The man himself! I ordered a Klein bottle from you years ago when you first appeared on Numberphile and to this day it continues to bring a smile to my face every time I see it. I'm proud to be from the same city as you, my favorite Buffalonian celebrity. Your passion and infectious excitement is a true gift to the world. Keep doing what you're doing.

Chat GPT can't follow links or read files. If it came up with the title of the manual, it was because that link and the title are somewhere in its training data.

I hope it's real and we make a major breakthrough in materials science. But if it's not, I'll console myself by knowing that BobbyBroccoli will probably make a really interesting documentary about it.

The short answer to "is this illegal" is "we don't know because it hasn't been adjudicated yet" and that's what courts are for. The argument for it being illegal is pretty straightforward - Microsoft is redistributing code without following the licensing terms of that code. Sounds simple. But the counter argument is that Microsoft isn't actually redistributing code, but using that code privately without redistribution (which is allowed by virtually every open source license I'm aware of) and the resulting product doesn't actually contain substantial portions of the original code. The legal comparison will be that I can use GIMP (or a custom derivative of GIMP) to create a logo for my enterprise application without open sourcing that application _or the custom derivative I used_. That sounds completely unrelated, but think of it this way: The training system that actually creates the model is almost certainly covered by all manner of (probably incompatible) license terms. It's includes, in whole or part, millions of pieces of copyrighted code. It is a derivative of all of that code. But they aren't redistributing the training system. They are redistributing the thing the training system created, which is a product of the code, in the same way that an image is the product of GIMP. Will this argument fly in court? Who the hell knows. My personal thought on the matter is the cat is probably already out of the bag, and was even before this whole AI thing. Legally, if you go on Stackoverflow and pull a 2 line snippet of code or a regex, do you now have to open source the entire 10-million line enterprise application you put it in? Yes, I think so. But everyone (except maybe a handful of companies with extremely tight regulatory requirements for sourcing code) just ignores that and does it anyways.

They'll never be able to pass a bill called AICOAOAMA. What does that even mean? Should have called it the FREEDOMONLINE act and everyone would be onboard.

Step 1: Assign people tasks and document those tasks publicly Step 2: Have people document when the tasks have been completed Step 3: Have someone review the tasks that have been completed Congratulations, you now know what work is getting done. Healthy companies don't care how much time you spend playing candy crush on the toilet if the money they are paying you translates into equivalent demonstrable return value.

"Does it even know who Richard Stallman is?" So one point in favor of the AI then?

This is literally an instrument designed to generate arbitrage. It dips below the dollar (so you buy) and then the anchor yanks it back up to a dollar and you sell. Rinse and repeat until it crashes down. Anyone who is selling these at less than a dollar and buying them back at a dollar is the mark and a fool.

Sanctions aren't a punishment in the way prison is - it's not a retribution thing meant to hurt the specific entity that did the bad thing. Instead, it's a pain compliance thing meant to cause enough damage quickly enough that the entity that did the bad thing is forced to stop doing that thing. As such, they are intentionally indiscriminate to be effective. So that's not really a problem, _per se_ (within the limited conversation around how sanctions are intended to work and not if they really do or not). Removing access to infrastructure that people rely on to live and work, as well as confiscating their IP will clearly cause pain, so that's working as intended. The thing that might be an issue is that due to the very nature of collaborative software, a lot small contributions work their way into a lot of big projects. I think it would be fair to say that the intention of the sanctions is _not_ to inadvertently break half the internet because some Russian developer wrote a function that made its way into, say, React at some point. Again, not assigning value one way or another to the utility or justness of sanctions; just pointing out that this is working as designed, in the same way that it stops random Russians from having McDonalds for dinner or being able to purchase basic goods or listing their services on Fiver.

Did Elon wake up and think "You know what hasn't happened in a while? The SEC threatening to fine me over things I said on Twitter." And then he decided to do something about that.

You will never convince me that "hypernetworks" and "hyperparameters" are real things and not something made up for the new Cyberpunk expansion.

Politicians don't normally manage their own social media presence. His social media manager was being lazy and should have gotten fired over this.

Given that these things were promised in a settlement agreement to a lawsuit about jail conditions, I would say that they are as far as the legal system is concerned. The inmates are legally entitled to these things.

"It's too hard right now" isn't a free pass to commit human rights abuses. When the government doesn't have the resources to do x when I exercise my right to do y, they just don't get to do x - they don't suddenly get the right to tell me I have to not do y so they can keep doing x. Being in jail doesn't change that.

Do you have an actual link to the law? This topic comes up here relatively often and everyone seems to have a different take on what the law actually says. Also, how the hell would anyone know if a drawn image depicts a real person or not? I guess it's supposed to be a catch all for someone rotoscoping actual images but it seems like that would be whack-a-mole at best.