Reddit is one of the few remaining platforms where reach is determined democratically. You demean that populous by calling it the singular "reddit", as if it's a single hive mind. But you'll probably get your way, I imagine the billionaire censorship campaign will come for it too eventually. Then you'll only ever get to hear thoughts you agree with, glorious day ey?
Why was that we used to be able to ban environmental toxins such as leaded gasoline without this weird psycho-social-political analysis that has become the fashion in some circles?
One thing I learned from this article is that even though the plastic particles themselves are poorly studied the chemical additives, such as phthalates and bisphenols, are very well studied and are known toxins. So even if the tiniest plastic particles (smaller than the ones your friend studied, that can cross from your gut into your bloodstream), don't affect your health at all, you still don't want to ingest these things because of the other chemicals in them.
Ugh. Ok if you're going to push one-sided propaganda I'll push the other side.
Google forced its researchers to retract an already submitted paper because it undermined its strategic and commercial story around large language models. The "we just accepted her resignation" is just a lie. Google made harsh demands with opaque reviewers that made vague objections, and then Jeff Dean moved very quickly to get rid of Gebru. Other Google researchers reported that they usually got to work through objections, Gebru got no such opportunity. Google showed that AI labs will not tolerate internal research that seriously criticizes technology central to its business.
When math is so divorced from science and engineering that there's no conceivable way that it will ever be applied in the real world then it is just a complex puzzle game that a tiny group of people play. It doesn't really matter much. If the 200,000 line Mathslop proof has no real world application and it doesn't help the puzzle solvers then it is double useless.
The product manager example doesn't make sense to me. Basically all that's happening is that the work of deciding is being centralized in one person instead of distributed across the team. If there are dozens of projects going on Hopper is not going to be the one that's best suited to decide if a project warrants more review. Plus he's going to be spammed with all the little projects meaning he's more likely to make mistakes and let things slip through.
Have they ever said what the difference is between Sonnet and Opus? Are they trained differently? Different architectures? Is Sonnet a distillation? Is it just that Sonnet has less resources for inference?
None of the other labs are doing this kind of long lived two model series.
Not to mention that this bias is a self-fulfilling prophesy. Thanks to all the ideologues that get elected to government by insiting that government can't work, surprise, surprise, government doesn't work.
The last time California Republicans had some power around 2009-12 they used it to manufacture crises and shut down the government. Since then voters have shut them out of power, including by passing 2010's Prop 25 which stripped the minority of the 2/3 veto they used to have.
Photocopiers and printers have included anti-counterfeiting tech for decades, so there is precedent for this kind of thing. And this is addressing a real growing problem:
I don't think that's true about Ds being incapable. For example Biden's Inflation Reduction Act has been very successful at boosting clean energy industry, creating jobs, lowering emissions. Including huge benefits to red states. The problem is a lot of times Democratic solutions, like Obamacare, are too technocratic, or the benefit is too diffuse for people to notice. There's also the issue of the huge growth of the right wing's propaganda power successfully hiding any D achievements from voters.