When I realized that an LLM can process all the traffic in Slack that overwhelms me daily and give me a manageable digest. How long until they intermediate most of our social interactions? Sooner than we can possibly adapt, I think.
What you're describing is what's known as an "end to end" model that takes in image pixels and outputs steering and throttle commands. What happens in an AV is that a bunch of ML models produce input for software written by human engineers, and so the output doesn't come from an entirely ML system, it's a mix of engineered and trained components for various identifiable tasks (perception, planning, prediction, controls).
This is a pretty old idea that goes back to debates over David Ricardo's theory of comparative advantage in the early 1800s.
> “Free trade is God’s diplomacy. There is no other certain way of uniting people in the bonds of peace than by the bonds of interest.”
- Richard Cobden
This topic is covered in the book "Dark Side of the Moon: The Magnificent Madness of the American Lunar Quest" by Gerard Degroot.
As I remember it, the author says that Eisenhower was focused on strategic value of investments into space, and he felt that the US had a large lead in areas that mattered (ICBM technology) and chose not to invest in what might be considered vanity projects.
I believe he also claims that letting the Soviets be first to launch a satellite was intentional, since it allowed the Soviets to establish the precedent in international law that it was OK to send satellites over your neighbors, and if the US had been first the USSR might have protested.
I guess there is a "that which is seen vs. that which is not seen" (Bastiat) argument to be made, but government investment into engineering research seems to be a pretty good deal.
I saw a quote the other day I liked: "If the rest of the world wants to emulate the US model, they should do as the United States actually did, not as they say they did".