big if true

Astroturfing just got orders of magnitude cheaper with the advent of LLMs. This, along with spam and advanced phishing are some of the true real and present dangers of this technology. It’s a battle between content platforms and any bloke with an axe to grind, and it’s probably a loosing battle for the content platforms.

Genuine human to human interaction online is going to become rare and tedious. Can’t even imagine what kind of captchas they’ll have to come up with to fool the next generation of multimodal models.

There’s some movement in the rust space, the main advantage being that you can compile to wasm and serve models in any browser. There are several efforts in this direction. This can also be linked to edge-computing, with more services starting to use wasm/wasi etc. There’s a world where you have your entire codebase in rust, and you get to deliver models either to browsers or wasm “VMs” in an edge provider.

Q2) minute 27:43 Tool Use (Browser, Calculator, etc. ) Anyone has links for similar implementations for llama and how is done or what kind of tech/frameworks are used ?

The naive way is to use langchain, but that’s hit and miss for several reasons, and whatever you build will be held together by duct tape and prayers. Alternative frameworks include Haystack and Griptape.

I’ve found that for local models the best tool-usage you can get is by using an advanced control library. This gives you a lot of flexibility in organising the prompts and “helping” the local models a lot. Guidance and LMQL are two such libraries.

Multicast, if your clients can connect to it.

So you list the gnu stuff, and then add “censored”, but that’s not goalpost moving? Come on.

0,1,2 and 3 ALL apply with an apache 2.0 license. Saying this is not open-source at this point is being contrarian for the sake of being contrarian, and I have no energy to type on this subject.

Quoting your own post fron gnu: Take the sourcecode, plug in c4 or redpajama or whatever, pay for the compute and you can get your own product. With the posted source code. I got nothing else.

Nope, RNN without attention, with some tricks for enabling parallel training.

Not looking to start drama here, but I feel we’re moving the goalposts a bit here… Source available and under a permissive license is opensource.

I feel the discussion around training sets is too risky at this point. Everyone is doing at least gray stuff, using dubious-sourced material and I feel like everyone wants to wait out some lawsuits before we can get truthful stuff about datasets.

and the use of AI for political far-right ideology would increase.

Uhhh… And the record didn’t skip when the professor said that? For anyone in the course?

New CEO, who dis?

This suggests that judging a model based on a single benchmark might not provide the full picture.

Duh… This has been a recurring problem with all these “benchmark leaderboards”. It turns out that “training on the testing set is all you need”…

Use the proper tools for the job. Either guidance (reborn just last week) or LMQL are two frameworks that can “enforce” any local model to output a json object.

Emmet Shear

I hadn’t followed everything but a quick google away I found this, that I assume you’re referring to:

PPS: Before I took the job, I checked on the reasoning behind the change. The board did not remove Sam over any specific disagreement on safety, their reasoning was completely different from that. I’m not crazy enough to take this job without board support for commercializing our awesome models.

Interesting. I think we’ll know more as things settle.

oops, derp me. I’ll leave it as it’s funny, but thanks!

So it seems that a bunch of ex-openai core people will move to MS, and possibly work on the same stuff, depending on what MS has access to based on their openai investment deal. They’re most likely to focus on bringing stuff to the enterprise market fast.

At the same time what’s left of openai will put the breaks on stuff and possibly emphasise safety and slowing down. I think everyone will come out of this kerfuffle having “won” what they wanted. No idea how this will impact the overall field, but possibly some players will get a free catching-up opportunity. Maybe.

SFTP into the box and download them would be the most secure way. Setting up a private S3 bucket would be the next best thing.

Ceci n’es pas un subscription.

Intuitively diffusion-based models for code generation make a lot of sense, glad to see people spending time on it. Really curious to see what can come out of it, even if it’s an intermediary step to be used in conjunction with LLMs (i.e. the diffusion model works with pseudocode and LLM translates pseudocode into actual language-specific implementations)

I can probably develop a stripe clone by myself.

=)))))))))))))))))))