The Coming AI Startup Bust

[@QwertiusMaximus]

When B-list and has-been celebrities star In Super Bowl ads for AI companies, it's time to sell.

[@RameshKumar-kb7hp]

Well said.

[@DavidOrlandBrown]

Potentially the most profound and true market indicator I’ve ever read in a comment on KZbin

[@Squidlark]

Sidenote: I like your username.

[@DavidOrlandBrown]

@@Squidlark da full gov my friend. SELL MY DATA youtubez. haha, thank you 🙏🏼

[@rtothec1234]

[✍️✍️✍️] _When … Larry… David … or … Jim … from the Office… does Superbowl… ads … it’s time…. to … get… out_ Got it. Thanks.

[@jorgeroman2058]

More like an LLM bubble, unluckily now AI is synonym of LLM , we have a lot of AI and machine learning models doing more specific tasks from a few years now, for instance defect detection on chip manufacturing.

[@obsidianjane4413]

This. Even in these comments you see ppl conflating the two. The ML technology is far more important than just chatbots.

[@MisakaMikotoDesu]

It's definitely not just LLMs. Look at self driving, medical AI, landlord price fixing AI, etc. LLMs are in the limelight, but people have been using the "AI" buzzword for anything ML for years now.

[@danharold3087]

We had Rule Based Expertsystems doing useful work in the 80s. The iphone is stuffed with AI systems. Yes it is very sad that people say AI when they should be saying LLM.

[@alainportant6412]

like blockchain was synonym of crypto... youse a smartass nicca ☝🏿☝🏿

[@douglascodes]

Yup, they blew their load making hallucinating chatbots into the face of a really deep scientific field to make a few bucks. They won't get a second chance to make a first impression.

[@AvenEngineer]

One must first pump, before they may dump.

[@DumbledoreMcCracken]

So sayeth the Lord of Capitalism

[@tear728]

The Book Of Altman From pumpeth you were made and to dumpeth you shall return

[@arjavgarg5801]

Yes to dump the load, you gotta first pump

[@Senthiuz]

When the market slumps, you'll find the chumps.

[@Charlie-Em]

Quan Quan Quan Quaan

[@MoneyMacro]

Great video. Completely agree. So many bubble signs. Hardly any leverage. So, probably not a problem on the macroeconomic scale.

[@hackerbrinelam5381]

Wow I love your content but u also follow asianometry's videos too? NIce

[@raifikarj6698]

Hey money & macro can you make video about how gambling in macro economic scale affecting thing. I am from Indonesia and currently there is of epidemic of online gambling that siphon our economy for middle and lower class to other country it makes business not running because there is lost in extra income to gambling. Its even reported there is deflation happening, the theoritical gambling money that get siphoned is as low as 12,5 billion dollar to 40 billion dollar/yearfor perspective our goverment budget was 200 billion dollar so up to 20% of our goverment budget of economy worth is lost in the country. In relation you can connect the important of gatekeeping gambling to your own nation like Singapore it was common story of high ranking and business people in SEA/South East Asian gambling in Singapore it give boost massively to early Singapore economy and there is no leakage of economy because banning of gambling for Singaporean people to go outside of the country.

[@hackerbrinelam5381]

@@raifikarj6698 Wow that show interesting to know, so like macroeconomic effects of gambling as a industry?

[@raifikarj6698]

@@hackerbrinelam5381 yes the inflow and outflow of economic in gambling even in America we can see it in lottery where it was become some form of more tax as to the less intelegence and poor people it show it can extract an economy and became a capital with the amount it pool. It was happening inside but for Indonesia case it going outside particulary these day it was to thailand and vietnam, Singapore still retain the wealthy tho.

[@piccalillipit9211]

Its a problem for the US as many companies are not viable unless the AI BS pays off big time - Google just binned Prabhakar Raghavan. The whole US IT industry rests on AI paying off or line go down

[@kalomboC]

Those 'sophisticated investors' poured money into FTX.....😏

[@lance31415]

Sophisticated investors like Steph Curry in TV ads saying "Do your own research"

[@zyansheep]

on paper, FTX had a great position. Its hard to predict outright fraud lol

[@kalomboC]

Playing video games on a call with the said sophisticated investors is a hell of a red flag to miss.....

[@DomWood]

Masayoshi Son is a _visionary_

[@Compulsive_LARPer]

They got their money back + 18% in accrued interest.

[@shrek769]

"AI" at this point, when used to talk about products based on LLM technologies, oozes the same vibes that "blockchain" and "cryptocurrency" did in 2020-2021 right before everyone realized the emperor had no clothes for 99% of crypto products.

[@fibryrruihf]

Crypto didn't produce multiple companies with hundreds of millions (anthropic), or even billions of dollars (OpenAI) of revenue

[@danharrington9164]

​What about FTX, Mt Gox etc? ​@@fibryrruihf

[@chazzbranigaan9354]

​@fibryrruihf yeah just an asset class with 2 trillion+ in market cap. Did you just say millions 😂

[@clovernacknime6984]

@@fibryrruihf Just because somebody's willing to pay millions for a chatbot with a tendency to make stuff up doesn't mean it's actually worth anything at all, any more than somebody being willing to pay millions for an algorithmically generated picture of a cartoon ape makes it worth anything. It just means that snake oil salesmen found a new well. LLMs are unfit for production use because they're inherently unreliable, just like blockchains are unfit because they don't scale at all. Unfortunately, unlike crypto which was mostly contained by that inherent unscalability LLMs spill over to the Web and spam search results and user generated content sites with automatically generated garbage, which is becoming a real problem.

