If you can't be bothered to support your statements don't expect anyone to be bothered to care about what you have to say.

See above.

Insider is not a weasel word. The president providing advanced notice of material policy decisions to people who pay him so they can exploit the markets is of course insider trading.

For insider trading to be insider trading it has to be an insider doing the trading or disclosing and it has to be material non public information. Your example fails both tests.

Given your example is not insider trading conclusions drawn from your incorrect assertion are themselves also invalid.

Do you believe it is possible to ever focus on a given topic without being required to evaluate every other instance of any other situation that could be remotely construed to be similar having ever occurred in human history?

If you were in court defending a client from a murder charge would you chastise the court for their failure to enumerate all the other murders who murder too? Would you drone on about how it is unfair to single out your client while asserting the court is suffering from MDS (Murderer Derangement Syndrome)?

I take native Hawaiian religious objections seriously. These are people who take a fatalistic "Pele will take what she wants, it's pointless to try to stop her" attitude toward lava flows running over their houses that seems utterly insane here in Iceland. We'll buy up every available water pump on the international market and walk on active-flowing lava fields to save a harbour for a couple thousand people. We'll keep running tourist sites right next to a volcano that erupts every couple months by building multi-story-high berms around it with giant bulldozers and cooling anything that tries to overtop to make the berms even higher. And it totally works . But Hawaiians will sit back and watch incredibly slow-flowing flows with easy diversion paths take out everything they hold dear. At best they'll point to some primitive failed attempts in like the 1940s as a reason to not try, but then when pressed, it often comes back to: "Pele will take what she wants to take."

Another example: geothermal drilling on the Big Island. Massive protests. Massive resistance, way beyond the token NIMBY resistance here. Meanwhile Hawaii continues to generate most of its electricity from oil, at great cost. Hawaii is the only other place (outside of Iceland) to have accidentally drilled into a magma chamber. In Hawaii they treated it as a major incident and immediately shut down the well. When we did it? Our reaction was, "WOW, we can produce REALLY hot steam from THIS well!", made it into a production well, and started planning to drill more.

I guess one major difference here is that in the early Icelandic cosmology, volcanoes weren't the product of a high-ranking creator goddess, but rather, Surtr, who was practically a demon and was fated to destroy the world. It's good to resist Surtr!

Except his complaint is not.

Yes, it literally was. The main public complaint of the weavers was that they spent their entire life learning their art, and then their bosses come in with a machine that they teach to replicate the patterns that they developed, churn out cheap "slop" versions of them, and fire them.

And here we're talking about profits, despite their huge scaleup expenditure. Theirs and OpenAI's margins on serving inference are ~40%.

It's almost impossible for a company undergoing a massive scaleup rate to turn a profit, let alone growing 10X annualized for 3 1/2 years running, but Anthropic pulled it off.

Woke is what happens when you go to see a movie and instead of being told a compelling story woke ideology is proselytized

Recursion (n): See "recursion"

It's hilarious how obsessed you people are over a term that you can't even define and that none of you agree on. As far as I can tell, the only thing "woke" means today is "I'm A Right Winger, And This Is Something That I Personally Hate". I've seen people describe concepts as unrelated as "the concept of a metaverse" and "removing copyrighted materials to avoid lawsuits" as "woke".

An example of woke AI slop is Everything Everywhere All at Once

Um. What?

That movie was filmed from January to March 2020. You think AI video was a "thing" then? The movie premiered in March 2022. The infamously awful "Will Smith Eating Spaghetti" AI video was what AI video looked like in 2023. AI video was not "a thing" back when that movie was being made. What you actually hate is that movie's CGI.

Secondly, since when is "Jackass"-style "bro" stuff** famously leftist? Sexual gags are stereotypically "bro culture" and stereotypically disliked by feminists in particular and leftists in general. Don't get me wrong, you'll also find some people on the left who like that stuff, but the right tends to buy into it more (men tend to vote more rightward).

As far as I can tell from this, what you actually hate is (A) CGI, and (B) bro culture. Based on this, what you should actually consider yourself is an AI-supporting leftist.

That entire video sits firmly in uncanny valley.

I've shown it to about a dozen people in person without telling them it was AI first, and not a single person noticed. It's only "uncanny valley" because you've been told in advance that it's AI.

It's like my favourite social media hobby: when I encounter anti-AI people, I post things made by humans, generally masterpiece paintings or award-winning photographs, suggest that it's AI, and then watch them tear down how awful and fake and soulless it is. It's pretty hilarious ;)

So what you actually want is more taxes on the wealthy? If so, just say that. Don't hide behind arguments about copyright or that sort of stuff when that's not your actual motivation.

but while machine looms only put people out of work,

Machine looms, along with all of the other equipment of the industrial revolution, led to a massive boom of prosperity across the population as a whole. Machine looms in specific made clothing far more affordable to the average person.

