverthenextfewyearsandthere'samillionshortageofengineerswhocandosemiconductordevelopment.Soveryhighskillsneeded.Ithinkhe'swrongonbothcounts.IthinkifyoulookattheUSasanexample,unskilledimmigration,especiallyofyoungpeople,andthisiswhereIdisagreewithFrankaswell,andIagreewiththeTimesarticle,unskilledimmigrationtodo,forexample,agriculturaljobsiscompletelynecessaryandneededintheshortterm.Andsosomeoftheanti-immigrantsentimentisirrational.Toputitmildly,touseaniceword.AndIthinkTrumpunderstandsthat.Now,Germanyisatotallydifferentcase.Germanymoreorlesshashadopenborders.Anditsculturehasbeencompletelytransformedbythoseopenborders.AndnormalGermansarepissedoff.
Speaker 2
I mean, I don't want to get lost in this. I would strongly disagree that Germany has open borders. You sound like some AFD person. Germany doesn't have open borders.
Words and timings
Imean,Idon'twanttogetlostinthis.IwouldstronglydisagreethatGermanyhasopenborders.YousoundlikesomeAFDperson.Germanydoesn'thaveopenborders.
Speaker 4
It totally does. If you look at the numbers of immigrants.
Words and timings
Ittotallydoes.Ifyoulookatthenumbersofimmigrants.
Speaker 2
It's more generous with immigrants. It doesn't have open borders. Not anyone can just go to Germany if they feel like it.
Words and timings
It'smoregenerouswithimmigrants.Itdoesn'thaveopenborders.NotanyonecanjustgotoGermanyiftheyfeellikeit.
Speaker 4
Well, I think you might find that's not true.
Words and timings
Well,Ithinkyoumightfindthat'snottrue.
Speaker 2
Well, let's leave that one.
Words and timings
Well,let'sleavethatone.
Speaker 4
They failed to pass a law yesterday to try to stop it. The CDU put up and it failed. That's another subject. Anyway, let's move on. Wait, wait, wait, because you always avoid the difficult points. If you deny normal people the right to feel like there's too much immigration, that doesn't make you... I don't think you're a racist to say there's too much immigration. You're a racist if you're a racist. But there is a legitimate conversation about immigration and society.
Words and timings
Theyfailedtopassalawyesterdaytotrytostopit.TheCDUputupanditfailed.That'sanothersubject.Anyway,let'smoveon.Wait,wait,wait,becauseyoualwaysavoidthedifficultpoints.Ifyoudenynormalpeopletherighttofeellikethere'stoomuchimmigration,thatdoesn'tmakeyou...Idon'tthinkyou'rearacisttosaythere'stoomuchimmigration.You'rearacistifyou'rearacist.Butthereisalegitimateconversationaboutimmigrationandsociety.
Speaker 2
Yeah, there is, but it doesn't... And this has nothing to do with tech. It doesn't make you necessarily want to vote for the AFD or the... National Front in France or Farage's party in the UK or Trump in America. You can have a sensible policy. And the Democrats and the left anyway are not open borders. That's another issue. But yeah, let's move on to other things. Your favorite, favorite girlfriend, Keith, came up in an interesting FT piece. Lina Khan warns of catastrophic consequences if Trump gives free hand to private equity. What's Khan's warning here? And is she right? I know you generally don't agree with Lina Khan.
Words and timings
Yeah,thereis,butitdoesn't...Andthishasnothingtodowithtech.Itdoesn'tmakeyounecessarilywanttovotefortheAFDorthe...NationalFrontinFranceorFarage'spartyintheUKorTrumpinAmerica.Youcanhaveasensiblepolicy.AndtheDemocratsandtheleftanywayarenotopenborders.That'sanotherissue.Butyeah,let'smoveontootherthings.Yourfavorite,favoritegirlfriend,Keith,cameupinaninterestingFTpiece.LinaKhanwarnsofcatastrophicconsequencesifTrumpgivesfreehandtoprivateequity.What'sKhan'swarninghere?Andissheright?Iknowyougenerallydon'tagreewithLinaKhan.
