IanCal

It might be that what we consider a basic and very hard puzzle are extremely close together on a more absolute scale. The difference is often for us what proportion of humans can solve it. And the low end of that is still quite high up - animals that can solve things that are very basic for the vast majority of humans are pretty rare and known about, yet are capable of quite complex actions and learning and aren’t wildly different in scale of neurons to us.

Going from 1m to 1T params is also a scaling of a million times. It’s like going from a human brain down to one percent in size in each direction or just a few mm.

It’s not about cookies. It’s about what the site is allowed to do with your data. Cookies are an implementation detail.

It’s baffling we’re having this misunderstanding on this site in 2026 still.

In a way it is - or rather it wouldn’t t change anything if we added more features.

The default is “no”. Without explicit consent you can’t do a lot of things.

You can’t have a default yes, because how can you agree with consent but automatically to everything?

And if it’s a no, are you saying you can’t ask a user for permission to use their data for a specific purpose?

And if you can ask, that’s what we have right now.

Utter nonsense.

There’s no way you could get models as smart by fine tuning. I couldn’t throw a problem like “build a pokemon database with UI to teach my son sql” and get a working system, nice ui, tests (which it iterated on) examples and explanations in one shot.

There weren’t thinking tokens. Maths is now dramatically better, making actual contributions when before they were mostly mocked for making extremely basic errors. Smearing is also something say is very rare in frontier models.

If you think they have barely changed you’ve either forgotten what they were like or not used them more recently, or you’re just being obtuse.

> I've worked with these systems for four years now and they have not meaningfully improved in that time frame.

Not meaningfully improved?! Four years ago was gpt *3.5*! ChatGPT hadn’t been released!

Huge fan of the non-seat/monthly based pricing. Do you need it to be higher to not just cover costs but also make a profit?

> What am I missing?

They’re testing the same models on the same task but different ways of organising the swarms and the new approach works better.

I disagree, you’re optimising many different things and there is not a simple way to compare them. The very concept of Pareto optimal is about this! There is not one uniquely perfect solution to a problem most of the time.

Spending too much time gathering requirements can be a bad business decision too. Those requirements are often not even accurate.

Building a perfect system may not be over engineering to solve the problem perfectly… but what if you don’t need to solve it perfectly? What if it’s better to fail on some requirements to deliver sooner or cheaper or deliver this and some other project?

The Chinese room argument takes it even further as it does not claim the system is unable to anything a human can.

And stochastic parrot isn’t a good description, certainly not any more with how LLMs are trained, but even without that it’s just wrong. They claim it can’t have a world model because there isn’t that in the training data.

You have access to php and a database there.

I think you're mixing up the lyrics and the video.

> It's funny when the nerd is pointing at the old JS logo and star trek's klingon icon

The lyrics are funny. The video is almost entirely converting each near standalone statement into a very direct scene. "I'm fluent in javascript as well as klingon" is the funny line about his nerdiness, and the video is literally just him pointing to those logos.

Key and Peele locking the car doors is one of the few things that happens that's in the video and not explicitly said in the lyrics.

> I won't even try to stop you

You don't need to, because I've not done that. This may be why you're reading my comments as obtuse because you think I'm saying something that I am not.

I initially thought you were mixing up a discussion about the lyrics vs creating a music video for a song. However now I'm not sure if you think that Weird Al wrote the lyrics after the video was created?

> If you want to believe that satire and parody only consist of blindly showing things mentioned in the lyrics

The songs are excellent parody, and then things like the music video for white and nerdy is pretty much a series of scenes showing exactly what the lyrics are saying. Almost every shot is explicitly the thing being said.

> Al mentions that his rims don't spin not because the rims of the car in the frame aren't spinning but because he's a nerd and why would a nerd have spinning rims?

The lyrics were written first, explaining that his rims don't spin. This is because he's a nerd. The music video was then created, and at the point this is sung he points to his stationary rims.

People aren't powerless. They keep voting for people that do these things and/or keep giving them money to do it again and again. That's why they have money, that's why they have power, and as long as people keep rewarding them for doing it they'll keep doing it.

That’s the date of the conference. The proceedings appear to have been published the year before.

There’s two sides to what you’ve said - I’d forgotten lots of the things like trying to lock the car doors, which is an additional point although the scene is still pretty direct.

The other things you’re saying are about the song and storytelling there which is not relevant here. We’re talking about the translation from the song and lyrics to a video. The things I’d remembered about the video were very direct translations from the lyrics, because the lyrics are very clear and to the point.

> Al mentions that his rims don't spin not because the rims of the car in the frame aren't spinning but because he's a nerd and why would a nerd have spinning rims?

