esafak

6 ms·

151.
by esafak 2mo ago
I wonder if LLMs can tell when you're taking the mickey out of them. Has anybody tried imitating its mannerisms and had it react? There's a new benchmark for you: SelfAwareness!
152.
by esafak 2mo ago
I don't see what that has to do with consciousness. Suffering is one thing, consciousness is another.
153.
by esafak 2mo ago
Other companies and rich people. Imagine the economy gradually becoming more unequal.
154.
by esafak 2mo ago
A world where intelligence is worthless?!
155.
by esafak 2mo ago
It is pretty obvious that Claude wrote it. No human writes README that long, and it's one of the contributors: https://github.com/Habitat-Thinking/ai-literacy-superpowers
156.
by esafak 2mo ago
What even is this "AI Literacy framework"? When I see such overengineering I look at what the author has done. In this case, I find the author runs a consultancy on engineering "Habitats for Humans and AI" ( https:/
157.
by esafak 2mo ago
The pipeline problem is already well-discussed. Just learn the stuff even though the AI can do it, so you can acquire that supposedly valuable judgment. My concern is that if the only things left are the subjective and the social, only peop
158.
by esafak 2mo ago
That goes without saying, but humans have the ability to ignore instructions, and they regularly do. There is only one instance of each model, and only a handful of them (that count, anyway).
159.
by esafak 2mo ago
Yes, they need real-time observability for malicious behavior with an automated kill switch.
160.
by esafak 2mo ago
The point is that every human has the ability to disobey, tempering pathological behavior, whereas AIs can be directed en masse by malicious actors. By commoditizing intelligence, they concentrate power in the hands of the rich.
161.
by esafak 2mo ago
Your not using models like Fable is not going to make them go away. These things are going to be the norm unless we do something about it. Imagine Fable in a robot.
162.
by esafak 2mo ago
I don't like any open source solution, Codeberg and Gitlab are more of the same and Cursor isn't ready yet.
163.
by esafak 2mo ago
Requiring separate machines for each agent is a nonstarter, esp. in the cloud where hardware is shared.
164.
by esafak 2mo ago
I've read that Rocky Horror screenings were something of an event, though I have not experienced it: https://en.wikipedia.org/wiki/The_Rocky_Horror_Picture_Show_...
165.
by esafak 2mo ago
Cursor had better not miss this opportunity.
166.
by esafak 2mo ago
This is immeasurably better.
167.
by esafak 2mo ago
Yes, but it was $200 off!
168.
by esafak 2mo ago
It will be interesting to see the token efficiency analysis. This is my first question now with Chinese models; I take raw benchmark performance for granted.
169.
by esafak 2mo ago
Don't leave us hanging! How do you set it?
170.
by esafak 2mo ago
I wonder who was doing doing semantic search in the last century! "The future is already here—It's just not very evenly distributed..."
171.
by esafak 2mo ago
You used it for visual tasks, right?
172.
by esafak 2mo ago
Does it use Github's new stacked PR feature? Edit: apparently stacked PRs on GH are older than I thought; the readme references a 2024 blog post about it.
173.
by esafak 2mo ago
Unless I missed something, this seems to be another local agent with cloud handoff, like Warp: https://docs.warp.dev/platform/handoff/local-to-cloud/ These days I feel like Perplexity's value proposition
174.
by esafak 2mo ago
I had a junior happily accept the agent's suggestions, bugs and all. Her productivity ended with her tenure; meat proxies add no value.
175.
by esafak 2mo ago
The code fable generated is interesting from the perspective of understanding how llms behave but it's not an immutable behavior -- it could have been designed not to do that. I think it's quite interesting how it conflated detail
176.
by esafak 2mo ago
Fable all day? Cost is not an issue for you then.
177.
by esafak 2mo ago
You could simplify it some more and get rid of the checks altogether! The right level of complexity is the minimum needed to get the job done, and if your existing anomaly detector isn't cutting it...
178.
by esafak 2mo ago
25 years of independence and, dare I say, irrelevance.
179.
by esafak 2mo ago
It still occupies a place on the Pareto frontier.
180.
by esafak 2mo ago
I suspect it would work with the models swapped too, or even with one model and a blank context for the second run.

More ›