Generative AI took the world by storm in November 2022 with the release of OpenAI's ChatGPT service. A hundred million people started using it virtually overnight. Sam Altman, the CEO of OpenAI, the company that created ChatGPT, became a household name. And at least half a dozen companies raced OpenAI in an attempt to build a better system. OpenAI itself tried to surpass GPT-4, its flagship model, introduced in March 2023, with a successor, presumably called GPT-5. Virtually every company rushed to find ways to implement ChatGPT (or a similar technology, created by other companies) into their business.
There's just one thing: generative AI doesn't actually work that well, and maybe it never will.
Essentially, the engine of generative AI is filling in the blanks, or what I like to call “autocomplete on steroids.” Such systems are excellent at predicting what might sound good or plausible in a given context, but cannot understand what they are saying at a deeper level; an AI is constitutionally incapable of fact-checking its own work. This has led to huge problems with 'hallucination', where the system claims things that are not true without any qualification, while inserting obtuse errors into everything from arithmetic to science. As they say in the military: “often wrong, never in doubt.”
Systems that are often wrong and never hesitate make for great demos, but are often lousy products on their own. If 2023 was the year of AI hype, 2024 is the year of AI disillusionment. Something I argued in August 2023, to initial skepticism, is being felt more often: generative AI might turn out to be a dud. The profits aren't there – estimates suggest OpenAI's operating loss could reach $5 billion by 2024 – and the $80 billion-plus valuation doesn't match the lack of profits. Meanwhile, many customers seem disappointed with what they can actually do with ChatGPT, compared to the extremely high initial expectations that had become commonplace.
Furthermore, essentially every major company seems to be working from the same recipe, making bigger and bigger language models, but all ending up in more or less the same place, namely models that are about as good as GPT-4, but not a whole lot better. What that means is that no individual company has a 'moat' (a company's ability to defend its product over time), and what that in turn means is that profits are declining. OpenAI has already been forced to cut prices; now Meta is giving away similar technology for free.
As I write this, OpenAI has been demonstrating new products, but not actually releasing them. Unless it comes up with a major advance before the end of 2025 that is worthy of the name GPT-5 and that is decisively better than what their competitors can offer, the rose will be off the charts. The enthusiasm that supported OpenAI will wane, and since it is the example of the entire field, the whole thing could soon go bankrupt.
