Osmium

Can you expand on this? I was only ever taught the "turn around" explanation.

Overall energy usage, from the perspective that energy usage is the primary immediate harm of AI. This is a thesis, not an opinion.

What would regulation look like if it was based on energy usage rather than capabilities?

A guardrail on mass deployment that is not linked to specific model size or aspects of model performance that are difficult to quantify.

Good, the ad was really disturbing. An ad is just an ad and not the biggest deal in the scheme of things but that was really unpleasant to watch. For me, it was the visual of needless destruction and waste as much as the meta-message.

> “only creating 3 comics a week”

As if this is not an absolutely breathtaking creative accomplishment, especially so consistently and after so many years!

I agree with another commenter that it probably does get easier with experience, but nevertheless impressive.

Or better regulation.

Treating refined sugar as the addictive substance it is, and keeping those products in a separate part of the store rather than at the checkout line. Removing subsidies for unhealthy ingredients, subsidising healthier ones. More education in schools. Better funding and requiring school districts to meet certain quality standards for the food they provide to children. Ensuring all children have the right to a healthy meal. Where previous initiatives have fallen short, critically evaluate why and make them better. Thinking about social psychological factors for our collective mental health and thinking about how this influences our dietary and exercise choices.

There are lots of actions that could be taken that are not just pharmaceutical.

It's interesting that this analysis implies intentionality on behalf of TikTok, and does not allow for the case that its users genuinely enjoy the service and therefore do not want to see it banned.

It was profoundly different. This is all a matter of public record of course, Ars Technica had a very good write up on it at the time it happened.

HFS+ was the legacy file system, from 1998, itself based on HFS from 1985. APFS was a wholly new, modern filesystem and quite radically different.

Filesystem experts please correct me.

For all the reasons that this might not take off, what a thrill that people are trying something new--and it looks really nicely designed too.

I think this is easy to dismiss at first glance, but I genuinely believe they're trying to think about a new mode of interaction. The idea that "the computer will disappear" is probably accurate in the long term. Except for content delivery (reading, photos, movies), most tasks we achieve via computers and phones do not strictly require a screen. It's probably a good thing if computers did a better job of getting out of the way, and stop so loudly disrupting human interactions.

Whether this will be the solution is unclear; the privacy/creepiness angle is still real with an outwards-facing camera. Latency and battery life limitations might be too significant. The cost will be a non-starter for many (it is for me).

But I'm still impressed because there was a vision here. The conversational interface has never worked before for many reasons, but that does not mean it cannot work in principle, or that the ideal implementation would not be spellbinding. I'm glad they're trying. Also, the laser display is neat!

I am really glad arXiv is getting more funding. It is an essential resource.

For me personally, it’ll be really interesting to see if the frontend changes. No doubt there are some important improvements that can be made (a website can always be made more accessible, moderation tools, support for name changes as mentioned, etc.) However, to first approximation, arXiV’s website already seems almost like a platonic ideal. It reminds me of Craigslist. Simple HTML, loads fast, has the information and features you need but otherwise gets out of your way. I love it.

The arXiv team deserves a lot of credit for what they’ve done to get it this far. It’s difficult to overstate how useful and transformative preprint servers have been to science.

Scientists already have well-defined terminology for many of the crystal structures described, the LK-99 is just a shorthand for this synthesis product (and a convenient one at that, for anyone trying to keep up with the latest). I wouldn’t be too concerned!

Part of the reason we’re not using a full formula yet is that we don’t precisely know what it is (or, indeed, if it’s just one thing at all; the linked paper makes a good argument that this will necessarily be a mix of multiple phases if synthesized as-described). Once we know what it is, we can use one of several existing names.

> calling Nature (Nature!) an "online sensational clickbait magazine"

Not far from the truth, talking as someone who is in the field. Unlike Science, which is published by AAAS, a non-profit, Nature is a for-profit publication. They have an incentive not to miss out on something huge so that they can retain their status as the place to go for big results, but this also means they have an incentive towards selecting more sensational research for publication. That doesn't mean that research published in Nature is bad--often it is excellent--and I'm sure their editorial staff sincerely try their best, but they often make quite bizarre editorial decisions (personal opinion).

That said, Nature attracts far more scrutiny than other journals because of their ability to make and break careers, so many people feel resentment towards them as a result. Not all criticism of Nature is entirely fair.

No comment on this particular story :)

I think it's important for discussions about this to distinguish between how things are and how things should be.

How things are is that grad student labor is critical to the running of the university: critical for teaching of undergraduates, critical for performing research to reach deliverables on grants. Indeed, very little research is performed directly by faculty. And grad students are hurting, often at or below the poverty line. Typically a grad school program in the UC takes 5+ years, during which time wages might stagnate around $30k, without raise or career movement that would be possible if they spent those years in industry, despite by all objective measures their day-to-day work being very similar to a research job. Sunk costs make it very difficult psychologically to leave early. Abuse is rampant.

How things should be is very much open to debate. Perhaps a grad program should be as some claim it currently is: a privileged learning role, where you can focus on learning from world-leading experts. Perhaps university institutions should not rely on grad labor as much as they do, and invest more in faculty. Perhaps individual grad students would be better off not in grad school. There are lots of good suggestions in this thread.

How we get from how things are to how things should be is also an open question, but I'd advocate we do it with the least amount of harm to the specific individuals already living within this system. This proposed contract is already weak. The grad students deserve and need a lot more. And increasing the stipend to, say, the $54k proposed would also place incentives on the institution for change.

Finally, without living wages and good contracts, there will always be diversity and inclusion issues in academia. Someone from a wealthy background may be able to slog through a few years with support from their parents, but people from a poorer background will not. It simply will not be an option for them. There is a deeper tragedy here, for everyone involved--the individual, and the institution--for what is lost when so many intelligent, driven souls are not able to participate.

For people who've been around longer, is it wise to take a new job with a company that's undergoing layoffs?