Speaker 4
You know, I think she's not reading the room when she says things like this. In a week that Deep Seek undermined billions and hundreds of billions of dollars of investment in USAI. She's focused on the danger represented by the American companies, not the need to out-innovate. And it's the wrong focus. It's just a mistake. And it comes across to me as ideological, primarily, not practical at all.
Words and timings
Youknow,Ithinkshe'snotreadingtheroomwhenshesaysthingslikethis.InaweekthatDeepSeekunderminedbillionsandhundredsofbillionsofdollarsofinvestmentinUSAI.She'sfocusedonthedangerrepresentedbytheAmericancompanies,nottheneedtoout-innovate.Andit'sthewrongfocus.It'sjustamistake.Anditcomesacrosstomeasideological,primarily,notpracticalatall.
Speaker 2
yeah well i expected that you're not a big fan one other piece before we get to start up of the week which i thought was particularly interesting was bill gates wants to tax the robots that take your job and some say it could fund ubi universal basic income to replace lost wages gates is a very thoughtful man what do you make of this I think it's pretty closely aligned to Sam Altman's WorldCoin. And Albert Wenger's ideas on a world after capital.
Words and timings
yeahwelliexpectedthatyou'renotabigfanoneotherpiecebeforewegettostartupoftheweekwhichithoughtwasparticularlyinterestingwasbillgateswantstotaxtherobotsthattakeyourjobandsomesayitcouldfundubiuniversalbasicincometoreplacelostwagesgatesisaverythoughtfulmanwhatdoyoumakeofthisIthinkit'sprettycloselyalignedtoSamAltman'sWorldCoin.AndAlbertWenger'sideasonaworldaftercapital.
Speaker 4
Yeah. And as you know, I'm a big fan of the idea that if capitalism can create abundance through automation, you know, it would be a tragedy if the abundance was concentrated in a few rich people's hands. What capitalism can deliver is a true liberated human race if the proceeds of automation, robots in Bill Gates' language, are used to raise everyone up. And I anticipate there's gonna be huge abundance where using classical economic language, value is no longer created by labor because labor is unrequired. And so that abundance has to be distributed. Funnily enough, Marc Andreessen last week disagreed with Sam Altman that the social contract is going to have to change. But then he went on later in the week to talk about the distribution of abundance as well. So I think ultimately everyone has to focus in on that discussion.
Words and timings
Yeah.Andasyouknow,I'mabigfanoftheideathatifcapitalismcancreateabundancethroughautomation,youknow,itwouldbeatragedyiftheabundancewasconcentratedinafewrichpeople'shands.Whatcapitalismcandeliverisatrueliberatedhumanraceiftheproceedsofautomation,robotsinBillGates'language,areusedtoraiseeveryoneup.AndIanticipatethere'sgonnabehugeabundancewhereusingclassicaleconomiclanguage,valueisnolongercreatedbylaborbecauselaborisunrequired.Andsothatabundancehastobedistributed.Funnilyenough,MarcAndreessenlastweekdisagreedwithSamAltmanthatthesocialcontractisgoingtohavetochange.Butthenhewentonlaterintheweektotalkaboutthedistributionofabundanceaswell.SoIthinkultimatelyeveryonehastofocusinonthatdiscussion.
Speaker 2
I have an interesting interview, and I forgot to connect this with you, with Steve Bannon, who is very, very critical, unkind about Andreessen and the rest of Silicon Valley, distancing himself as... one of the MAGA figures from Andreessen and Musk and Gates. And I think that this is the opportunity for people like Gates and maybe even Andreessen and Altman to start thinking about a new politics of taxing the robots, for example, that actually offers an alternative to the MAGA movement, because it's time, it's one thing to come up with theories, but there needs to be some politics here. So I thought that was a particularly Interesting piece.
Words and timings
Ihaveaninterestinginterview,andIforgottoconnectthiswithyou,withSteveBannon,whoisvery,verycritical,unkindaboutAndreessenandtherestofSiliconValley,distancinghimselfas...oneoftheMAGAfiguresfromAndreessenandMuskandGates.AndIthinkthatthisistheopportunityforpeoplelikeGatesandmaybeevenAndreessenandAltmantostartthinkingaboutanewpoliticsoftaxingtherobots,forexample,thatactuallyoffersanalternativetotheMAGAmovement,becauseit'stime,it'sonethingtocomeupwiththeories,butthereneedstobesomepoliticshere.SoIthoughtthatwasaparticularlyInterestingpiece.
