I doubt that an LLM would do very well, but other forms of ML (like a model trained to work in business rules and economic outcomes rather than stringing words together) could probably be used. Hell, they probably won’t need to use the planet-destroying mega data canters to do it; some very effective specialized ML models can be run on a Raspberry Pi.

The biggest problem with AI is they’re convincing people they can think and replace workers by using extremely large language models trained on stolen works provided as cloud services asking for exorbitant subscription fees.

In other words, tech bros figured out how to take something useful and make a grift out of it, because you can get rich by gatekeeping something of value. Why do you think they went so far as to threaten to ban DeepSeek when a free model that can run on a desktop PC from a foreign “adversary” appeared? They all shit their collective pants because the jig was up.