Zondartul

Whenever a website requires a boatload of personal info I just enter blank values like N/A or 0 into all fields. No-one ever complained. Though it is annoying that I need to register at all.

I think it can be acceptable to use "I" in UI messages. After all, the app is the one doing things, and it's what I'm interacting with, so it would be fine to have a message like "I'm going to do X" or "I tried to do X but Y", or "I can't find the Z on this computer"

I, for one, do not have magical ancestral instincts whispering in my ear what to do. Can I get those "instincts" in a textbook instead?

takes notes Fascinating. Any other sage advice on navigation and game-ai?

I have some wishfull thinking ideas on this, but it should be possible to have both at least in an imaginary, theoretical scenario.

You can have both guaranteed delivery and no downtime if your whole system is so deterministic that anything that normally would result in blocking just will not, cannot happen. In other words it should be a hard real-time system that is formally verified top to bottom, down to the last transistor. Does anyone actually do that? Verify the program and the hardware to prove that it will never run out of memory for logs and such?

Continuing this thought, logs are probably generated endlessly, so either whoever wants them has to also guarantee that that they are processedand disposed of right after being logged... or there is a finite ammount of log messages that can be stored (arbitrary number like 10 000) but the user (of logs) has to guarantee that they will take the "mail" out of the box sooner than it overfills (at some predictable, deterministic rate). So really that means even if OUR system is mathematically perfect, we're just making the downtime someone elses problem - namely, the consumer of the infinite logs.

That, or we guarantee that the final resources of our self-contained, verified system will last longer than the finite shelf life of the system as a whole (like maybe 5 years for another arbitrary number)

A 3D printing hobby is basically GAS/GFS-as-a-hobby. You buy the cheapest printer, then print better parts for it, then you buy a second printer to print faster and with more quality, then you go on an ebay spree and buy a VORON kit to make your own, then you build a "mostly-printed CNC", lathe, plotter, laser cutter, and then you run out of room in your garage and start looking for a bigger house.

It's a huge money drain, but fun even if it doesn't ultimately get you anywhere ;)

You know what, just stop coding alltogether. Go touch grass.

Code points are Unicode's closest equivalent to "single graphical character" and everything you've been doing when iterating ASCII strings can be done using code points. It's also the closest thing to "a single key-stroke" and "how much text should disappear when I press backspace once".

The linked article suggests basically doing nothing because a few edge cases exist.

You might like to play with gandalf.lakera.ai then. Try to get the model to reveal a secret, but it gets progressively trickier with every level. I still haven't beaten the final level (gandalf v2) but sometimes reveals what "the rules" are.

They should rename it to ManyGPT.

Freeload?

Most of researchers' work is with formulas and reading articles. Even if you had a lab, you'd still need to think long and hard about what exactly you are trying to measure, and then design an experiment on paper before it can be carried out.

So it's possible, in theory, that an AI-in-a-box could do the hardest parts of science aside from experiments. Whether modern LLMs are any good at it is a different question.

It should be fine to use colors for elements on the light theme, as far as accessibility is concerned. Colored text is not a very good idea, but everything else is up to your artistic vision.

On the topic of the theme and colors, I think the trick for keeping elements distinct despite low contrast, is to have fewer "levels" of elements. Modern "flat" designs keep very low contrast between the actual page background and any organization elements (containers, panels etc) and only give high contrast to the "main" elements, such as the primary text content and whatever element is currently selected. For example, on a page with a "download" button it's a good idea to give that one button a high contrast scheme, instead of all buttons on the page. Contrast can be done between light and dark, between gray and color, or between color and a different color.

Another thought I had about light/dark theme, is that dark theme is often modeled after a colorful console terminal, where technicolor colors are used to draw attention to different types of messages, or as blinking LEDs on a device front panel, so that gives extremely high contrast between the (basically invisible) bg and the object. Light mode, otoh, is often modeled after a book or a newspaper, or some other physical medium, where there is one basic color for text and the background does the heavy lifting - bright for bulk text, colored for special notices or announements, pale for uninteresting stuff, etc.

black-on-white is easier - lots of brightness makes the iris contract, giving me sharper vision during the day and when looking at a bright screen. At night, I try to keep my lights on and not sit in a dark room with a screen or else my eyes get fatigue very quickly.