Speaker 4
Also interesting there, Andrew, is that that is a point at which Reid Hoffman, Sam Altman, Andreessen and Musk could theoretically converge.
Words and timings
Alsointerestingthere,Andrew,isthatthatisapointatwhichReidHoffman,SamAltman,AndreessenandMuskcouldtheoreticallyconverge.
Speaker 2
Yeah, because they're going to get I think most people assume that at some point Musk and Trump are going to lose interest in each other. So we'll be seeing lots of interesting stuff coming up. Meanwhile, you're Startup of the Week, Keith, is a very interesting company. I've used it in the past. 11 labs. They're valued now at $3.3 billion with a new $180 million Series C funding. Is this an example of AI now becoming sort of consumer-friendly? Yeah, you talked about the importance of revenue. I mean, are 11 labs making money? I've subscribed to them in the past. I don't currently, but I probably should. Yeah, that's what they are first, because not everyone's going to know.
Words and timings
Yeah,becausethey'regoingtogetIthinkmostpeopleassumethatatsomepointMuskandTrumparegoingtoloseinterestineachother.Sowe'llbeseeinglotsofinterestingstuffcomingup.Meanwhile,you'reStartupoftheWeek,Keith,isaveryinterestingcompany.I'veuseditinthepast.11labs.They'revaluednowat$3.3billionwithanew$180millionSeriesCfunding.IsthisanexampleofAInowbecomingsortofconsumer-friendly?Yeah,youtalkedabouttheimportanceofrevenue.Imean,are11labsmakingmoney?I'vesubscribedtotheminthepast.Idon'tcurrently,butIprobablyshould.Yeah,that'swhattheyarefirst,becausenoteveryone'sgoingtoknow.
Speaker 4
Yeah, what they are is they focus on audio and video. So for example, Andrew and I could use Eleven Labs to learn our voices. We already do, Keith.
Words and timings
Yeah,whattheyareistheyfocusonaudioandvideo.Soforexample,AndrewandIcoulduseElevenLabstolearnourvoices.Wealreadydo,Keith.
Speaker 2
This is a complete AI product, right? We could also... You're not looking at the real Keith Teer or Andrew Keane here. Keith is an Eleven Labs produced bot.
Words and timings
ThisisacompleteAIproduct,right?Wecouldalso...You'renotlookingattherealKeithTeerorAndrewKeanehere.KeithisanElevenLabsproducedbot.
Speaker 4
Well, that was my second point. We can actually make avatars of ourselves with lip sync. So you can give it a script, a voice or voices and avatars, and it can create a completely digital version of this podcast or this video blog.
Words and timings
Well,thatwasmysecondpoint.Wecanactuallymakeavatarsofourselveswithlipsync.Soyoucangiveitascript,avoiceorvoicesandavatars,anditcancreateacompletelydigitalversionofthispodcastorthisvideoblog.
Speaker 2
We could do Azim Azhar and then we could just steal everything, all his wisdom and create a bot. Or we could make him say things he doesn't agree with as well. Yeah, well, you do that with me all the time. So is Eleven Labs, we should start playing with it again. Is it really a revolutionary product?
Words and timings
WecoulddoAzimAzharandthenwecouldjuststealeverything,allhiswisdomandcreateabot.Orwecouldmakehimsaythingshedoesn'tagreewithaswell.Yeah,well,youdothatwithmeallthetime.SoisElevenLabs,weshouldstartplayingwithitagain.Isitreallyarevolutionaryproduct?
Speaker 4
It's very good, let's say that. It's leading edge in what it does. There are a couple of competitors emerging now. I think it's a British company, if I'm not mistaken.
Words and timings
It'sverygood,let'ssaythat.It'sleadingedgeinwhatitdoes.Thereareacoupleofcompetitorsemergingnow.Ithinkit'saBritishcompany,ifI'mnotmistaken.
