My experience has been that out of the box, AFM has a hard time with tool calls. Even with pi, I had to trim down the system prompt and update the tool descriptions to make it succeed in even the simplest of tool calls.
I personally enjoy the news feed part of it. It cleanly summarizes the major global news stories across several sources into 3 bullet points, and it doesn't show you any ads disguising as stories (eg product reviews, tips, etc)
Where do you draw the line, and where does someone take final responsibility? I don't have an issue with allowing AI to do everything with limited to no human verification for hobby projects. I have an issue with allowing AI to write and ship medical device software with no human review. What's the middle ground?
You still take the full responsibility, I don't think leveraging AI reviews removes any layer of responsibility. The challenge is building the feedback loop (or the agent that builds the feedback loop) to make the review process actually catch what it needs to catch.
I think the line is wherever you feel comfortable with pushing the changes yourself. You feel like adequate testing had been done, basing it on previous testing periods or how successful AI runs have been.
Whether you or AI wrote the code, you decide when it's safe enough to put your name on it. Like it's always been!
> A galactic algorithm is an algorithm with record-breaking theoretical (asymptotic) performance, but which is not used due to practical constraints. Typical reasons are that the performance gains only appear for problems that are so large they never occur, or the algorithm's complexity outweighs a relatively small gain in real-world performance. Galactic algorithms were so named by Richard Lipton and Ken Regan, because they will never be used on any data sets on Earth.
It's more that it demonstrates that it's possible at all. We now know that the floor isn't an exponent of 2, which makes pursuing further improvements way more valuable.
Yes because now it opens the door for future algorithms to chip away at that exponent where as in the past it may have seemed that an exponent of 2 was the floor.
I don't really get it, seems like a gate keeping exercise for the academic community.
Of course OpenAI and others will hype any LLM led/assisted research results, and the media will find juicy headlines to write about it. But in the end they are publishing findings, and the academic community can gauge the developments and decide what they want to do with them. To me that seems like research working as expected.
> That’s assuming that you actually get those 50 people to work without blockers, stepping on each other, or other coordination issues. The coordination complexity alone is astounding.
This is a question of exceptional management, which needs to be present both in the Claude and human cases, and is scarce. Not everyone given the Claude tokens would be able to deliver the same result.
This seems like a great way to facilitate data gathering for improving LLMs coding performance.
If previously you needed to take action 1, 2, 3 to go from state A to B, all you saw was the change from A, B. Now you see intermediates 1, 2, 3 and can train the models to skip straight to B with the added context of the intermediate states.
Yes. Make sure you’re not using the Gemma sparse models since they don’t have a small model to use. Also I removed all the image models from the workspace.
All 4 gemma-4-*-it models, regardless whether they are dense models or MoE models, have associated small models for MTP, whose names are obtained by adding the "-assistant" suffix.
I've gotten it to work with other models. They've got to be perfectly aligned usually, in terms of provider, quantization etc. Might be a bit before you can get a matched set.
reply