[@seancatacombs]

The vast majority of B2B SaaS startups were always designed to be bought out for their IP and/or acquihires within X amount of years. Very few have coherent roadmaps towards profitability, or could ever reach profitability in the first place.

[@kebabsaurusrex1601]

Great comment - agree that most companies are looking to be bought out

[@exponentmantissa5598]

IP - the only real measure is patents and very few of these have any of substance. Look at any large successful hi tech company and I guarantee you they will have a formidable patent portfolio. A lot of startups claim they have valuable IP but in reality they have nothing the others dont have.

[@YPeezy]

It’s an effective exit strategy

[@neskey]

i like these simple essays, it's nice to just hear your thoughts on things

[@foobar9220]

The biggest issue about any use of AI in the sense of LLMs is the fact that hallucinations are not just an obscure outlier, it is the direct consequence of how they work. That limits any practical use to applications where looking correct is more important than being correct (like "do my homework for literature class"). Whenever correctness is required, it can only give a hint to humans, and that hint must be severely limited in size or the human will be unable to check it. Considering, for example, that most of the costs associated with software development occur in late stages of development and maintenance, replacing a mostly correct software that people understand with a mostly correct software that nobody understands, will spell disaster. The other big issue is that LLMs can only remix information that is already there. If you doubt that, ask any of them to write a program in an obscure programming language, for a non-standard problem or both at the same time. It is easy to train for programming quick sort in Java because there is already a gazillion of tutorials and examples on the internet. But that again pretty much limits how much additional value it can provide, as one can just copy-paste that from any tutorial

[@qqqqqqqqqqqqqqqqqqqqqqqqqqw]

Pseud

[@Phobos11]

You are never supposed to think the results from the LLM are 100% accurate, that is not the use case. If you think this is the use case, the you’re using it wrong

[@foobar9220]

@@Phobos11 But what are the applications for something that is not correct? Please give some important applications with economic impact. You can do your literature homework, as this often does not have an intrinsically correct solution. You may be able to replace some FAQ on your customer service page by a chatbot, but only for the simple stuff. But you cannot build an airplane/car/... if every part is 95% correct and you know that you cannot rely on it. If I have a question on programming, I can just google it and end up of Stackoverflow most of the time. I have the primary question, the answer and people commenting on that. I know that this might not be 100% correct, but I have a lot more information on judging whether this may be correct or whether this code snipped is even related to my question. LLMs just throw the answer without the context

[@zorintoto1167]

AI hallucinations have been completely defeated just a few months ago , you need to stay up to date

[@foobar9220]

@@zorintoto1167 Not sure which sales guy pitched that but from a technical standpoint it is impossible to prevent hallucinations. LLMs do not have any understanding, they are just a large probability machine for the next words and that makes hallucinations a direct consequence of how they work. For example, they still happily generate me programming snippets that print all numbers whose square is a prime. Something that is impossible by definition

[@2livenoob]

AI replacing humans is like the fantasy of flying cars. We do not yet have flying cars, nor is it safe to do so. What we think the future will be is wildly different that what it ends up.

[@XieRH1988]

The bubble can't burst soon enough. The sheer amount of lengths that companies are going to when it comes to trying to sell AI to the masses like it's the next big thing reeks of the same desperation we saw when companies were touting metaverse, crypto and blockchain

[@Knowbody42]

I mean, crypto as a currency is still kind of useful. Even if things did go off the rails with things like NFTs and all the scams, pump and dumps, etc.

[@oadka]

@@Knowbody42 It's most useful for criminals and circumventing sanctions. I don't know of other use cases where it is meaningful.

[@nonamehere9658]

IoT, don't forget IoT!

[@belstar1128]

yea while i think ai is not that bad its at least doing some cool things they are taking it too far using it for things its not intended for

[@dankprole7884]

Don't forget "voice", the new old UI everyone is using now instead of a keyboard.

[@zangarkhan]

I think we are in a a overall SAAS bubble with a lack of focus on automation, near-shoring, building manufacturing ecosystems, focusing on industry specific logistics/manufacturing.

[@cinemaipswich4636]

Price to Earnings is important. If the PE is 10:1, that means it has already given you 10% return this year. You'll double your money in 10 years. If it's valuation is 200:1 that means Wall Street values it that way, even though no dividend has been paid so far, if ever. It's "fake it till you make it". Personalities are invited to join the Board of Directors. It's like Crypto Influencers selling junk.

[@obsidianjane4413]

AMZN

[@bryce.ferenczi]

CAGR is more important for early stage companies, you'll be doubling your (paper) gains a lot sooner than 10 years. PE is a good measure for giants that are well established and move slowly like GE and ExxonMobil.

[@gus473]

​@@bryce.ferenczi Agree, although I wouldn't use GE as an example of much of anything anymore. The "catchy" CAGR when I was closer to things was about 30%, sustainable.

[@GabrielCazorlaPersson1]

Stocks are valued on their potential future earnings, not current ones. I am invested in a company (not in AI) with P/E of 200, but its revenue is growing 17% year on year or even higher. Then if you consider that earnings=Revenue-costs and the bulk of the costs is depreciating assets (which is just accounting, not real costs) and financing costs, you can have a company with high P/E ratio and still quite cheap.

[@Mordecrox]

​@@GabrielCazorlaPersson1it's all fine and dandy until the assets deprecate themselves regardless of your accountant's opinion, and those can't be replaced because line goes up, production costs go down no matter what, maintenance costs be damned. Right now I'm sitting in a bubbly place with lots of fully deprecated assets that should be replaced but we must use until they crumble to dust even if there's evidence that there will be productivity gain in replacing them with new assets (except windows 11, accountants and legal department still haven't fully realized how to write 11 in the books)

[@TzSik]

My favorite part is how Ai trained on human content spread to spam internet and another generation is trained on that Ai spam

[@yesyes-om1po]

they curate the datasets before training

[@Tom_Quixote]

@@yesyes-om1po They will probably curate it using an AI.