Speaker 2
That's a lot of money, Keith, for an audio startup. $3 billion valuation for an AI audio product?
Words and timings
That'salotofmoney,Keith,foranaudiostartup.$3billionvaluationforanAIaudioproduct?
Speaker 4
Well, if you think DeepSeek spent $1.5 billion, 180 million ends up not being that much. Arguably, 11 Labs has easier training and doesn't need as much capital. But I can tell you this, the UK is really bad at finding the ability to write checks. DeepMind is the good example they sold to Google for just about double what 11 Labs just raised. And 11 Labs and Mistral, for that matter, in France, neither one of them look like they can survive due to the European Inventing an ecosystem.
Words and timings
Well,ifyouthinkDeepSeekspent$1.5billion,180millionendsupnotbeingthatmuch.Arguably,11Labshaseasiertraininganddoesn'tneedasmuchcapital.ButIcantellyouthis,theUKisreallybadatfindingtheabilitytowritechecks.DeepMindisthegoodexampletheysoldtoGoogleforjustaboutdoublewhat11Labsjustraised.And11LabsandMistral,forthatmatter,inFrance,neitheroneofthemlookliketheycansurviveduetotheEuropeanInventinganecosystem.
Speaker 2
Can one of the players, could a Meta or a Google or Microsoft, could they acquire 11 Labs and get away with it now, especially in the Trump regime?
Words and timings
Canoneoftheplayers,couldaMetaoraGoogleorMicrosoft,couldtheyacquire11Labsandgetawaywithitnow,especiallyintheTrumpregime?
Speaker 4
Well, they could definitely get away with it. The question is, if I was 11 Labs, I wouldn't want to sell. I'd attract American capital, which, by the way, is what they did. And I'd effectively become an American startup, even though you started in the UK. It's the only way to get the capital.
Words and timings
Well,theycoulddefinitelygetawaywithit.Thequestionis,ifIwas11Labs,Iwouldn'twanttosell.I'dattractAmericancapital,which,bytheway,iswhattheydid.AndI'deffectivelybecomeanAmericanstartup,eventhoughyoustartedintheUK.It'stheonlywaytogetthecapital.
Speaker 2
Well, I think it's a good product. Interview of the week. Keith is always kind enough to give one of my interviews with my old friend Nicholas Carr. Has a new book out. Superbloom, how technologies of connection tear us apart. Not Bloomer. Carr is a... a tech skeptic, a critic, the original one, everybody else now is following him. Do you like Nicholas Karki? Do you think he's an important... Of course, he wrote the original book on the decline of attention.
Words and timings
Well,Ithinkit'sagoodproduct.Interviewoftheweek.KeithisalwayskindenoughtogiveoneofmyinterviewswithmyoldfriendNicholasCarr.Hasanewbookout.Superbloom,howtechnologiesofconnectiontearusapart.NotBloomer.Carrisa...atechskeptic,acritic,theoriginalone,everybodyelsenowisfollowinghim.DoyoulikeNicholasKarki?Doyouthinkhe'sanimportant...Ofcourse,hewrotetheoriginalbookonthedeclineofattention.
Speaker 4
Yeah. So look, I think he's a throwback. I mean, his point is quite simple, that face-to-face communications can't be replaced by digital. And even though digital scales you to more relationships, it alienates you from the other people and therefore you become less social, not more. At the core, I really disagree with that. I think Zoom and what we're doing here is a great example of the opposite. So I don't really agree, but I can understand why someone from his generation feels... Well, you're from his... You're older than he is. I've spent my whole career in tech learning it and using it, so I'm native to it. He's really looking at it from the point of view of an intellectual who feels undermined by it.
Words and timings
Yeah.Solook,Ithinkhe'sathrowback.Imean,hispointisquitesimple,thatface-to-facecommunicationscan'tbereplacedbydigital.Andeventhoughdigitalscalesyoutomorerelationships,italienatesyoufromtheotherpeopleandthereforeyoubecomelesssocial,notmore.Atthecore,Ireallydisagreewiththat.IthinkZoomandwhatwe'redoinghereisagreatexampleoftheopposite.SoIdon'treallyagree,butIcanunderstandwhysomeonefromhisgenerationfeels...Well,you'refromhis...You'reolderthanheis.I'vespentmywholecareerintechlearningitandusingit,soI'mnativetoit.He'sreallylookingatitfromthepointofviewofanintellectualwhofeelsunderminedbyit.