Right, a thing in the lyrics.

It does depend on what they’ve introduced though, the player saying they noticed an npc has glowing eyes doesn’t seem like quite the right split (caveat - of course always do whatever seems fun, fun is the point).

This is really interesting. While I know others have posted about fixes I think it’s a very useful thing to see regardless about how well they can follow initial directions and understand what should happen.

I think you could create an interesting benchmark for this, you could likely have models trying to to derail it and another scoring. Detecting when it’s happened shouldn’t be too complex for a model. I understand why LLMs do this, but ideally they wouldn’t.

When my wife has a thing she says she needs to do and I can help, I now ask “do you want this done or do you want to do it?”. I think this is a similar kind of split.

Sometimes I want to cook, that’s a thing I want to actively do. Sometimes I cook because I want to put dinner out, dinner being out is the thing I want and cooking is just a required step.

Sometimes I want to solve a problem, sometimes I want a problem solved.

Here’s the tricky part for me now and I think others are hitting it - when a machine can solve the problem does that devalue the feeling of doing it by hand? Solving a sudoku feels good even though I know I have multitudes of machines in my house that could solve it faster than I could pick up the pen. Games that place a dollar value on some item I can also achieve makes me feel like the effort is only worth $ though. This isn’t logical but I’m ok being human.

So for a personal project do I get the same feeling doing it by hand? Will it feel like I’ve just made my life harder for no reward or will it be a nice satisfying thing?

As the models get so much better the goalposts shift too, the less I direct the less I was needed.

It’s a weird time. Fascinating, exciting and definitely useful - but so much of what I’ve learned is rapidly becoming less and less important for many tasks. Still, I’ve argued for many years that more people should code because it’s such a powerful tool even used basically, I guess I’ve got my wish (and that side I genuinely love, seeing people make things with their domain knowledge and not having to learn exactly how brackets work in order to automate something)

> Taking your logic to its conclusion fishes (or descendants of fish) regularly recite Shakespeare.

I believe that was the point.

Lots of the shots in white and nerdy are literally showing what the lyrics are describing, is there another level to them with references I’ve missed?

The first conference proceedings were published in 1976 which might be where this is coming from.

https://www.abebooks.com/first-edition/First-West-Coast-Comp...

Accidental things and silent things are very different. Accidental means you didn't mean to do it, silent means you don't know you've done it (or might not if you want to get picky, you could notice).

> The weird thing for me is the prompt optimization loop? Why not fine tune the model instead of AI generating the prompt?

Why is it weird to optimise the prompt? Whether you optimise the model is a separate issue.

If you use any closed models you can’t fine tune them, which is another reason for most but here they also fine tuned models.

How were you penalised? Was it just losing the ability to flag?

If however you see things isn’t lining up well with how they want things flagged, it makes sense to remove that from you. Unless there’s more punishment attached this seems very sensible regardless of how genuinely you believe in things.

No, GDPR spells out in much more detail what consent means - and burying things in a EULA doesn’t work.

I found that telling Claude I was going to bed meant it continued on making assumptions for longer rather than asking lots of questions or stopping part way.

The prompt is interesting, I can’t help but wonder how many times it was run and extra instructions were added (don’t return if x, etc).

They aren’t allowed to do whatever they want with your data, there’s strict restrictions on requiring consent for things you want to do with user data.

If you're measuring it as "how many people own superyachts" that's probably different from how most people want to measure "how well is the planet being run".

> And she's probably more efficient in spending it than the practical alternatives.

Why? This seems like an odd statement. Why is a singer more efficient at doing this than a team of analysts?

> Yes to all of the above. I assume.

I have to point out that you are likely looking at the world in a drastically different way to most.

You think that

1. We should be pushing for more superyachts, as an immediate measure of efficient allocation of resources

2. Taylor Swift is being rewarded for efficient allocation of capital and not her songwriting, singing, etc.

3. Taylor Swift, who has been singing for many years at lower amounts of wealth, would stop if her fortune dropped and she had to release songs in order to gain more wealth

4. Taylor Swift is making music because there is something she wishes to purchase that she is saving up for. She does not wish to make music, and is putting up with it to finally afford something.

I think most people believe that musicians who are extremely popular for their music would continue to make music if they could live an extremely lavish lifestyle without a care in the world for money but their net worth didn't increase. I think that most people would prefer some measure of how well things are going that look at their own life, or how many people have food.

> Also, if that was the goal, wouldn't it be better to tax (or break up) the corporations rather than the shareholders? It comes out of their pocket either way,

If you do it via the corporations you're saying it should be linear (so everyone taxed flat). Doing it via the shareholders allows it to be non-linear.