[@yesyes-om1po]

@@Tom_Quixote Its not recommended to use AI to curate datasets for AI, maybe for visual models, but for LLMs its gonna create a feedback loop pretty fast.

[@Boris-r3j5o]

this is "Self-learning AI"

[@andreikersha4060]

it's a ponzi scheme or musical chairs, the bubble goes up until the music stops, the last investors will hold the bag.

[@bryce.ferenczi]

LLMs seem to be a comodity and is being given away for free by many large companies. None of these companies are worth anywhere near what their current valuation is and won't exist in a decade.

[@tzardelasuerte]

Yeah 99% of the people that know anything about AI are talking about llms and mainly chatgpt. People don't even know the amount of other models being used for other industries.

[@DomWood]

commoditisation is exactly the way to look at it - it's basically an attempt to try and repackage the insane glut of computing power we have in all our devices. it's so cheap and available and accessible that it's WAY ahead of what people actually use their devices for in consumer terms - just device manufacturers and investors trying to find a way to package the incredible tech advances that semiconductors have made ahead of consumption.

[@lugebeatzz8747]

​@@DomWood I think the tech industry will always have something to sell. AI or not, the technology industry will always hire people to solve their problems. Tech companies have people in the "future is AI" religion. It's easier to sell the future in this era than it has ever been.

[@tomh9553]

@@DomWoodexactly, chips keep getting better but most people only use their computers for ms office and google chrome, if companies can make these ai features valuable in the eyes of consumers they unlock another decade of consumers who’ll upgrade their phone every year.

[@chazzbranigaan9354]

@DomWood i find this a very interesting premise, but if the processing is done on a data center, how does this solve that? Using an LLM on a 500 dollar device or 5000 dollar device is all the same?

[@rairaur2234]

A year ago when my LinkedIn feed started overflowing with AI talk I thought to myself "welp, everybody and their mother are pushing their product using a single magic-sounding technology... Where did I hear that before?"

[@littleones-yeahh]

well if you're talking about blockchain then we've got another 6-7 years before the ai bubble bursts

[@citizen3000]

If you think this is anything like that, you’re a midwit.

[@mytech6779]

@@littleones-yeahh There wasn't much real investment leveraged on blockchain, it was just being used as a marketing buzzword. Thus it slowly faded with no bubble to burst.

[@Dannyboi91]

"The money might go to robotics" I wish. The AI boom is actually affecting robotics already too. The sudden out of nowhere surge of humanoid robotics is one of it. I've worked in a humanoid robotics startup (Before LLM craze) and I suspect most of the "new money" is just people that really believe that if we make human shaped robots, do VR headset teleoperation and collect data, we will have humanoid chatGPT. I wish it was that simple.

[@mosesracal6758]

Humanoid robotics are honestly too limiting. Its literally the closest we will ever be to being pseudo-gods and the first thing we just want is another version of ourselves? You can literally reinvent the wheel to something else in that space.

[@Nehpets94]

@@mosesracal6758, "We just want another version of ourselves?" Well, yeah, we're the best-designed living organisms. We're the ones who shaped the world around us and left this rock. Just as the human body was able to accomplish all it had, a robot will now automate it. I don't think you're considering what a humanoid robot can accomplish.

[@ambrosebechtel709]

@@Nehpets94To a robot, which can house its sensors and processing equipment wherever it is most convenient, a head is a cumbersome and unwieldy appendage.

[@Xeonerable]

Its mind boggling that so much money floats around for something nobody even knows will actually be any good.

[@imacmill]

I believe 'subscriptions' for AI services are well beyond what the average Jane/Joe is going to be willing to pay. And the privacy issues with AI is off the charts (but sadly something almost nobody will care about).

[@agxryt]

Average joe checking in - OpenAI feels like a bargain to me and I plan to keep it... Until they have to increase it to make a profit

[@NOLNV1]

​@@agxryt How average? Are you in a rental and work as a cashier or nurse etc? Because I have tried but I have no use case whatsoever thus far.

[@imacmill]

@@agxryt Yup, and that increase will come. Brave Browser has their own 'private' AI service (that doesn't track you), and it costs $40US/month. All AI services will end up costing more and more, as AI uses enormous amounts of energy to function, and advertising won't come close to covering it.

[@agxryt]

@@NOLNV1 I work in IT, live in a low-income Canadian province, and I rent. ChatGPT is really good for learning things. It's like having a personal tutor - on everything. I use it to help me learn web development, but also hobby interested like chemistry and pharmacology. I also use Bala when I'm feeling a lil freaky. Good creative writing outlet

[@johnxina1681]

@@agxryt neet

[@BlakeEM]

I'm am a software developer (25+ years experience) that often works in and with AI daily. I've been saying it's a bubble for the last year. As a software developer, the tools are very useful and are becoming better, but will never replace a developer due to fundamental limitations in how they work (they lack any reasoning ability). This overselling of something, that is too complex for most people to understand, is why this is happening and is why I saw it coming early. It's the exact reason that fully self driving cars are always 2-5 years away, the machine learning models do not "reason" and so can not deal with new situations, something that comes up often when doing anything useful. They will replace other simpler and predictable work, like service jobs.

[@cajampa]

Then you are going to be surprised what is coming. It is your limitations that makes you blind to it. You know what they say "It is difficult to get a man to understand something when his salary depends upon his not understanding it."

