I still think AI will eat the world

Even if the utopian predictions prove false, Large Language Models are still a big deal

For the last couple of years, I’ve been out of step with many opinion-havers on my corner of the internet.

Though Silicon Valley might be populated by true believers in AI, my social world is… less enthusiastic. Some people are justifiably sceptical of some of the more expansive claims about the implications of this technological moment – and others are outright deniers, who refuse to entertain the idea that AI is little more than the next overhyped bust after the crypto scams, NFT frauds, and metaverse nothings of previous hype cycles.

This is why I’m in a strange position: Since I first tried ChatGPT, I’ve been convinced that this is a profoundly important technology. And since it has felt like every week, this conclusion has been renewed by watching AI make possible things that were previously inconceivable. I’ve catalogued many of them in my Odds and Ends links newsletters.

However, I will concede there are plenty of reasons to be sceptical. Earlier this week, my friend and podcast co-host Martin Robbins published a blisteringly great piece explaining why he believes that fundamentally, Large Language Models (LLMs) are a bust:

[T]hey are profoundly limited in many ways. They contain no knowledge and have no mechanism for reasoning on the knowledge they don’t have. They have no independent agency or driving insight beyond that provided by the user. These are not bugs to be fixed but fundamental limitations of this kind of technology.

But here’s the strange thing. Even though I basically accept that all of Martin’s arguments are correct, I still think this technology is both extremely useful, and a really big deal for humanity.

So I thought that, in a similar spirit to our debate about AI on The Abundance Agenda this week, that I should explain why – despite everything – I’m still bullish about AI.

I’m not an AI bot, so if you value my 100% pure, human-written writing, please consider taking out a paid subscription to support my work, and read the rest of this piece.