pbohun

Have any of the Bioware execs even played a game?

This is what happens when you over-optimize and over-automate. You should just be able to plug those freezers into power and turn a knob to set the temperature. There is no reason for them to ever have an internet connection.

It would be much better to use LLMs to create ordinary programs than to use an LLM over and over to emulate something that's thousands or millions of times more efficient.

As an example, is it faster to multiply 2 numbers using a single MUL instruction or multiply matrices with millions (or billions) of elements multiple times? Multiply that overhead by every single thing you do and that's the overhead of using an LLM to emulate a program.

>Overall, nesformer runs 150–290× slower than the real console.

I like these kinds of projects because it helps to show the massive compute requirement between LLMs and ordinary programs.

A while back I saw some people saying we would just have LMMs hallucinate an entire operating system every single frame. I guess technically it's now proven to be technically possible, but it would be thousands of times slower while requiring thousands of times more compute (a multi-million x difference from an ordinary OS!).

Forget all the benchmarks, this is how you know a model is actually good.

This reminded me of a moment from Star Trek V: "I don't want my pain taken away. I need my pain!" ~Captain Kirk

https://www.youtube.com/watch?v=9ABLYnvpAso

Okay, that's some of the most impressive feats I've ever seen. Congratulations to the engineers!

The natural question from this is to ask if there's a way to create conditions where organs "become young" again. Of course, people have been trying to figure out aging for a long time, but this is encouraging that solving aging might be possible.

I think agents have the capability to multiply this problem times 10. It doesn't have to be that way though. I think a skilled person will create great software regardless of the tools they use.

Iryna Zarutska's killer had 14 prior arrests. There are many examples of people having dozens of arrests for violence before they murder someone. There are dozens videos of people taking trash bags into stores and wiping them out because "It's less than $1k". People's cars are broken into and things are stolen (while the cars are occupied!) and the police don't even respond.

There's not some mystery about who is doing crime. They're arrested all the time and let go! If we simply prosecuted the people who are arrested crime would fall by 90% over night.

This is so cool! I mean, we really wouldn't need it if we had 100% ipv6 (no cgnat), but this is the next best thing. I think people underestimate the innovation that could happen if we had trivial p2p.

Russ also has a great blog, with the most recent entry on fast floating-point printing and parsing. I believe these performance improvements have landed in the latest version of Go. https://research.swtch.com/fp

I feel the same way. I think the difference is that new tech used to be an improvement to your life. New tech is there simply to suck life (your attention) and money away from you.

There are exceptions of course, But the general trend is worrying.

An idea I've played around with in the past is to have some sort of law that limits how much revenue a company can make from ads. For example, a law that says "A company cannot make more than 10% of its gross revenue from ads or selling customer data".

This kind of law would wreak havoc for certain companies (and content creators). However, just like prediction markets, ads distort and ruin the very thing they are trying to optimize.

People are finally understanding consequentialist vs deontological ethics. All the worst criminals in history were consequentialists.

I read a sci-fi story years ago that had prediction markets in it that helped markets become more efficient. I thought it was a good idea at the time.

What neither I nor the author anticipated was that the easiest way to make the markets "efficient" was to rig them.

In hindsight it's obvious, but optimism is one of the most powerful blinders out there.

If all you care about is the end product and not how it's made, then you never become a better person. Your capacity for thinking and solving problems will never grow. You will never be an expert. Experts care about the process.

Of course, you may not care, and that's fine. It's important that people not be delusional about this though. A person with no medical knowledge and an AI will never be as good as someone with prodigious medical knowledge and an AI. The same is true for any profession, whether programming, music, writing, etc.

There are some good points here, but I disagree that programs are slow because they are accessible. Computers are ridiculously fast, and it doesn't take more time to display a string in Japanese than it does English.

The problem is ultimately with how the software is structured from the the start. Pressure to release fast, and the fact that most software engineers are not experts, means that nearly all software systems and libraries have poor design. They will be bloated and buggy, and difficult to work with.

Poor design means that development velocity dramatically slows as a project matures. Frameworks ensure that you start development fast, but eventually it takes a week to add a button.

Honestly, I think starting from win32 again would be a breath of fresh air. Also, note that I described it as an option, no one would be forced on Windows Lite. It was my own speculation that it would become popular.

It's kind of crazy that we have a hobby OS that supports Linux infrastructure so that can support Wine infrastructure to support a Windows application. Very impressive!

I looked into writing an http server based on iouring myself, but all the resources I could find said iouring is less safe from a cybersecurity perspective.

Is there a safe way to use iouring for a webserver, or is libuv the better way to go, even though it has less performance?

No, the idea is that the actual key is the XOR of 3 completely independent keys. I think you were thinking of XORing a key with itself 3 times, which would just return the original key.

In the book, there is a cargo ship carrying 1/3 of a OTP. Other two other ships from two other companies are carrying the other thirds. This actually is a fairly decent method of transporting a OTP (I'm assuming there's some kind of physical security preventing tampering).

The book even talks later on about how only using the pad isn't enough, since it provides no proof of authorship or tampering. Vinge did a pretty good job w/compsci in the book.

No, the GPUs used in AI data centers don't have the capability to do normal graphics. They're completely useless for games and content creation. Additionally, B200s can consume up to 14kW of power. There are no consumer power supplies that could power such a thing.

The sad thing is all that silicon will basically be worthless.

I really like Iosevka, and though I haven't used it, I do like Berkley mono. Maybe I'll incorporate this into my list of terminal fonts in the future. It does look pretty sharp.

How many tokens did this port consume?

There need to be some serious regulations around prediction markets. I'm not sure they should even be legal. These "markets" are encouraging large amounts of fraud and manipulation. As far as I can tell, they offer no positive effect to society.

One possibility is that most of the code that matters is being written by hand, while enormous amounts of code are being generated for other things. People are being evaluated on their AI usage after all.

This makes me think of the color scheme of Plan9. I think they chose that color design for similar reasons.

You probably also don't use a calculator because it uses a scary language called arabic numerals. Why write 123,456 when you could write out in english: One Hundred Twenty-Three Thousand Four Hundred Fifty-Six? English is your programming language and also your math language, right?