Generative AI took the world by storm in November 2022, with its release OpenAI service ChatGPT. One hundred million people started using it, almost overnight. Sam AltmanThe CEO of OpenAI, the company that created ChatGPT, has become a household name. And at least half a dozen companies are racing against OpenAI in an effort to build a better system. OpenAI itself has managed to overcome it GPT-4its top model, introduced in March 2023with a successor, probably called GPT-5. Almost every company is trying to find ways to apply ChatGPT (or a similar technology produced by other companies) to their business.
There's just one thing: Generative AI doesn't really work that well, and it probably never will.
Essentially, the tool of creative AI is fill-in-the-blank, or what I like to call “auto-completion on steroids.” Such systems are good at predicting what sounds cool or reasonable in a given context, but are unable to understand at a deeper level what they are saying; AI is constitutionally incapable of fact-checking its own work. This has led to major problems with “illusions,” in which the system asserts, without qualification, things that are not true, while inserting blunders into everything from arithmetic to science. learn. As they say in the military: “always wrong, never in doubt”.
Systems that frequently fail and never doubt make for great demos, but they are often terrible products themselves. If 2023 was the year of AI hype, 2024 is the year of AI disillusionment. What I argued in August 2023, to initial skepticism, has been felt more often: Generative AI could turn out to be a dud. The profit is not there—suggested estimate that OpenAI's 2024 operating loss could be $5 billion — and a valuation of more than $80 billion doesn't mean a lack of profitability. Meanwhile, many customers seem disappointed with what they can actually do with ChatGPT, compared to the extremely high initial expectations that have become common.
Furthermore, basically every major company seems to be working on the same formula, creating larger and larger language models, but all in more or less the same place, which are the models. about as good as GPT-4, but not a whole lot better. That means no individual company has a “moat” (the ability to protect a business's products over time), and that means profits are shrinking. OpenAI was forced to reduce prices; now Meta is giving away the same technology for free.
As I write this, OpenAI has introduced new products but hasn't actually released them yet. Unless it makes some major progress worthy of the GPT-5 designation before the end of 2025, which is significantly better than what its competitors can offer, the roses will be in full bloom. Enthusiasm for OpenAI will wane, and since it underpins the entire field, the whole thing may soon go bankrupt.