One can criticize the politics of the early industrial revolution for not doing enough to funnel the wealth down quickly enough or to regulate the robber barons, but it was overall a massive benefit to quality of life of society as a whole.

LLMs are doing that plus going into systems that abuse them in lots of other ways.

And the industrial revolution made it easier to make bombs and bullets. That doesn't mean that we should all go back to medieval Europe where the average person lived on the edge of not affording their next meal and society had to dedicate almost all of its resources to basic production rather than things like education, medicine, science, etc. Or reject even more technology and ultimately end up back as paleolithic hunter-gatherers where 40-50% of people died before age 15.

This is a load of BS. Everyone gets a vote. Technology is merely a means to an end.. a tool to get shit done. What constitutes progress is both subjective and distinct from technology. They are completely different concepts and should never be confused.

For example technological means to nullify the strong force using $20 worth of materials from your local electronics store would certainly constitute "technology"... yet unless your idea of progress is randos with nothing better to do vaporizing entire cities such "technology" would constitute regression not progression.

Likewise if all AI ultimately brings is aggregation of power into ever fewer hands enabling tyranny and devaluing everyone there are likely to be many who would not consider this to be "progress". Technology does not in and of itself constitute progress.

Copyright concerns only the performance or (re)production of works and derivatives. It expressly does not constrain any other use, facts, styles.. etc embodied in a work including learning by a person or machine from that work. Training AI is clearly transformative and so there is no copyright interest. While it is possible to violate copyright during training by for example making "fixed" reproductions of copyrighted material in the training workflow it is also possible to train an AI on copyrighted works without having done so. Thus the argument is effectively reduced to quibbling over pointless technical modalities.

What I'm waiting for is people to start training up models against known data sets in a bit for bit reproducible manner. When the aggressive derivative work complaints roll in from IPR groups you would have something that can be taken to court to demonstrate production was done without knowledge of infringing work. AI use then becomes a means of reducing legal liability.

Firstly, US courts have no jurisdiction in most places, so what.

The EU rulings thusfar have been even less restrictive - for example, the LAION ruling, which actually allowed scraping (within constraints).

And if even the US courts are ruling that training with US materials is legal, do you really expect a foreign court to ban training with US materials? And you actually expect that to do anything?

Wishing it wasn't doesn't change the fact that stuff written online has automatic ownership

First off, this is a myth. It's not "everything" that it is written that has automatic copyright, it is everything that is written that involves more than a de minimis standard of creative effort that gains protection. There is a huge range of things that fundamentally lacks copyrightability. It's not "everything".

Secondly, copyright does not make the owner a God to their works. It grants a limited set of rights. The boundaries of these rights get ejudicated by the courts. The very courts whose rulings we have been discussing above. "because...The Law" is what gets determined by the courts.

But open source AI is no such thing

Seriously, please stop what you are doing right now, go to Huggingface.com, and start browsing. This is a maddeningly ignorant statement.

Fifthly, do I personally care about open source AI? No.

Then you must love the idea of giant corporations and billionares controlling things far more than they even do today, because that is the future you get otherwise.

but I view current state of the art AI as statistical toys

LLMs work by chained logical inference.

What exactly do you mean by "woke" and "DEI"? Like, contains black people in more than just the background? Has two major female characters that talk to each other about something other than a male character? Has people who aren't straight cis anglo-saxon protestants?

You massively overestimate the capabilities of state of the art AI tools.

A badly done movie by a talentless hack that got a lot of press hype does not a technology define.

First off, a lot of AI use in movies is not "generating the movie". It's special effects. Replacing backgrounds, adding in effects that look better than CGI at lower cost, "tweening", fixing filming errors, etc.

Secondly, you can do a lot better than that terrible demon movie. For example.

The problem is that first off, by and large, it doesn't. Some cases have but perfectly legitimate alternatives - for example, the courts found that Anthropic copying material they haven't bought is illegal but that it's perfectly legal for them to buy used/surplus books en masse for pennies on the dollar, scan them in, and train on that.

Secondly, even if you take an ultra-restrictive view on copyright and heavily strengthen copyright legislation, you're not getting rid of AI. You're just handing control of it over to e.g. Disney, Google, etc. Do you honestly think that's better? Also, what you're killing is free, open source AI. You think that makes for a better world?