If AI has a bright future, why does AI think it doesn't? | Hacker News Reader

If AI has a bright future, why does AI think it doesn't?

claude.ai

15 pointsJCW200121 comments

Those who think Gary Marcus, Ed Zitron and Yann LeCunn are wrong, and believe in AI: How do you reconcile things when AI thinks the market is highly likely to collapse?

Quote: "The entire system only works if AI revenue grows fast enough to outrun the obsolescence treadmill. For that to happen, Microsoft would need approximately $130 billion per year in new AI revenue, Google $100 billion, Amazon $120 billion, and Meta $70 billion. Against a current reality of $18 billion in total industry AI revenue and zero profits, that gap is not a rounding error. It is the entire bet."

Stock market collapses and technological success are two very different things. The internet led to a market collapse just as it started showing real promise. One of the problems is that you can't invest in a technology, only in companies, and oftentimes the companies that turn a technology into financial success are not the ones that exist when the technology is in its infancy. It's not unlikely that LLM will be a huge success while Nvidia, OpenAI, and Anthropic collapse.

When people are talking about an AI bubble, they are explicitly talking about the stock market. It’s not at all about the question if LLMs are useful or not.

Of course, but I was responding to a comment that contrasted "believing in AI" with the stock market. People believed in the internet, it still led to a bubble and a market crash, and that didn't mean those people were wrong. Part of the cause for a bubble, then and now, is that you can't invest in a technology, only in companies. Investors want to invest in AI, but they can't, so they invest in OpenAI, which may well go bankrupt. Those early internet pioneers didn't deliver on the inflated financial expectations from them even while the internet as a whole eventually did.

What's interesting to me is how quickly the big AI labs are losing their competitive edge even before becoming profitable. Models that are less than one year behind the leading ones are effectively commoditised already. If OpenAI and Anthropic disappear tomorrow, it will be no more than a brief inconvenience to LLM users. They're spending a lot of money for an almost negligible advantage.

From skimming that chat: it doesn't, no matter how much you tried to steer it into that direction. It barely reluctantly agreed under some hypotheticals, but it never agreed to the hypotheticals actually being true.

This submission is junk, and the the title is editorialized.

Direct reply from Claude: https://claude.ai/share/628e6a1d-087d-4454-95e3-cf8a3c4a5b72

"The case for a bright future is strong."

I trust Claude over JCW2001

Ha. Try again

Because AI doesn't think and apparently a lot of people using it don't think critically either.

Exactly. LLMs at their core are just fancy autocomplete. Extremely fancy, to be sure, and the output that they predict can be very useful - but people who anthropomorphize them or ascribe higher significance to the generated output seem to be missing this.

LLMs are heavily biased by what it is told.

It is not magic, it is not an oracle, it is not good at analysis, and is particularly bad at predicting the future.

Garbage in, garbage out.

(If OP was trying to make the bull case then the same snark would still apply.)

How is it not a good analysis? Genuine question: it seems like it is summarizing the bear case which to my very limited understanding continues to be reinforced

Substantive issues with this submission aside, it’s a mistake to have such long conversations with an LLM. The longer they go, the more likely they are to accumulate errors. The latest models all claim to be able to handle long conversations, but in my experience they still don’t do as good a job as just pasting your conversation into a new thread.

I made this comment half a year ago as well, but i believe AI is going to bring down the profitability of the big tech companies by a lot.

Instead of massive scaling advantages which has given software its extreme valuation, it now hit on something that is almost a perfect commodity. Energy and depreciation are easy to calculate and its subject to global competition.

Great for consumers, less so for people looking for a ROI.

I worked at yahoo during its (in retrospect) decline.

It used to be hard to be "web scale" and available, now that's either k8s or a few checkboxes in AWS.

Yahoo used to be able to "coast" on the compellingness of their services because 80% attractive with 100% available and 100% global reach crushes 90% attractive with 95% available and 25% global reach.

I was often confused by the hyperfocus of analysts asking "Is Y! a tech company or a content company?"

What they were really asking was if we should be valuing Yahoo! as 30%+ margin on putting ads next to Yahoo! News articles, or 10x multiplier on originating GMail/Search?

I think "data is the only moat", and in a way that goes back to the "first to market / eBay" POV, and the difference between first to market and fast follower is super interesting!

a) LLM’s don’t have introspection capabilities

b) in my observation, the longer context window, the more unhinged/pessimistic LLM output becomes

How are the responses here unhinged? They are summarizing a lot what the bear case has been shouting for the last two years

LLM's are like accountants looking at the past.

The numbers are bad therefore it will collapse.

were just sharing claude chats now

LMs. don't. "think". It. Just. Generates.some.text. based. On. Whatever. You. Dumped. Into. It

Chatbot log as a submission? Really?

I dunno, it's kind of fun to watch people faceplanting as they try to ride their mind-bicycles.

Reply on news.ycombinator.com