The Coming AI Startup Bust

Links:

[QwertiusMaximus]

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

[jorgeroman]

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.

[AvenEngineer]

One must first pump, before they may dump.

[MoneyMacro]

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

[kalomboC]

Those 'sophisticated investors' poured money into

[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.

[shrek]

"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.

[livenoob]

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.

[TzSik]

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

[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.

[foobar]

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

[neskey]

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

[XieRH]

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

[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.

[andreikersha]

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

[Xeonerable]

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

[cinemaipswich]

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.

[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.

[Dannyboi]

"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.

[breezyx]

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