[@henkondemand]

@@cajampa So... you have no substantiated criticism on his comment. And no, a vague reference to some hypothetical future development doesn't count.

[@digitalsoultech]

You clearly haven't used pythagora then...

[@cajampa]

@@henkondemand Don't you worry it will be very evident sooner than you think

[@Krbydav328]

I can see your cardboard sign even now…

[@breezyx976]

If the bubble pops, maybe graphics cards will become easily affordable again?

[@lewiswood1693]

Nah crypto mining will suck them up again. Graphics cards will never be cheap again

[@phoenixfire6559]

The AI bubble refers to there being too many mouths and not enough slices of pie. But the pie is real and growing i.e. AI will still be here when the bubble bursts. Until a better technology than GPUs and CUDA come along to handle the AI, GPU prices will keep rising.

[@TiBiAstro]

crypto isn't mined on GPUs anymore ​@@lewiswood1693

[@lucasrem]

reezyx976 You invested in weirdo projects ? GPU computing you do ?

[@pf100andahalf]

@@lewiswood1693 There's a chance that it'll be a long time until the next boom. It would be nice for things to settle down and get normal for 10 years. I can dream can't I.

[@thisnotdog]

the hdd boom is an apt comparison, these startups are built on the same logic. there is something strange about SMEs abandoning ship on monumental projects just to start smaller ones though.

[@DumbledoreMcCracken]

Rats leaving the sinking ship

[@PetarMakesNoise]

SMEs leaving big companies and projects are usually people who figured out they have enough skills and connections to start the same thing, but keep the bigger piece of pie for themselves. Usually SMEs are in the company the ones pulling the strongest but often getting very fast neglected as the companies grow. Nothing unusual, it's seen in every industry.

[@vernacular1483]

This channel is the the highest tier of quality content. Outstanding.

[@wt29]

I saw a YT, I think it was Slidebean, that shows a symbiotic connection between the startups and VCs. The hype bubble(s) are a requirement to keep those industries turning over. We have been promised industry revolutions before - crypto, nfts, big data etc - all driven by VCs and in the end reduced to solid, small niches but not revolutionary.

[@SurmaSampo]

Big data is used across pretty much all large enterprises and is essentially the reason why we have cloud. It is now just another part of the mature enterprise architecture and tech stack. It is neither small nor niche.

[@ArawnOfAnnwn]

Neither crypto not nfts were made by VCs. Sure they were both hot air, but VCs weren't responsible for that.

[@wt29]

@@SurmaSampo you seem to forget how it was going to be the answer to productivity gains by unlocking all the hidden value in corporate data. The number of VC funded startups was extreme and now?

[@Terszel]

Yeah its very disheartening especially when starting a company. I learned only start a company when you already have a job, that way you can launch and pay for the uptime costs and focus on making something useful, then you can think about how to monetize it and only when it is profitable, leave your job. Chasing VC money is very soulless, they focus only on hype, not utility. Have spoken with several VCs and its cringe to hear them list the startups they invested in and when I look them up im like who tf is this, and they are out of business 1 year later.

[@nedprin3357]

​@@Terszel your profile description and profile pic is not what i imagined from your comment lmao

[@mauzen_]

"I think we are in a AI bubble" big understatement

[@carloschu7127]

💪 Asianometry is joining the stock market.

[@sirkiz1181]

He didn’t say that. He said AI startup bubble. Very different

[@MichaelBabich]

Do you mean my AI-enable toothbrush priced $500 is a gimmick? 😂

[@LathropLdST]

Reading comprehension, child.

[@usedcolouringbook8798]

I feel that this bubble is more akin to the "Tulip Mania" back in the 1630s, since it's all hype over 1 specific form of machine learning and not helping advance anything but what is related specifically to LLMs.

[@riffsoffov9291]

I thiink the LLM bubble has some similarity to the 19th century Railway Mania in the UK - investors mostly lost money, but it helped get infrastructure built that benefited other industries.

[@chazzbranigaan9354]

How is it similar to tulip mania at all, tulip mania=nfts this isn't really anything like that

[@ChinchillaBONK]

These billionaires just gonna keep making new startups and sell them off and run before it makes an actual MVP. Or milk the money from investors.

[@Zuranthus]

the guys working at the top AI companies are making $500k-$800k, they'll be fine

[@Memepolicedoggo]

He's not talking about the top AI companies, he's talking about all the little startups making glorified chatGPT wrapper apps, and getting paid in stock options

[@obsidianjane4413]

@@Memepolicedoggo No one said capitalism wasn't messy.

[@kabirkumar5815]

Lots are much smaller

[@Zuranthus]

@@Memepolicedoggo if that were the case he wouldn't be telling them to save their money, cause they're not making any

[@musicdev]

Alternative title: Venture Capitalists Still Haven’t Learned Anything

[@Cris022]

VCs know exactly what they are doing. These cycles are practically by design, and every time one happens, they only get richer in the long run.

[@Snp2024]

Alternative comment musicdev don't know what he is commenting

[@michaelnurse9089]

Venture capital has surprisingly been a successful industry. They only have to hit 1 in a 100 to break even and 1 in 10 to buy everyone a yacht.

[@astroNexx]

I’m pretty sure they are pumping the bubble and that they will have their exit at a very profitable place

[@N8O12]

I try to keep up with news surrounding AI and I didn't even know half the companies you mentioned EXISTED. This video made me realize there's way too many and it'll be a good thing when many disappear and only the best remain. An interesting thing about AI is that there's already high quality open source alternatives and soon enough they will catch up so close to the paid models that they will become the better choice for most people. And then there will be free and uncensored AI for everyone.

