People might not want to be on update treadmill of proprietary weight models, for enterprise things. Things change a lot between models and they can't guarantee backwards compatibility like you can for deterministic software
I think there is a rich seam of work for computers to help humans spot when there are anomalies in the hypotheses humans hold that can help humans form the corrections to hypothesis that fit our aesthetic sensibilities
I've thought of a service that scans new websites and GitHub repos and looks for things that don't look like anything else (using something like hdbscan for outlier detection), and creates a feed for people to follow.
Have you come across the lethal trifecta [1]? I'm interested in decomposing tasks to avoid any agent doing all three and having limited data pipes between them.
I think it is great for experimenting, and proving concepts. Alphas and personal projects, not shipped code.
I've been working on wasm sandboxing and automatic verification that code doesn't have the lethal trifecta and got something working in a couple of days.
I've wondered if LLMs can help match people. People give the LLM some public context about their lives and two LLMs can have a chat about availablity and world views.
Use AI to scaffold relationships not replace them.
Fair, I don't know how valuable it would be. I think LLMs would only get you so far. They could be tried in games or small human contexts . We would need a funding model that rewarded this though.
I'm not sure. We can use LLMs to try
out different settings/algorithms and see what it is like to have it on a social level before we implement it for real.
I've been thinking about using LLMs to help triage security vulnerabilities.
If done in an auditably unlogged environment (with a limited output to the company, just saying escalate) it might also encourage people to share vulns they are worried about putting online.
https://claude.ai/public/artifacts/d12af7c5-3eb0-4218-a505-e...