I just read this very AI-skeptical article, that basically says that today's generative AI has no credible business model and is unlikely to improve significantly enough to get one. While I am aware of counter-arguments to the technical aspects, I wonder if there are more positive takes on the financial/business aspects, coming from anyone else than AI vendors.

Carl Shulman’s perspective is interesting - I think he’s someone who really takes the “look at the trends and then keep extrapolating” thing seriously, which leads him to very extreme conclusions about the expected impact of AI. Could be a bit of a different discussion though, as he’s not talking about today’s models. Anyways here’s a very long interview with him. https://80000hours.org/podcast/episodes/carl-shulman-economy-agi/

That's extrapolation with a big dose of speculation. We have no idea for now if AGI will ever happen. So it's not a basis for today's business plans. Even if AGI happens, and even if it happens soon, it's probably neither OpenAI nor Anthropic that will benefit economically from such a development.

Saying that LLMs don't have a commercial future today is like saying automobiles don't have a commercial future 6 months after the model T is released, because the roads are bad and gas is hard to find. Society evolves in the context of its technology. GPT4-level LLMs have been generally available for, what, 18 months or so? It takes time for behaviours to switch. The fact that the easiest and fastest thing you can do is "rub a chatbot on it" doesn't mean that's the best or only thing you can do. in my day job, or clients give us problems, and we prototype LLM-powered solutions over a couple of months that improve the quality of the output over human work, reduce the time required, reduce the expense involved, and eliminate untold hours of drudgery. We demonstrate the potential on one project, and two months later there are six projects of the same size. The limiting factor is us. There isn't enough human labour to implement LLMs everywhere they have value. And of course not. It has only existed for 18 months.

I agree, but the long-term,perspectives were not the topic of my question. The model T was economically viable from the start. Today’s LLM deployments are not, they require large subsidies, if the article I cited is correct. The question is not so much about a commercial future, but about a commercial present.

That AI is not currently profitable is true, but looking at that, and M365 copilot, is not a rational way of determining if it is economically viable. Does the author know what percentage of the losses at OpenAI are attributable to operations of existing models and what portion are for R&D? No. But do you know who does have that data? The investors who just poured another $7B into OpenAI last week. It's click-bait nonsense.

Thanks @Jason Morris , that’s the kind of reply I was hoping for! Not sure I’d trust investors to make good forecasts about economic viability though. They can also hope for short-terms due to the hype.

I use these models every day in my work. My instinct is that they are extremely economically viable. Chat bots may not be, but contrary to what you might gather from the current crop of products, "talking to people" is not the highest and best use of LLMs. Not by a long shot.

I most definitely agree on that! And yet, chatbots probably make up for most of today's LLM use, and are therefore a central issue in economic viabiliity.