4 Comments

I enjoyed this. Some nice suggestions.

Expand full comment

I like the article, like as a mirroring contrast to Marcus’s, et al. But like all the material on “AI”, a misnomer, we have no clue so long as we depend on LLMs for “AI.” It will continue to be 1. An Onomatopoeia; 2. A self-confirming feature-enrichment tautology of Bayesian incompleteness by definition of Bayesian (you can never enrich enough, there is always something sub in an atom and surrounding it by tangents, beside it); 3. An utter, dismal unacknowledgment of the constitute social basis of a topology, where vector decomposition also occurs on - NVIDEA, LLM mathematics, etc.). It’s a lost cause and a waste of time until and even if those matters can ever be solved. AI doesn’t exist. We impute and imbue it, both in our modelling and in attitudes to it, which go to an equivalence class of their very own making. Goodby Columbus.

Expand full comment

What an unusual article! Instead of fear-mongering and making up fantasy scare scenarios, you have written a thoughtful and helpful piece on benefits and approaches to reducing risks. It seems that there are far fewer of those us taking this approach than there are of those hyping both short-term and long-term dangers, with rarely a word to say about massive benefits.

Expand full comment