No I'm not. It's been impressive but it seems to have plateaued. It's not particularly reliable, and we're at the point of diminishing returns with the current technology and scalability. We're not seeing big jumps in models capabilities in the ways we did at the start, and the investment needed to get incremental increases in power are now looking insane.

We're in the middle of a stock market bubble, where everyone is hyping up their products and speculators have been throwing money at AI, desperate to have a slice of the "winner". But the productivity gains we've been promised haven't followed. If you use the products even a little, you really see the limitations - they "hallucinate", they don't reliably produce the most efficient answers, they often tell you what you want to hear.

A lot of products have been badged as "AI" but they're really either LLMs or more targetted generative AI. They are certainly useful, and some are undoubtedly going to be profitable, but they're not anywhere as near useful or powerful as the proponents are trying to make out. Companies like OpenAI for example are burning through money at an alarming rate, and not really producing much gains now. Companies like Microsoft, Meta and Google have deep pockets and have thrown many $100s of billions at AI, yet to what end? Little chat bots, tools that help you make emails briefer/concise, tools that sort-of help you search the internet? It's not as revolutionary as it first seemed.

It's telling that companies are quietly backing away from hyped up mega-deals with OpenAI (Microsoft, Nvidia), and that Microsoft has pivoted away from Copilot due to backlash and lack of user uptake. Their own stats shows only 3% of users were using Copilot, yet they've rammed it into almost every corner of Windows and Office, and to the detriment of their actual core product and users.

General AI is likely on the way, but not from this initial burst of technology, and probably not from any of these companies. LLMs and other current techniques are interesting and have their uses, but the idea that we just scale it up and have General AI is seemingly nonsense. The biggest proponents of that idea are companies like OpenAI that are desperate that the technology they have is the one that dominates. Instead I suspect it's the researchers and companies quietly working on technology that better mimics the actual biological brain that will get us to something like General AI, and that seems to be quite a way off.

Complex ideas like this are not just about throwing money at the problem; there is plenty of resource going in to AI development but what you actually need is good quality research and time. Much of the money is very likely being wasted on duplication, pointless infrastructure (like more data centres for current models to scale), bloated salaries of research staff due to competition, bloated salaries of non-research staff and of course the stock-market shenanigans and deals that really produce nothing for AI research.

Don't be worried about AI; be worried about this stock market bubble popping.