> The implicit social contract between mathematicians and AI companies deserves further attention
It certainly does, and locally I have been disappointed in the how slow we are to build that social contract. For some years, and through two international reviews (2004 and 2010), the gap between mathematics and computer science in the UK has been identified as a weakness. AI should be seen as a novel and well-supported source of new challenges and new connections capable of being a benefit to the intellectual health of research in mathematics. By this I don't just mean the application of AI systems to mathematics, important as this might be. I mean particularly the development of the theories, techniques and tools for principled understanding, reasoning and quantification of AI systems.
Your note raises important questions and I encourage everyone who reads Silicon Reckoner to read it. It was published in a newsletter whose slogan begins "Cultivating optimism about technoscientific progress…," an attitude that I am quite far from sharing, but I read your note as a specific warning against uncritical optimism. I have to say that, having read far too many press releases emanating from Silicon Valley, I am not surprised that "almost nobody understood" your point about the order of training data. If engineers have trouble with operations that don't commute, imagine the level of confusion in the boardrooms.
the usual label of "communism" from libertarians, who have no idea about the latter (profits? 10 years of Gulag!). It's a bit of socialism here that's proposed, at best (cf. Medicare etc)
> The implicit social contract between mathematicians and AI companies deserves further attention
It certainly does, and locally I have been disappointed in the how slow we are to build that social contract. For some years, and through two international reviews (2004 and 2010), the gap between mathematics and computer science in the UK has been identified as a weakness. AI should be seen as a novel and well-supported source of new challenges and new connections capable of being a benefit to the intellectual health of research in mathematics. By this I don't just mean the application of AI systems to mathematics, important as this might be. I mean particularly the development of the theories, techniques and tools for principled understanding, reasoning and quantification of AI systems.
I wrote a note about this at https://sotaletters.substack.com/p/mathematical-challenges-from-ai-ethics
Your note raises important questions and I encourage everyone who reads Silicon Reckoner to read it. It was published in a newsletter whose slogan begins "Cultivating optimism about technoscientific progress…," an attitude that I am quite far from sharing, but I read your note as a specific warning against uncritical optimism. I have to say that, having read far too many press releases emanating from Silicon Valley, I am not surprised that "almost nobody understood" your point about the order of training data. If engineers have trouble with operations that don't commute, imagine the level of confusion in the boardrooms.
Thanks!
> We should all own it and share in the profits.
Communism. You're talking about communism!
the usual label of "communism" from libertarians, who have no idea about the latter (profits? 10 years of Gulag!). It's a bit of socialism here that's proposed, at best (cf. Medicare etc)
You shouldn't let yourself be confused by labels.
https://en.wikipedia.org/wiki/Alaska_Permanent_Fund
Or if you prefer clean energy,
https://comptroller.nyc.gov/services/for-the-public/confronting-the-climate-crisis/public-solar/