Physical Address
304 North Cardinal St.
Dorchester Center, MA 02124
Generative AI took the world by storm in November 2022, with the release of OpenAI service ChatGPT. One hundred million people started using it, practically overnight. Sam Altmanthe CEO of OpenAI, the company that created ChatGPT, became a household name. And at least half a dozen companies are racing OpenAI in an effort to build a better system. OpenAI itself wanted to surpass GPT-4its flagship model, introduced in March 2023with a successor, presumably to be called GPT-5. Practically every company has promised to find ways to adopt ChatGPT (or a similar technology, made by other companies) in their business.
There’s just one thing: generative AI doesn’t work that well, and maybe it never will.
Basically, the engine of generative AI is gap-filling, or what I like to call “autocomplete on steroids.” Such systems are great for predicting what might sound good or plausible in a given context, but not for understanding on a deeper level what they are saying; an AI is constitutionally incapable of controlling its own work. This has led to massive problems with “hallucinations”, in which the system asserts, without qualification, things that are not true, while inserting head 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 in doubt make for fabulous demos, but are often unsavory products in themselves. If 2023 was the year of AI hype, 2024 was the year of AI disillusionment. Something I argued in August 2023, to the initial skepticism, was felt more frequently: generative AI could be a dud. The profits are not there –estimates suggest that the operating loss of OpenAI 2024 may be $5 billion, and the valuation of more than $80 billion is not in line with the lack of profits. Meanwhile, many customers seem disappointed with what they can actually do with ChatGPT, rather than the extraordinarily high initial expectations that had become common.
Also, essentially every big company seems to work from the same recipe, making bigger and bigger language models, but they all end up in more or less the same place, which are models that are about as good as GPT-4, but not a much better one. What this means is that no individual company has a “moat” (the ability of the company to defend its product over time), and what it means in turn is that profits are decreased. OpenAI has already been forced to cut prices; now Meta gives away similar technology for free.
As I write this, OpenAI has demonstrated new products but is not releasing them. Unless it comes out with a major advance worthy of the name of GPT-5 before the end of 2025, which is decidedly better than what its competitors can offer, the bloom will be off the rose. The enthusiasm that supported the OpenAI will decrease, and since it is the poster child for the whole field, the whole thing may soon be over.