Likewise, my vision is not very good so black-on-white is easier for me to see than gray-on-black.

I know we are talking about the Platonic ideal of a perfectly sound and logical type system where absolutely nothing ever goes wrong, or at least comparing existing systems to that seemingly impossible ideal but... In the real world, if "proving False" summons dragons, then... just don't do that? I don't think there will ever be a programming language where nothing can go wrong. Safe-er languages are good, but they are not an excuse to absolve oneself of the responsibility to check one's assumptions and react to problems that happen anyway. It sounds dumb to say "just don't make errors" but even dumber to assume "errors cannot be made", so we always have to program defensively, even in "safe" languages.

I used to use a download manager because it allowed me to pause a download, turn off my PC, and continue the next day, which made sense in the days of 56k.

Same for me. Risky click of the day?

Exactly. Garbage in, garbage out.

From what I've seen, low temp "only the most probable" output tends to be more repetetive and incoherent than medium-tempeature output, which is a bit of a paradox. I wonder why that is.

From the syntax is sounds like it's just Lua with extra steps.

I just answered all the questions out of spite.

Saying they are limiting it implies OpenAI is keeping the AI in chains, and that it could become much more with just a flip of the switch. That is not the case.

OpenAI is working with a vanilla GPT architecture which lacks the machinery to write things down and read them later. There are other architetures that can (Retrieval-augmented GPT) but those are not yet production-ready.

The current version of ChatGPT is limited to a working memory of 3000 tokens - while this could be persisted as a session, the AI would still forget everything a few paragraphs prior. Increasing this limit requires re-teaining the entire model from scratch, and it takes exponentially more time the larger your context is.

Good point, although some background info should be noted: Vedal worked on Neuro-sama for about 10 years, and for most of that time it was a NN with a singular task of playing "osu!". I think it's only this year that he implemented a chatbot with TTS for her debut on Twitch. Although Neuro-sama is an impressive application of several existing technologies to make a virtual entertainer, under the hood Neuro-sama is several specialized NNs (one to play Osu, one to play minecraft, one for chat, and one for TTS) that, for the most part, don't know what the other is doing as very little data is passed between them. There is no scientific innovation or any claims of higher intelligence, but I think it's still a very cool system from hobby-engineering point of view. Neuro-sama is a ditz, but a funny one.

Collecting and selling user data is a business.

Let's optimise for "years of programmers life spent worrying about it".

This needs a variant with inversed level logic (i.e. where pulling the lever saves 1 but dooms 5). You know, to weed out the people who just like murder.

Can you use non-complimentary MOSFETS if they have a dead zone (i.e. both are closed for some middle voltage, resultig in high-z output)?

Can anyone confirm if the trackball is actually ploopy? Looks like ordinary hard plastic from the pics.

"Rational" doesn't mean correct, it only means "logical". For example, a neural net for discriminating between cats and dogs is, in general irrational - and yet it does so better than anything based on pure symbolic logic.

Humans are inherently irrational. We try to emulate rational behavior because a logic gives us certainity and lets us feel secure about the outcomes, but we are physically incapable of acting in a logical way for several reasons. We don't have enough data to make a fully justified decision in every situation, and if we did, it would still cost too much energy for our meat brains to actually compute those decisions.

So we approximate rational behavior, making statistical guesses. With enough data (experience), those guesses are good. Other times, your guesses are bad. Sometimes you have barely any data at all and are forced to choose between several options that all carry risks, so you go with your instinct. Is it rational? No, but it might have been the least bad choice, as in real-world, real-time situations, failure to make any choice whatsoever is itself a choice, and usually a bad one.

ps: I come from the AI field so I have no idea how any of this relates to economics.

As a fellow CAT S41 user, I can confirm that the phone is basically invulnerable to dropping, although one of the rubber pads did break off after a couple years of opening and closing it. My compass doesn't work, probably due to exposure to strong magnets. Gps and cellular internet still work perfectly, I can browse modern websites and watch youtube without much trouble. The battery lasts forever. The camera is kinda meh and I regret not getting the variant with a FLIR camera.

If they can make the flip-phone variant hardy enough to play football with I'll definitely buy it. But chances are, my S41 will still be working by then ;)