[@ryanmckenna2047]

I really appreciate your videos and the time you put into them! Most new industries eventually end up with a few conglomerates monopolizing the industry by lobbying the government to enforce protectionist policies masquerading as regulation to prevent new companies from competing whereby new entrants to the industry will simply be bought rather than become competition. It would have been good to consider this angle.

[@sunnohh]

For the money LLMs are not creating any business value, my firm spends hundreds of millions on LLM automation that does not even come close to working

[@samalateja]

Interesting… I am looking to get a job in GenAi , LLMops side. I am a recent grad(Masters) and building RAG apps and upskilling on AKS for productionzing the apps in Azure cloud.But very hard finding an opportunity. Any advise will be very helpful….

[@bloodaid]

LLM’s are at best highly sophieticated automation script generators

[@rafhq8172]

In which field does your company provide products and services? Could you provide some info on what are they trying to solve with llms

[@WallaWallazz]

​@bloodaid and even then, they are quite expensive for mid to low tier quality scripts.

[@everybot-it]

LLMs sort of democratise the IT industry, as someone like me, completely self-taught with huge impostor syndrome, can output stuff faster, particularly shortening bugfixing times.

[@BobSpector-up7lw]

Thanks!

[@Jebbis]

Just had my Q3 all hands and my CEO said many disparaging remarks about AI and it’s ability’s, while this time last year he was raving. The rich are trying to soften the blow to their wallets once this comes down.

[@elijahprobst]

😂

[@WhyInnovate]

Says you! Bubbles are where valuations are not supported by revenues, here corporations are really ramping up spending. So the null hypothesis for your assertion is “that valuations will be supported by future revenues” test that first

[@WhyInnovate]

Totally have to look at the null hypothesis

[@setlonnert]

I was in the middle of two "bubbles". One was in the beginning of the 80s, and what was called "the fifth generation" of computers and software: AI by means of logic programming (Japan) and functional (USA). I was a participant in "the 2:nd logic programming conference". Eventually it all died. The second, was around 1999-2000, I was invited to USA, money was thrown around. But it also died in 2001. I expected the second bubble, which I learned from the first. Now I can also see some signs. But as you also underline: some good things may come out of this. Maybe also some bad ones. Nonetheless it has been a fun ride!

[@rabokarabekian409]

Companies with shares (stock option bonuses) work for gamblers. Employees, facilities, and delivering products, and services all cut into profit. "You got to know when to hold up, know when to fold up."

[@Arcticwhir]

i like it! well balanced and rounded video. I read AI research papers quite often - usually products follows 1 month to 1.5 years after. I say this because the research I currently see shows that there are still many knobs and avenues that will increase AI reasoning, innovation etc.. It might be a funding bubble, but the tech is real and quite useful (at least to me). AI is still relatively quite open - which surprises me, around 80% of the research is out there, meta still releasing many projects and knowledge, stability ai (even though not the best), still release openly, open ai even and claude...

[@ericvincentofinowicz5610]

The last 120 to 60 seconds of this was beautifully written. Well done.

[@stefanschneider3681]

9:42 this might be the most beautiful way to dream for the tech-world, but also for human society overall. And it's legit to hope for the good - whilst not ignoring the realities. You did a fine job in this task 👍

[@ParkingLotStudioGames]

the bubble "forming" man we've been at peak market top vibes for 6+ months easily

[@recklessroges]

The more bubbles that I see burst the less I believe that the innovation is worth the destruction and fallout.

[@shaunashton5434]

Add in government subsidies and tax breaks and yes the game is wasteful and rigged

[@matthewwilliamson6229]

LLMs are a B2B product. They will be found in every software application and seep into all application GUIs. If millions of businesses are paying for LLM access in order to operate their websites, I can easily see how it could be profitable.

[@martin_emrich]

„I just make videos“. I think that makes you one of the more wise people in that sphere right now.

[@WolfmanDude]

I am so tired of all modern "tech". Just give me a better CPU and affordable graphics card. I want a powerful PC, not a terrible knockoff Ipad bolted to my cars dashboard. I want to use a mouse and keyboard, I dont want to talk to my electronics like some lonely weirdo. I dont need this app cloud chatbot trash in my life. Its fake innovation powered by greed and buzzwords

[@NinaYahsika]

You must've been born before 1990

[@mcs131313]

Do you not think these things That you do want have also advanced immensely?

[@fappylp2574]

Voice interfaces are actually a pretty big deal. A lot of younger people can barely read an write and will need ways to interact with software other than written text.

[@nmayfield]

Another great video, you are one of my favourite channels on youtube.

[@dathes]

Sophisticated VC's?? You remember the decision criteria applied by VC's for investing in FTX? 😂

[@gus473]

Substitute "accredited investor" (as interpreted by SEC regulations and case law) for "sophisticated." But yeah, point made! SBF's making new friends, some people got some money back, all's good! 😉😅✌️

[@lol109109]

Thank you as always for the level headed insight and commentary without any clickbait antics.

[@marnig9185]

And all what they make is real warm air.

[@duncanluciak5516]

With more CO2 in it.

[@rubyciide5542]

And reduced water

[@zeg2651]

I am working on an AI startup myself and I think you are right. It's very easy to copy each other. I believe that the ultimately determining factor will not be the product itself, but the business operation behind it

[@ah244895]

Imagine the NVIDIA GPUs you will be able to pickup at fire sales in the future. Play your video games with H100 and H200s...

[@Sven_Dongle]

Those arent graphics GPU's, but they would probably pump out a decent frame rate nonetheless. They have tensor cores and a transformer engine, so most of the $30000+ device would be scratching its ass as you busted caps at zombies. Oh, and you would have to go through remote desktop since it has no outputs, no HDMI, no DP. Rock on!

