The worst thing about addictive app (and website) design is, that companies employ real psychologist to help them design a product that makes kids addicted (and possibly makes them trick their parents in paying for in app purchases). I always wonder if the psychologists don't have any kind of ethics code, because you can't tell yourself you're helping anyone when you design a phone app to maximize usage time.

There are no tools to prompt a movie. Every AI movie you've seen is the result of creating a lot of short (5-30 second) clips step by step, often using image generators for key frames between the sequence is interpolated. If you hope to "prompt a movie" you will be disappointed now and in the future.
While you could try to automate some steps (prompt image generators with LLM and so on) the result would be terrible and uncreative.

Reading should be no problem. You heat the surface to be able to write narrower tracks, but you can read them as usual. The only question is, if the drive will then behave like a printer telling you "I have no yellow, I cannot printer your black and white document!"

And require *very* small blocks in the scandisk interface.

If it keeps scaling like this, the plane is no problem. Currently LLM continue to get better at a large rate. Both getting better and allowing for smaller models with the same intelligence. The question is, if there is any wall to hit. Currently the main challenge seems to be the quadratic attention, which requires scaling up compute and memory, but is no theoretical limit to what a model can do given enough resources, but there may be other limits we did not find yet.

What should be false? You put the archive URL before the link and get the archived version. You still see the link of what was archived in the URL.

It is perfect, useful, for most people to understand and a leaky abstraction. In addition it already lead to a lot of dangerous issues and will continue to do so in the future. We also don't have any better option due to how current computer hardware works.

It would be really interesting to do floating point computations on a device that is not binary. If you've had a seamless voltage instead of on and off and can do computations with it, you're much closer to the real numbers.

Every heard about floating point numbers? Or the Pentium bug? Things may be deterministic, but they are not perfect.

"AI has (among other things) a random number generator at its core. It will produce different outputs, randomly, for the same input. Certainly not "perfection"."

That wrong on so many levels. First you use the term AI, which is a huge field in which many thing have nothing to do with random numbers at all. Let' be fair and put LLM there. Then you assume a LLM must be sampled stochastically, which is one method but not the only one (and in many tasks one uses low to zero temperature). And finally you assume that you sample with non-zero temperature and do not fix the seed even though you desire repeatable output.

A neural network is *literally* a function. It maps one large vector through matrix multiplications to another large vector. If you want (and one does it because one wants it) a random component to get diversity in the outputs, *you* are adding the randomness intentionally.

I wonder if they know how "fair" Spotify distributes the money.

I'd rather say everyone who can afford it should have a day per year where they consider what open source projects they valued most in the past year and then distribute the amount of money they can afford between these projects. That also allows to reward projects that treat you nicely and show the projects that treated you badly that you remember.

No, Nvidia will not die if OpenAI considers if there are also other graphic card companies. No, a proof of concept world model will not create games on its own. Sometimes one thinks people investing at the stock market don't have *any* idea about what the companies actually do when buying the shares.

Yeah, you can type ten synonyms, or get a LLM find the information no matter if you don't find the right synonym. Hey, nobody is taking ctrl-f away from you, but some like to have modern options as well.

You see. If you don't have anything not dynamic, you don't profit from caches.

"Open-source maintainers monetize through documentation visits"

Open source is monetized through documentation visits? Not only most open source programmers know how to use an adblocker, but I also saw very few (thanks god!) documentation pages with ads. What are these people talking about?

Your post is unclear. Do you think he's unemployed or do you think he's got a marketing job in China?

No matter how much AI ads you see now, in a few years most ads will be AI. Too flexible and cheap while becoming very good for companies not to use them. If you're in the part of marketing that is about making graphics and already was only slightly above just using clipart, you should now look for a new job. The people who plan how a good ad needs to be designed to work will still have their job, the persons who photoshop a lemon slice on a glass of cola won't.