version_five

That seems awfully harsh, the comment I replied to had done the usual "let's find some reason this is discriminatory" thing and I was pushing back on it. I'm happy to stop commenting.

Right, much better for everybody to be screwed in the name of equity.

Lots of decisions in the last 50 years have made "the system" more fragile. This makes me realize that outsourcing trust is another one. This would never happen now because businesses have no connection to their customers, either a payment provider or credit bureau does the vouching which becomes meaningless when things don't operate as usual and human judgement is needed

I have an emergency radio with crank for power outages. I know it probably takes a lot of cranking to charge a phone, but I think that would reduce complexity and make it more universal than having to occupy a specific kind of bike sharing bike

The Haskell version is the best one imo: https://m.youtube.com/watch?v=ADqLBc1vFwI

Is there a precedent for locking people up for being victimlessly deviant? Maybe drug use?

That's not logically consistent. Someone selling solutions to a nonexistent problem is a swindler.

I think I mostly agree, but what kind of credentials or track record would be automatically credible in "AI safety". There are conflicting views and not an agreed upon body of practice for what it means, so yes there are charlatans but also just different viewpoints. Arguable the charlatans stick to some of the more mainstream stuff (if your model passes a "bias" test it's good to go). Whether someone is good or bad at AI safety is still much more of a holistic assessment. Whereas in established industries it would be obvious.

"AI" doesn't work reliably enough to be used for most stuff. Safety from that perspective is real.

There are grifters who talk about existential risk, which is pure BS and people who play up identity politics and call it safety.

Why do you have to share your contact information to download the model? Otherwise it is licensed under cc-by-sa (gpl-style, changes must be shared) which is a good choice.

(Sorry just rereading and I meant demodulation, autocorrect got it the first time)

To receive AM you just need a non linearity and the x^2 term iirc will do your rectification. So any element that electrically provides some kind of nonlinear behavior (ie is not a pure R, L, or C) should give some sort of modulation.

It's complicated. Startups are always dysfunctional, and it gets magnified by people who aren't used to it. What they say on these sites it almost certainly true, but it's relative, not absolute, so it's hard to get a sense of how outside of norms it is.

Like if someone says none of the execs know what they're doing, it's going to fall apart any day, are they a new grad, someone who's only worked at big companies with everything formalized and an established business, or are they startup veterans. You don't really get that context.

So the best you can do is try to judge it based on your experience. If you find those comments concerning, maybe it's not a good place for you. If you think "yeah everyone says that" then maybe it's ok.

Your company needs the maturity to manage that engineer. Companies with products have customer success and sales etc and their own engineering management and product hierarchy to translate your end user issue into a software fix. That doesn't happen automatically by staffing and engineer in house. How would a company even figure out the profile they need?

Alignment is also a distraction, its OpenAI marketing and something people who don't understand ML talk about, not a serious topic.

Like I said, discussing model politics have a place but bringing it up every time a model is mentioned is distracting and prevents adult discussion. It would be like if every time a company comes up, the thread gets spammed with discussion about the worst thing that company has ever done instead of discussing it in context.

To be fair, it's a stupid distraction from discussing the model. If every thread just turns into politics it would not make for good discussions. People can start threads about specific ideologies that language models have and they can be discussed there (and have been). Bringing that up every time a model is discussed feels off topic and legit to flag (I didn't flag it)

Edit: but now I see the thread has basically been ruined and it's going to be about politics instead of anything new and interesting about the model, congrats everyone.

> This reminds me of a comment elsewhere I also replied to today

Right, looks like you made fun of / were condescendingly dismissive of my comment in another thread, I wouldn't have replied here if I'd realized you were the same person.

I briefly played with this a couple months ago and it looked promising. Have there been more developments?

Lol is that the strawman that people have come up with, that not liking metas "only do what we allow" license must be anger about competition?

No. A good parallel would be if Microsoft (say) wrote their own linux clone that was compatible but had some proprietary enhancements that made it desirable over open source distros. The only catch being, it wasn't gpl licensed (they wrote it from scratch) it had a proprietary MS license that says you can only use it for things MS approved of, and are using it at their pleasure, to be revoked at any time.

People don't care about the license, they call it open source and move away from gnu/linux to the proprietary MS version, and now we're only doing what they allow us to.

That's exactly what's happening in the ML model world right now, but people are happy with the shiny models Facebook lets them use so they say "what's the big deal".

The license is a wedge that's destroying the meaning of open source, it's worth complaining about, and it's evil to have done it that way. I would have preferred a commercial license that was at least honest instead of a scorched earth ecosystem takeover like they've done. In a sense it's an extension of the big tech "provide something notionally free that's too good not to use and use it to destroy competition" model.

I'd be curious if you have any hard data about use. Mine is anecdotal too, but I see that llama.cpp is the very close second highest starred repo with llama on the name, after meta llama. Additionally, all the HF models seem to have ggml / gguf quantized versions . I'm not aware of a competing format for quantized models. There are also python bindings which are used in a lot of projects. What is a competing framework, other than pytorch, that's getting more use? Or is it all just pytorch (and some hf wrappers) and the rest is a rounding error?

See https://github.com/search/advanced there are various date options

Github has pretty varied filters, you can just search llama and sort by stars or recent activity etc. It doesn't look like it's possible to exclude python, but doing so might get you the "edge" ones. (Except they usually have python utilities for converting pytorch models)

They didn't mention llama.cpp or show it in their picture, that's hopefully an oversight, it feels like a major slight. It's a (the?) major reason for llamas popularity.

I have mixed feelings, llama is great but it's perpetuated it's shitty license. They could have done so much more good if they'd used gpl style licensing, instead they basically subverted open source, using an objectively good model as leverage.

Alexa and voice assistants, chatbots etc are useless because they are neutered ability-wise, not really because of a tech problem. So I doubt generative AI will change anything.

Short answer, possibly, especially if it was say part of a rag system or some other architecture like that. There is room for more. Nothing particularly special about llama.cpp as the llm back end though. It's optimized for running on lower-end hardware which matters less if you're serving models as a service. But it has many strengths.

Llama.cpp / ggml is the open core of ggml.ai founded by GG and funded by the guy from github, so they have some monetization plan for it.

Otoh the superficial social justice warrior types self-disqualify so you don't have to put up with their politics. There are some reasonable objections - say if your job was to make a chemical weapon more deadly - but lots of good military tech that keeps people safer. If someone wants to paint it all with the same brush and manipulate people into clicking ads or algorithmically denying insurance or whatever other tech thing instead, I'd be happy not to work with then.

See previous post flagged to death about the same time this new one was posted: https://news.ycombinator.com/item?id=37667852

Hahaha quite possible, I really didn't think it was satire and it's indistinguishable from a lot of serious comment I see, especially wrt covid.

Edit: looking at the op's recent posts I also found

  If you write critical software or drivers in ‘23 and aren’t using rust you should be sued for negligence.
  Take your “safe” subset/library of C/C++ and put it in the trash already.

So I think this may be a professional pursuit

Lol, you're experiencing what's called "crimestop" the automatic tendency to stop thinking when anything encroaches on the orthodoxy you've been fed. Why does it "sound harmful"? Because it contradicts the covid propaganda that's been pushed on you, or because there is a real first principles problem with it? If it's the latter, cool, let's hear it. If it's the former, I encourage everyone to try thinking for themselves.