[@yesyes-om1po]

I don't think the H100 uses normal PCIE interface, even so, thats an 80gb ultrafast inferencing card, its never going to be cheap.

[@WingofTech]

A real, levelheaded and unbiased analysis. I still marvel at how sheltered we are as kids from the markets! :)

[@jimmurphy6095]

AI has already been integral to the winners of the 2024 Nobel in Chemistry for protein folding predictions. That's a good start.

[@EverFlores-w2y]

those are not LLMs

[@jamiehardt3061]

There's a ton of great applications for ML but the Bubble is centered on all the startups that are promising to make chatbots and NLP prompted content creation into consumer products, and none of these applications are economically proven.

[@MarkMcCarty-i8n]

An interesting look at the hype and the reality. About halfway through, I started wondering about the need for all these data centers and voila, you addressed that, too. Nicely done!

[@3800S1]

Whenever there are celebrities involved, I never know who any of them are, which gives me negative interest.

[@foughtthelol]

Are you boasting of your ignorance?

[@throwback19841]

i manage dns security for my employer. we block non-approved gen ai/llm tools. i have to keep updating the list of blocked tools. it was about 70 at the beginning of the year. its over 700 now. that is a bubble. there isnt room in the market for that many players.

[@lostintranslation150]

There's more to this topic unfortunately. MSFT, NVDA revenue roundtripping scheme, where credits are passed off as "investment" NVDA funding Startposition that use their chips, generating "revenue" l. Just as Cisco did back in the internet bubble. SMCI with questions in their finances as pointed out by Hindenburg and then they delayed their 10K. We only scratched the surface yet.

[@elibarbq]

Not sure what you're talking about. Could you please expand on your comment?

[@snage-thesnakemage]

@@elibarbq same here mostly replying so i also get a notification for when they give a more through exlpaination. I get that your are talking about companies using their stock names but what do you mean by only scratching the surface?

[@adeptpeasant6161]

Ai is the new Narc on all our electronics

[@lenOwOo]

And because common people have less money, most people don't want pay for ai

[@sanesanyo]

The issue here is also that their is far too much private capital that VCs have to deploy and AI is the only theme they seem to be investing in, so a lot of money chasing few opportunities has also contributed to this. The bubble will pop as soon as the hyperscalers stop betting the house on data centers and public market valuations will start correcting which then will have an impact on private markets.

[@Standbackforscience]

No real product, no real revenue, I guess.

[@michaelnurse9089]

The have a product. The revenue is lacking. The counter-argument is that Google and Facebook were in this camp for the longest time.

[@ArawnOfAnnwn]

@@michaelnurse9089 What is their product? What does it actually do that's useful? If can't code on its own. It doesn't seem to be replacing lawyers or doctors or even customer service. So far it's mostly just made some kooky chatbots and some admittedly impressive art - that no one seems to want to actually use. So what's their product for then?

[@manysnakes]

@@michaelnurse9089 What did Google and Facebook end up as? Advertising companies. If that's where LLM ends up, these valuations will indeed be fantastical.

[@someperson5252]

​@@ArawnOfAnnwn Who said it needs to replace software engineers, lawyers or doctors? The tools are there to save time in its current form. It helps speed up tasks that usually would've taken a person days down to hours. It's been 2 years since LLMs have become mainstream and we already have people expecting them to replace professions entirely like you do here. That's saying something about the technology. I can count on one hand how many of these startups have claimed that. There's a few startups that have really high valuations that's because they were either first or hae some insane founding team. The rest of the valuations look normal to me. Sure most are going to fail but is that just the usual startup game or is it related to the AI hype? The next generation models haven't even come out for us to judge whether the models are reaching their limits. If you hate LLMs like some developers I've met do then just say that, don't go saying the technology is useless because it hasn't "replaced doctors and lawyers".

[@knoepfler34]

⁠@@someperson5252👍 Overhyped doesn’t mean it’s not useful.

[@vincemccord8093]

"What we're dealing with here is a concept. You only have a problem when you have a real product." From the play, "Control-Alt-Delete"

[@bengmo64]

I'm already sick of ai being jammed into anything with a chip in it

[@tuckerbugeater]

this video was produced by ai

[@RllXeron]

Best yt Channel ever. Thx for this analysis

[@XiaosChannel]

i think you got one thing wrong that is most people just use the AI provided instead of training their own nowadays, given how llm can perform generic workloads

[@Disent0101]

the difference between this boom and every other one, is this will be the last, its a boom to replace human effort and the mad scramble to make some money before we're all redundant is what we are witnessing...

[@makego]

Very thoughtful essay. "They're already rich," as you said. Extra risk is something they can do with their extravagant excess, but mostly they are greedy and trying to get richer.

[@lord_voldemort44]

love how you include that this technology is real. this is not rose petals. this stuff really has the potential to change the world. we have no idea how far it can go

[@thewatersavior]

Robotics feels like the likely end goal - already seeing tools pop up everywhere. Next wave of companies will build the omni-connectors that can talk to any robot

[@PukuBuffet]

the transformers paper really tipped the industry over the edge and launched it into the heavens- robotics need a completely new foundational model for it to fuel the next cycle of the industry. that paper is still being written...unsupervised sensorimotor learning is as though of a problem to crack as any other problem out there. but so was nlp as one point....

[@thewatersavior]

@@PukuBuffet oOo.. yup re nlp at one point - as we shorten the traning feedback loop this should improve. I am guessing that we need really more sensors or a new way of sensing combined with the smart approximation and trial and error that we use to learn. In the words of 2 minute papers.. what a time to be alive!

