I’ve been trying ChatGPT plugins & they only work okay so far, but it is clear that they are going to be a big deal. But I can’t help but return to the OpenAI white paper that discussed how connecting GPT-4 to other systems can lead to unpredictable risks. (Look at example, too)
Emergent behaviors are the scary variable. The systems are sufficiently complex that Gödel’s incompleteness theorem and Turing’s prof on undecidability are gonna wreak some havoc. You cannot predict the phenotype from the genotype, and we are looking at some incredibly powerful DNA molecules here.
@emollick Fully autonomous AIs that generate tasks for themselves are already ready to go. As soon as people realize these make money, everybody will want one. A future economy with millions of shoggoths intertwining their tentacles with each other is certainly... interesting.
@emollick Ethan, what's your take on prompt engineering as a field? Any course/path recommendations for folks interested in it? Thanks for your work.
@emollick Feauring this in ‘What Did OpenAI Do This Week?’ whatdidopenaidothisweek.substack.com/about
@emollick As soon as they released the API, this was inevitable, though.
@emollick Microsoft smelled Google's blood... Who cares about the rest.
@emollick This is going to grow into a class room type situation. Kids passing notes around and playing dumb when they get spotted. We must assume its like schrodinger cat, but with danger and people taking the tool to use within their own sense of ethic that might not match yours.