This was only a matter of time. It'll be interesting to see how this will unfold. Potentially it could turn into lawsuit cases being built up, it could also mean content producers get a cut down the line... of course could be both. Since FOSS code also ends up in training those models I'm even wondering if that could lead to money going back to the authors. We'll see where that goes.

Definitely this! Major FOSS projects should think twice before giving their street creds to such closed systems. They've been produced with dubious ethics and copyright practices and since they're usable only through APIs the induced vendor lock-in will be strong.

There's the carbon footprint but of course there's also the water consumption... and with increased droughts this will become more and more of a problem.

This is impressive results. Clearly much less artifacts than on previous such models.

This is important. We need truly open generator models. This can't be left in the hands of a few with only API access, especially since they lack basic transparency.

I'm still doubtful about it but maybe I'm wrong so a counterpoint to my own opinions. Of course this is a purely productivity standpoint in here which overlooks my main concerns with how this is currently deployed and used.

Clearly aims to demonstrate the superiority of their specialized hardware for training. That said it's nice to have proper open models available (architecture, training data, weights... it's all in the open).

Now this is a truly impressive technology! This will make facial motion capture a really smoother process.

Now, this starts to become interesting. This is a first example of trying to plug symbolic and sub-symbolic approaches together in the wild. This highlights some limitations of this particular (quite a bit rough) approach, we'll see how far that can go before another finer approach is needed.

Training sets are obviously already contaminated... now it'll be a race of hiding such mistake under the carpet with human interventions. That'll be a boon for misinformation. That's what we get for a useless large models arm race.

Now this is a properly balanced piece which looks beyond the hype. Usable yes, if hallucinations don't have a high impact. Can the hallucinations be solved? To be seen, I personally have my doubts with the current architecture... at least banking it all on human feedback is being very naive about the scale of the task.

The climate constraints are currently not compatible with the ongoing arm race on large neural networks models. The training seems kinda OK, but the inferences... and it's currently just rolled out as shiny gadgets. This really need to be rethought.

Now, this is interesting research. With all that complexity, emergence is bound to happen. There's a chance to explain how and why. The links with the training data quality and the prompts themselves are interesting. It also explains a lot of the uncertainty.

The lack of transparency is staggering... this is purely about hype and at that point they're not making any effort to push science forward anymore.

Well, people asking relevant questions slow you down obviously... since the goal about the latest set of generative models is to "move them into customers hands at a very high speed" this creates tension. Instead of slowing down they seem hell bent at throwing ethics out of the window.

This is an excellent piece. Very nice portrait of Emily M. Bender a really gifted computational linguist and really bad ass if you ask me. She's out there asking all the difficult questions about the current moment regarding large language models and so far the answers are (I find) disappointing. We collectively seem to be way too fascinated by the shiny new toy and the business opportunities to pay really attention to the impact on the social fabric of all of this.

When they changed their statutes it was the first sign... now it's clear all ethics went through the window. It's about fueling the hype to drive money home.

That's a good set of questions to ask ourselves when in contact with a product claiming the use of "AI".

Interesting and surprising limitation. This makes a lot of sense when you think about the set of images used for training though. Also says something about our own art history.

Or why they are definitely not a magic tool for programming. Far from it. This might help developers a tiny bit, at the expense of killing the learning of students falling for it and the creation of a massive amount of low quality content.