[@ConversionCenters]

As always, thank you for another well done video analysis. One element is "money availability." With an "AI story" and some "AI people" meetings with capital can be quickly arranged and investment networks energized. It sounds odd, but if money is available the tech folks develop a way of getting it. My point is that this current "fad" is at least partially "banker driven." If your startup's pitch easily goes upstairs to the check writers? You will get money. I agree with your point that competition at this stage of a new market will be a challenge. The "killer app" of past years will offer these startups a target, but, this point is critical....the tech is moving really fast. So fast that it is difficult to define how quickly an AI market solution will be "obsoleted." From following the craze, I don't see an end except that a couple of diversified big players will end up with something as often happens. What they will "end up with" remains a mystery..

[@kabirkumar5815]

These unoriginal ones are 100% in a bubble, but its mostly due to VCs being bad at judgement and the AI space attracting hype beasts like very few others. Stuff like shrek sampler should be worth millions, but the guys making it seem to just be doing it for free behind anon accounts atm. The AI Celebrities thing is due to VCs being dumb and original again. What we will be able to see in the next two years, imo is the first serious AI companies who can actually bootstrap, as learning efficiency increases and the ability to make value increases.

[@OdinReactor]

AI is the new buzzword. Two years ago it was Quantum. 🤣

[@DeepfriedBaby]

There's always a bust in every bubble. Otherwise, it's not a bubble.

[@mururoa7024]

Copilot for programming appeals only to juniors. The code it produces it pretty inadequate. This is because it doesn't actually understand the code, it just cobbles something together from snippets it finds on the web. It fails to understand if there's any quality to it. If you ask it to optimize the same code 3 times, the end result is a complete mess. If you're a seasoned engineer you don't need copilot, as regular auto-complete was already good enough to save on typing. On the contrary, very often copilot stands in the way when you have to constantly cancel what it's trying to do.

[@CakeIsALie99]

Maybe one of the top dog LLM companies can hit the jackpot and become profitable, but I don't see any of the smaller startups that aren't on the cutting edge ever becoming profitable. And if the top dog ever makes it, you can be sure they will be milking it instead of trying to share their profitable product.

[@tear728]

With what products can they do so? Alex Karp at Palantir has said this many times. These companies don't have many pathways to convert LLMs into viable business solutions

[@BienestarMutuo]

@@tear728 Yes, Chinese companies have already won, they are 50 times more cheaper and as the video say are just a few months away from the leader. example deepseek

[@barryobrien1890]

Great perspective as always. The potential of ai and llm seems great but elusive. Current technology is incredibly unreliable but impressive and everyone is seeking the golden key that will unlock the vast potential. This key has proven very difficult to find and the current approach involves a lot of data, energy, hardware, and testing. To date the results are mixed with disappointment in almost every field. Ai is inherently unpredictable and prone to passing off misinformation / creations as fact. The best usecase seems to be Adobe which use ai to speed up picture retouching, with positive if low resolution results, but not quite matching the level of professional retouchers. The worst in my opinion is perplexity that just makes up incorrect results for web searches. Both tools have their uses as aids in speeding up workflow, but fail to replace human input, and fail miserably in the cost to results field, when comparing how much it would cost to train a number of humans to perform a task as it takes to train an llm. For similar investment the humans always win. The key is close but may prove as elusive as the proverbial magic wand, and then wei will need to find another potential bubble

[@ScowlieMeerkat]

Yeah the economy is going to boom so hard if AI actually succeeds at the executives' goal of *checks notes* no longer having to pay workers' salaries.

[@phillipsoltan9913]

AI is one reason that leaders in technology are starting to seriously consider UBI (Universal Basic Income). Companies using AI can wipeout massive numbers of jobs in a short period of time. Without government policy changes, massive job instability could become the new normal.

[@CC21200]

I think I've seen around fifty different KZbin advertisements for scams claiming that they'll use AI to decide what crypto to buy and you'll have free money.

[@michaelnurse9089]

My UBlockOrinin has now blocked 13.9M ads since install two years ago.

[@ajudicator]

As an ai startup founder hearing how non chalantly the video talks about failure where it’s my blood sweat and tears and sleepless nights hits different

[@mikemaldanado6015]

The problem is as computer scientists we know this is not AI. There is a reason why these models need more and more data because unlike us intelligent being that can learn and reason with just a fraction of the data they are being fed is because these are memorization machines, plain and simple. You know how openai is always going on about it's "ai" passing the bar exam, well after all the hoopla a couple scientists game chatgpt the same exact test but just switched around the answers on the multiple choice. Well yeah it failed miserably but research reports don't get much attention outside the comp sci community. I made millions during the tech boom in the early 2000's and am not short every company with the words AI in it because I know what it truly is. We are no closer to real ai than 20-30 years ago - as the great linguist noam chomsky said when he saw chatgpt 2 years ago "it's a parlor trick' - this coming from the father of linguistics, His argument which still stands today "but they have no idea what they are saying" , that is all u need to know really. Real ai will, its at the top of it;s real definition, So I am going short,,,,,,,,,,,,,,,,,,,,

[@jasonhemphill8525]

I can't wait!

[@bobbyc1120]

No way Americans spend just $26 billion a year on retail and restaurants. That's $79 per American.

[@rko2016]

Thats actually a statistical outlier, bill gates actually spends 10 billion on hambugers annually and should not be accounted for

[@austenol7570]

I think it means the commercial side, as in businesses like Walmart and Target spend $26 billion on maintaining and growing retail stores. 9:25

[@VijayPhone-he2ck]

It probably refers to construction of those things

[@kevin-jm3qb]

you are the gold standard of tech. Bloomberg needs you on

[@tangobayus]

Most of the people doing those startups know nothing about what really goes on in businesses. It's not their fault. They've gone from school to startups without spending any time inside a bank, hospital, or manufacturing company.

