Nuclear weapons, the maxim gun, lead paint, lead gasoline, basically all lead-based products, thalidomide, CFCs, the electric chair, agent orange, asbestos, oxycodone, zyklon b, refined sugar, high fructose corn syrup, disposable plastics, cigarettes, trans fats, ...

I think @PeepinGoodArgs@reddthat.com is doing a great job of pointing to the actual substance of the argument, so I'll leave that to them, but it's actually really easy to come up with a long list of technological horrors that absolutely did not benefit most people but had huge impacts on our economy.

I do think "impact on our economy" is a pretty squishy phrase that'll give you some wiggle room, but many of these nightmare technologies are inextricably and inseparably tied to the way we've structured our economy. Likewise, I think it's easy to define "technology" in convenient ways for these kinds of arguments, but also ends up being circular pretty quick.

I can, totally, see AI only benefitting the people who own the code and make policies for it. Despite the fact that it may be used to "benefit" most people, the ones who will benefit the most are the people who own it. Similar to targeted ads. It's a multi-billion dollar industry that gathers insane swathes of information on individuals, and that information is bought and sold to the highest bidder. You could make the argument that it's easier to buy shoes online, but is it worth having literally everything about you sold to whoever is willing to buy it? It's usually a ruse crafted by people with the ability to profit off of others, making the majority think they're benefiting in some minute way.

What Acemoglu is saying is fundamentally a Marxist argument, and I'm saying that with no value judgments attached, I'm just pointing out factually that he's essentially saying the same thing as Marx. In summary, technology tends to disproportionately benefit the people who can afford to implement it (the owners). AI is a means of production. While it's currently possible for anyone to download the means of production for free, no violence required, currently you've got several large AI companies (like OpenAI) trying to pull the ladder up behind themselves under the guise of safety and ethics concerns. They're trying to protect themselves from the tendency of the rate of profit to fall by angling to limit the pool of competitors. Indeed, much of whether a technology benefits society at large is dictated by the barriers to entry to using that technology. If they are successful, the barriers to entry will be made much higher, and it will all but guarantee that the benefits of AI stay at the top.