> In a way, RAM becoming more expensive has ushered in a new era of forced optimizations, which I'm really happy for

Isn't this mainly Cloudflare's scale though? That's literally also what's in the introduction written as the reason why they are doing the optimization

On day-to-day interactions with my coworkers, I had no problem working with Americans.

At the same time, a lot of them voted in favor of foreign policies that I very strongly disagree with.

Both can be true at the same time, and the former doesn't make me forget the latter. Especially when their president continues to voice threats against the sovereignty of my country.

I think this is speaking from the perspective of nations/governments rather than individual citizens.

For what it’s worth, as an American, I agree with them. The USA is proving to be unreliable, if not outright malicious, to our long term allies.

I always write my assembly by hand because nobody appreciates the craft anymore. It’s all hidden by those pesky compilers which generate extremely inadequate code that could easily be improved by some thinking on the programmer side /s

I do think a better programmer will be a better prompter just like a better compiler engineer will be a better programmer (when performance matters at least).

Agreed. AI doesn’t level the playing field but turns it more in a winner takes all situation. AI is currently best in Python programming from what I hear and this is because it has seen the most Python code.

No matter how much code it has “seen” during training, it still makes mistakes during inference. Therefore AI is most well-suited to write in languages with feedback loops that catch mistakes earlier.

Moreover, no matter how good it is at writing Python, Python is not a language you want to use everywhere. “The right tool for the job” still applies in the age of ai.

Good point. I had great success with Rust indeed, which can catch errors early.

If the CIA could figure out how to stabilize a country that would be nice, but seems they know only how to destabilize countries when looking at their track record [1].

[1] https://en.wikipedia.org/wiki/United_States_involvement_in_r...

> They’ve been really good at getting good press coverage

Same holds for most big corps from what I can tell. If you really do a deep dive into the scandals over the years, you’ll probably see most have barely reached the news or if they did then it’s usually a quite bland criticism like “anti-competitive practices” or a poor HVAC at a certain factory. I think a lot more is hidden than we think

Mojo to me seems like a high level language with some additional support for low level control especially around GPUs. A bit like Rust or C but with more streamlined Python integration and more low level GPU (matrices) support.

> Meta was using porn for AI training or …

Seems plausible. The material contains many hours of highly detailed human interaction plus I don’t expect Meta would have moral objections given their track record.

I suspect their goal is to train a model for detecting adult material, so that they can block it, which isn't necessarily an amoral goal

"Guys we can't sell a VR headset that doesn't do porn good, nobody will buy it" seems just as likely to me.

I don't follow your logic, VR headsets are fully capable of displaying 3D scenes. That's a solved problem.

Just never ever ask how they train the models for CSAM.

I'm not a fan of joking about those efforts. The people doing that work to catch predators and prevent their evil from reaching the general public are heroes.

But would still be a copyright violation

I'm not a lawyer, but in my view copyright violations center around distribution, not necessarily the method by which a work is obtained

e.g. if someone is reciting a poem they wrote in a public park, and I make a recording of that recitation, I have not violated their copyright in doing so, however I might violate their copyright by reconstructing that poem and then distributing it

As in all cases it depends on jurisdiction. For example where I am performers have a right to not be recorded so you may in fact not make such recording, you didn’t not necessarily violate copyright but you violated related rights.

And likewise if you distribute any sort of copy of the recording you made you are violating at least one if not more copyrights.

where I live, anything in public view is allowed to be filmed/recorded, which is how the news can record things on the street without getting a release from those that they filmed, and they certainly distribute those recordings

If the news reporter said "Someone at a park played this 4 minute copyrighted song, and today's top story is just us showing you the recording" I'd bet that would be illegal where you live.

I remarked upon distribution in a comment above

> however I might violate their copyright by reconstructing that poem and then distributing it

My argument is that the recording itself is not a copyright violation, not that distributing such a recording couldn't be construed as one (however I disagree on principle with that being illegal as well, as it puts an undue burden on the exercise of free speech, which should include the ability to accurately and entirely recite what I had experienced in a public place).

Which is why torrenting is the dumbest way to do this, since every downloader is also an uploader. That is what gets torrenters locked up. It is possible to leach only, but its not the default mode, and if this guy was broadcasting his IP address, then he was seeding... distributing.

You expose your IP address even when not seeding, and I don't know if this person was seeding these torrents

What is the moral difference between training on porn vs other genres of media?

The potential output. You train on movies and animation and you get a model that can

- produce your own movies/animation - identity movies/animation from context - identify actors/characters and cross-reference them in other works

You apply these parallels with porn, a more anonymous and intimate setting, and you can see how the kinds of results can be more morally dubious. And that's before getting into if the kinds of porn being trained on is even legal in your given jurisdiction.

> I am reasonably confident that there's essentially nothing that I am better than Fable at

Most things in the real world probably. I’m not saying AI can’t do it, but currently it’s bad. Try send an image of the inside of a broken toaster and how to fix. It’s laughable. Again, not saying AI will never do it, but am saying there are definitely large holes in knowledge.

Well.. as so many things in engineering: it’s a tradeoff. Mechanically, it’s also very important to let the engine warm up and cool down before and after heavy load. Primary reasons for warmup are that the oil is more fluid on higher temperatures and metal is also less fragile when 50 or so degrees above freezing. Cooldown is especially important on turbo engines as the turbo on hundreds of thousands revs per minute and high temperature needs oil to be pumped through the bearings. If you turn it off after heavy load on engines that do not keep the oil running after motor shutdown, there is a chance that oil burns and gets stuck in the bearings, degrading performance or worst case breaking the turbo.

Also for the warmup case, tolerances inside the engine are designed assuming a hot engine. At lower temperatures, the metal hasn’t expanded yet causing also more stress on various parts such as the piston (seals).

And for the turbo I forget to mention that if you leave a turbo engine idle for a few minutes that this allows both the temperature and speed of the turbo to reduce. Many trucks have a sticker in the dashboard saying “do idle engine x minutes before shutdown to avoid turbo damage”. (But also do not idle for extended periods which I’m not 100% what’s the reason. I guess it’s due to diesel engines in some cases getting too cold while idling.)

When a combustion engine is idling, it’s usually not burning anywhere close to 100% of the fuel that it draws, which means higher quantities of unburnt fuel are reaching downstream parts of the engine system (like the exhaust system) that are only designed to be moderately resistant to fuel.