Speaker 2
Well, another fellow from... To borrow from Keith, Nick Carr's generation is a certain Marc Andreessen. He has the post of the week. DeepSeek R1 is AI's Sputnik moment. He put that up on January 26th. Do you think he's right? Is it a Sputnik moment, Keith?
Words and timings
Well,anotherfellowfrom...ToborrowfromKeith,NickCarr'sgenerationisacertainMarcAndreessen.Hehasthepostoftheweek.DeepSeekR1isAI'sSputnikmoment.HeputthatuponJanuary26th.Doyouthinkhe'sright?IsitaSputnikmoment,Keith?
Speaker 4
In my editorial, I interpret that as meaning the trigger that forces everyone to innovate better. because Sputnik didn't end up being a winner, but it did catalyze JFK.
Words and timings
Inmyeditorial,Iinterpretthatasmeaningthetriggerthatforceseveryonetoinnovatebetter.becauseSputnikdidn'tendupbeingawinner,butitdidcatalyzeJFK.
Speaker 2
Right, so just to be clear, historically, Sputnik was a Russian-Soviet satellite launched in October of 57 that triggered what now we call the Sputnik crisis and the American response through NASA, a massive investment in space technology.
Words and timings
Right,sojusttobeclear,historically,SputnikwasaRussian-SovietsatellitelaunchedinOctoberof57thattriggeredwhatnowwecalltheSputnikcrisisandtheAmericanresponsethroughNASA,amassiveinvestmentinspacetechnology.
Speaker 4
Right. So you could argue that most of the late 60s and 70s wouldn't have happened in the same way were it not for Sputnik because it triggered everything. And we've seen this week with OpenAI launching 03 that they're already competing for the moonshot.
Words and timings
Right.Soyoucouldarguethatmostofthelate60sand70swouldn'thavehappenedinthesamewaywereitnotforSputnikbecauseittriggeredeverything.Andwe'veseenthisweekwithOpenAIlaunching03thatthey'realreadycompetingforthemoonshot.
Speaker 2
Yeah. And to come back to your friend, Lena Khan, there is no the difference from 1957 and the original sputnik moment and the crisis was that the state invested in technology through nasa today there is no nasa option is there realistically i mean even even the democrats Wouldn't be responding to the original crisis with the Sputnik moment by saying that we need to spend trillions of dollars. I mean, Joe Biden might have, but he's history.
Words and timings
Yeah.Andtocomebacktoyourfriend,LenaKhan,thereisnothedifferencefrom1957andtheoriginalsputnikmomentandthecrisiswasthatthestateinvestedintechnologythroughnasatodaythereisnonasaoptionisthererealisticallyimeaneveneventhedemocratsWouldn'tberespondingtotheoriginalcrisiswiththeSputnikmomentbysayingthatweneedtospendtrillionsofdollars.Imean,JoeBidenmighthave,buthe'shistory.
Speaker 4
Look, I think it's interesting that at a time when innovation is crying out for as much capital as it can get its hands on, The contrast between the Stargate announcement, which has state backing.
Words and timings
Look,Ithinkit'sinterestingthatatatimewheninnovationiscryingoutforasmuchcapitalasitcangetitshandson,ThecontrastbetweentheStargateannouncement,whichhasstatebacking.
Speaker 2
500 billion. Let's be clear. We talked about that last week.
Words and timings
500billion.Let'sbeclear.Wetalkedaboutthatlastweek.
Speaker 4
500 billion. The contrast between that and Elon Musk's narrative that the state spends too much is going to catch them with their pants down. You can't have it both ways.
Words and timings
500billion.ThecontrastbetweenthatandElonMusk'snarrativethatthestatespendstoomuchisgoingtocatchthemwiththeirpantsdown.Youcan'thaveitbothways.
Speaker 2
Either you want to create- That's the second metaphor you've used of Musk and pants down, Keith. Do you know something that we don't know?
