I love the project. Will be excited to participate when it comes to MA.

They lay out the case clearly here...and I agree. This was my one-sentence take back in 2022: https://twitter.com/andyreagan/status/1506294505930203151

> hot take: large language models (looking at you, GPT-3) are just lossy compression

> Training these models with extra data turns out to be incredibly expensive and relatively ineffective.

I can see that it's expensive, but have you tried it for effectiveness?

BTW, your approach is very cool here.

I've only done two experiments with it myself - training a tagging model on my blog's content and using that to suggest tags for untagged entries - and I found the results very unimpressive fur both a cheaper and the most expensive model.

I've seen a few other people suggest that time tuning GPT is unlikely to give better results than just feeding the regular model a few examples in a regular prompt.

I've yet to see anyone talking about a GPT3 fine tuning project that went really for them. Maybe I haven't looked in the right places.

what town in MA are you? I'm in Shutesbury, and we're building a new library in the coming years. would be great to do this, and have an example to base from.

Oh interesting! We wouldn’t be messing with any bits, but rather responding to dns queries. The opt out or simply setting your own dns server would mean we’re not forcing anyone to use our dns “service”.