[@gus473]

Yes, yet they can/hire/rent that expertise. And at some point they will, given enough "investment."

[@benjaminjackson8663]

This is one of the most rational and educated takes I've heard so far. Nice job.

[@JeffBilkins]

What are these crazy valuations even based on? There is too much money and greed just sloshing around.

[@Knowbody42]

Greed, and FOMO.

[@tuckerbugeater]

@@Knowbody42 money printing by the fed. this cash didn't come from nowhere!

[@jamiehardt3061]

Small trading volumes at inflated prices, and the prices are inflated by the expectation that these companies will all take over their markets and be in a position to charge monopoly rents.

[@bannedeverywhere]

>What are these crazy valuations even based on? Read about 'greater fool' theory.

[@qoncept]

Pretty good unstructured thoughts. Love the videos, man. Doubt I'll ever be able to meet for a tea.

[@felixwalton4612]

the telecom bubble peaked at 16 billion of private investment in one year and the internet changed the world; this year they'll be just 6bn in AI investment, we are not even close to the peak

[@juanbertone1941]

Are those 16b adjusted for inflation?

[@wwjccsd]

Except it’s not just private. Microsoft, meta, google, openai are throwing 50+B in AI this year

[@Marina-xg5jv]

20% of index funds are compromised of the tech sector and 40% of market returns in the last 15 years are from the tech sector. Grandma will lose her money.

[@bobweiram6321]

LLM is really an algorithm that depends on content theft. The massive availability of data is what makes it possible.

[@kazedcat]

The Internet is content theft. Ban the internet.

[@dissahc]

it's almost as if they're stealing something of such negligible value that it's not even theft. people assume that because AI needs a lot of data, that all of that data is equally valuable. it's just that separating the wheat from the chaff would be far too tedious. it's never the people who produce worthwhile art that worry about their stuff being copied, because they're used to generating novel ideas. it's always the people who produce art which is highly derivative and uninspired. they don't like AI beating them at their own game. also, if you witnessed the last 2-3 years of AI progress and think "content theft" is even worth worrying about, you're either a fool or a narcissist.

[@freddy4603]

​@@dissahc I mean, they did steal it... My reason for not caring is because I don't like how copyright laws work in the first place.

[@NOLNV1]

​@@dissahc What is "worthwhile art" and how have you come to the conclusion said artists don't care?

[@JohnCollins-vy4nf]

​@@dissahcGood work goyim, keep up the support for the head bankers.

[@zants_]

I love how the topic of AI is just a proxy by which Asianometry is able to provide this grounding perspective and advice for life. :)

[@cartmann94]

Just like the bust of N/F/Ts, my schadenfreude bristles with excitement. Hopefully a lot of these unnecessary ai apps like the one on Google search and the one Amazon uses (Rufus? Isn’t that the fat bloke from Street Fighter 4?) will be discontinued and moved to the wayside.

[@user-pt1kj5uw3b]

Comparing AI/ML and NFTs is ridiculous.

[@Twangaming]

@fullstackcrackerjackthis reads like you REALLY want AI to be valuable but have no examples. You know what’s not valuable? How the internet has been completely fucked with bots and slop content since this stuff hit the market

[@Twangaming]

@fullstackcrackerjack I am a programmer dipshit. And again, you fail to say what the value is

[@NOLNV1]

​@fullstackcrackerjack You are probably overestimating the importance of programmers, also pointlessly insulting others.

[@Durgenheim]

LLM’s are just another entry in a long list of products brought to us by tech companies that solve problems that either don’t exist, or that already have tried-and-true solutions.

[@kevinbarry71]

Absolutely a bubble; and the fact that this keeps happening sheds doubt on the idea that humans are intelligent. The very idea of an AI celebrity is absurd.

[@БеловБорис-у4щ]

It’s not useless, just now one really cares about it or knows how make it useful. Now it’s a stage of “making biggest model” so best minds are focused on it. Then this approach will reach its limits or investors will be not so happy about it best minds will refocus on utilizing models. Then we will see real impact of AI.

[@ChristianDoretti]

Well, Humans are complex, but not what you would call intelligent, humanity and its concepts are indeed a summary of an intelligent civilization.

[@thomasfalcon6350]

Blame the money printing by governments, this is the only thing that fuels bubbles.

[@Onii-chan_2465]

The idea of it being absurd is a social construction which you are aware of but refuse to accept within your worldview. It is the very same social construction which others embrace that makes the idiocy possible.... I dont disagree with you though, it's absurd. There's certain things about us as human being that we are better off leaving a mystery.

[@2CSST2]

Yes it does show that humans are not very intelligent because we have plenty of braindead people like you who think something is absurd right up until it exists, showing their total lack of foresight. People like you used to think the idea of everyone carrying small computers with them was absurd, that would be cellphones now. What about the fact we fly people all around the world? You can speak almost instantly to someone on the other side of the world. If something is physically possible and technologically within reach, it will happen. Yet with all this the most obvious thing that WILL definitely happen like an AI celebrity is apparently beyond your puny tiny little imagination. Jesus this is sad.

[@urisinger3412]

i am also about to bust

[@chrismasterith]

The hardware is good though.

[@Boko-Maru]

the best thing to come out of this

[@lucasrem]

you play games, hardware ???

[@jackhand4073]

Spam bot.

[@GlobalPenguin2012]

Great clip…. Thank you

[@Derpy1969]

AI hasn’t figured out how to make money.