Of Dave Lee

At the beginning of the year I wrote that Wall Street investors should be prepared for a “winter of artificial intelligence” in 2025. This is not necessarily a reduction in investment, and certainly not the advertising frenzy of companies, but of tangible progress. I had warned that patience would be tested.

Some recent developments make the question timely: Is the winter of artificial intelligence in front of the gates?

The GPT-5, Sam Altman’s long-awaited new model, was released earlier this month and was accepted with lukewarm reactions. If it is a step towards the so -called artificial general intelligence, as the company has repeatedly supported, it is definitely a very small step. The model was so poorly welcomed by some loyal users of Chatgpt that the company was forced into a shameful folding, restoring the older models. Altman’s claim that the GPT-5 was like talking to a “special doctorate level” quickly ended up anecdotal.

Last week, Coreweave Inc., one of the few “net” shares of artificial intelligence, sliding over 25%. The reason was the forecasts that terrified investors: the revenue is expected to lag behind in terms of capital expenditure. The end of the IPO lock period came to aggravate the fall.

And as difficult it is to determine how beneficial it was, or it will be, TN for the business world, a McKinsey & Company research has to put us in thoughts. While 8 out of the 10 companies responded said they were applying TN genetics in their business, the Counseling Company noted that many have said that there was no “significant impact on financial results”.

The reaction especially to the GPT-5 gave the AI old skeptics the opportunity to triumph. One of the best known, scientist Gary Marcus, celebrated the obvious “fall” of the GPT-5. “Rightly, Altman’s reputation should have been burnt completely,” he wrote, accompanying the comment with a list of seemingly insignificant work that the GPT-5 failed to perform adequately. Writer and journalist Brian Merchant noted that Altman was less willing to use the term “artificial general intelligence” now that his last model is obviously not still. “I don’t think it’s a particularly useful term,” Altman told CNBC. Merchant recalled that it is a term that Altman has often used, even last February on his personal blog. It has been useful to raise billions of dollars.

Altman has made other revealing statements in recent days. “Are we in a phase where investors overall are overly excited about AI? My opinion is yes, “he said last week to a group of journalists. “You have to expect that Openai will spend trillions of dollars,” was another statement. My favorite, from an interview with CNBC last week, was: “It’s nice not to be listed on the Stock Exchange.”

I don’t doubt! All this time I was wondering how investors would have reacted to the GPT-5 if Openai was a publicly listed company. I suspect that at least it would have been a week like Coreweave, as soon as investors took into account the rebellion of users, the “folding” on the issue of hypertension, and the prediction that it would need “trillions” of extra dollars. (Altman said he was sure that the company could invent “a new type of financial tool for funding and computing power” to support its rapid industrial development. We probably now know at least one of the books found in GPT-5 training data.)

However, the negative effect did not extend to other shares closely linked to Openai’s luck, such as Microsoft Corp. or Nvidia Corp. This shows that investor nerves have not been stretched. This may be due to some milder analyzes. One view is that shameful viral failures, such as the inability to correctly write the word “blueberry”, are insignificant tricks that lose the big picture: that the GPT-5 is more sophisticated in choosing the right model for a job, which may seem insignificant but actually practical and useful. Another view is that the possibilities have improved large enough for the innovative AI digital assistants who can undertake and perform specific tasks, are now very close and then the investment will begin.

Or may not, and this month be considered the beginning of some important thing. I don’t think we can yet talk about “AI winter” – but there is definitely a sudden cool in the air.

My conclusion from the release of the GPT-5 is that while AI companies can show overall performance in various benchmarks, they are becoming less and less significant. Until one who is not researchers AI, these scores have little value to the end user, whether it is the consumer or the CEO.

What shapes the narrative of the progress (or the absence of progress) of AI is the practical application, and here all companies are still failing. The GPT-3.5 existed for months before the presentation of Chatgpt impresses the world until we all discovered its many weaknesses, not through laboratory tests of millions of questions but with our own eyes. Digital assistants can be the next “moment of Chatgpt” if they do what they promise. They need better performance, otherwise investors may be confronted with a frozen winter.