Deep and useful. It plugs in well into the ideas from "The All-Knowing Father", an essay written more than 1 year ago, which also saw LLMs as a new kind of interface to knowledge. There are in it two analogies to LLMs, the map and the hologram. Going further with the map one, the two scenarios you talk about will translate to an inexperienced explorer vs an experienced explorer, both using a detailed but not verified map to reach their goal (it may include a lot of non-existing features and may omit existing ones). It is easy to see the possible result. Maybe the most important point is that "knowing what" is very different from "knowing how"; experience is much more than abstract knowledge. This will reinforce the specialization effect you describe. The essay is here, I genuinely think it will be of interest to you: https://antimaterie.substack.com/p/the-all-knowing-father
An excellent comparison, Nicolas! It is very useful to differentiate between technological and process innovation. Another example would be electrification: Right after the commercial production of electricity started, there was a surge in investments into power plants, and rightly so of course. But in the long run, the value creation was in the companies that successfully integrated electricity in their manufacturing processes and therefore became more competitive. It is not a question of doing one and not the other - we Europeans should invest in the application layer as well as the models, but I think the sequence matters: https://danielflorian.substack.com/p/the-unsolved-dilemma-in-henna-virkkunens
Deep and useful. It plugs in well into the ideas from "The All-Knowing Father", an essay written more than 1 year ago, which also saw LLMs as a new kind of interface to knowledge. There are in it two analogies to LLMs, the map and the hologram. Going further with the map one, the two scenarios you talk about will translate to an inexperienced explorer vs an experienced explorer, both using a detailed but not verified map to reach their goal (it may include a lot of non-existing features and may omit existing ones). It is easy to see the possible result. Maybe the most important point is that "knowing what" is very different from "knowing how"; experience is much more than abstract knowledge. This will reinforce the specialization effect you describe. The essay is here, I genuinely think it will be of interest to you: https://antimaterie.substack.com/p/the-all-knowing-father
An epic read.
An excellent comparison, Nicolas! It is very useful to differentiate between technological and process innovation. Another example would be electrification: Right after the commercial production of electricity started, there was a surge in investments into power plants, and rightly so of course. But in the long run, the value creation was in the companies that successfully integrated electricity in their manufacturing processes and therefore became more competitive. It is not a question of doing one and not the other - we Europeans should invest in the application layer as well as the models, but I think the sequence matters: https://danielflorian.substack.com/p/the-unsolved-dilemma-in-henna-virkkunens