Words and timings
Eitheryouwanttocreate-That'sthesecondmetaphoryou'veusedofMuskandpantsdown,Keith.Doyouknowsomethingthatwedon'tknow?
Speaker 4
Well, I think he has a blind spot when it comes to the relationship between government and the private sector. And I do think the US government is the world's most powerful agency, if it chooses to be, to make things happen. The fact that it doesn't, as you said, the fact that it doesn't make things happen and that private sector is doing it is a choice. It's not inevitable.
Words and timings
Well,Ithinkhehasablindspotwhenitcomestotherelationshipbetweengovernmentandtheprivatesector.AndIdothinktheUSgovernmentistheworld'smostpowerfulagency,ifitchoosestobe,tomakethingshappen.Thefactthatitdoesn't,asyousaid,thefactthatitdoesn'tmakethingshappenandthatprivatesectorisdoingitisachoice.It'snotinevitable.
Speaker 2
So do you think that they chose the wrong narrative, the Trump people and Musk? They should have said, look, we're going to save trillions of dollars or certainly hundreds of billions of dollars in wasted government bureaucracy, but we're going to put that money into some AI response rather than just giving it back to people with taxes.
Words and timings
Sodoyouthinkthattheychosethewrongnarrative,theTrumppeopleandMusk?Theyshouldhavesaid,look,we'regoingtosavetrillionsofdollarsorcertainlyhundredsofbillionsofdollarsinwastedgovernmentbureaucracy,butwe'regoingtoputthatmoneyintosomeAIresponseratherthanjustgivingitbacktopeoplewithtaxes.
Speaker 4
Yeah, we've talked about it on the show. I think the right narrative is modernize the government, not shrink it. You can do both because modernizing costs less, as DeepSeek has shown. So you can modernize the government, make it better, deliver better, including NASA, by the way. Why isn't NASA as good as SpaceX? Well, because it hasn't been modernized. And, you know, therein lies a big narrative. I generally am not anti-government, but I'm anti-waste. I align with Elon on anti-waste.
Words and timings
Yeah,we'vetalkedaboutitontheshow.Ithinktherightnarrativeismodernizethegovernment,notshrinkit.Youcandobothbecausemodernizingcostsless,asDeepSeekhasshown.Soyoucanmodernizethegovernment,makeitbetter,deliverbetter,includingNASA,bytheway.Whyisn'tNASAasgoodasSpaceX?Well,becauseithasn'tbeenmodernized.And,youknow,thereinliesabignarrative.Igenerallyamnotanti-government,butI'manti-waste.IalignwithElononanti-waste.
Speaker 2
Well, no one's for waste, Keith.
Words and timings
Well,noone'sforwaste,Keith.
Speaker 4
I mean... Well, unfortunately, we are all, you know, pathetically... you know, passive when it comes to waste. We all know there's massive government waste. And right now, Elon wins because he's the only guy calling it out and doing something.
Words and timings
Imean...Well,unfortunately,weareall,youknow,pathetically...youknow,passivewhenitcomestowaste.Weallknowthere'smassivegovernmentwaste.Andrightnow,Elonwinsbecausehe'stheonlyguycallingitoutanddoingsomething.
Speaker 2
Yeah, and there was a good piece from Gillian Tech this week in the FT about automating government and being sympathetic to Musk. We will come back to that. So much happening. That was certainly quite a week in tech, Keith. We will... We will congregate again. We will return on February the 8th, when no doubt many more hundreds of billions of dollars will have been invested in AI. But have a great week, Keith, and we'll see you next week. Thank you so much.
Words and timings
Yeah,andtherewasagoodpiecefromGillianTechthisweekintheFTaboutautomatinggovernmentandbeingsympathetictoMusk.Wewillcomebacktothat.Somuchhappening.Thatwascertainlyquiteaweekintech,Keith.Wewill...Wewillcongregateagain.WewillreturnonFebruarythe8th,whennodoubtmanymorehundredsofbillionsofdollarswillhavebeeninvestedinAI.Buthaveagreatweek,Keith,andwe'llseeyounextweek.Thankyousomuch.