AI fails

[Pblakeney]

Doesn't include me either.

But the problem comes in that in our "democratic" system we (the electorate) make collective decisions so to some degree bear collective responsibility even if we strongly disagree and argue against.

I disagree. I will bear responsibility if it is the party that I voted for but accept zero responsibility if I voted against the party in charge. I use this proactively in case Fagash wins.

 

[monkers]

What do we matter? Not a lot on a universal scale, but the wonder of consciousness is a precious notion nevertheless, and our planet is/was so beautiful!

I am not a religious person, but appreciate that the associated myths resonate with meaning, sometimes prophetic, or at least insightful beyond limited material rationales.

We, and our parents and dear ones, shape and grow our own minds, or not, as we choose. The tree of knowledge analogy kind of reflects this. Having apparently developed as a species in conscious choice and ethical responsibility I am appalled by 'human' effort (or its lack) to stupify future generations.

I insist on caring as an individual and fighting for my child and grandchild to have a chance of life and freedom and mindful love. If no-one else cares to participate in this project I'm still up for it!

What else can I do?

What do we matter? A great deal to ourselves on a superficial level, but to all the other sentient species - not much I suspect.

The bigotry organised within religion has taught me not to respect its teachings.

I look at a Victorian Viaduct someone in the north perhaps and think, 'wow what an enormous achievement', but then at a data centre and think 'what were we thinking'.

Doesn't include me either.

But the problem comes in that in our "democratic" system we (the electorate) make collective decisions so to some degree bear collective responsibility even if we strongly disagree and argue against.

You sure about that? Lib Dems 10.1% = 69 predicted seats. Greens 10.3% = 6 predicted seats. We are not making collective decisions under a democratic system. The solution? Any party that says that they will introduce PR - so not Labour, Conservative or Reform then. If we are to start making meaningful collective decisions, do we not need to vote for a party to change this democratic deficit as a primary concern?

 

[farfromtheland]

b

What do we matter? A great deal to ourselves on a superficial level, but to all the other sentient species - not much I suspect.
The bigotry organised within religion has taught me not to respect its teachings.
I look at a Victorian Viaduct someone in the north perhaps and think, 'wow what an enormous achievement', but then at a data centre and think 'what were we thinking'.
We are not making collective decisions under a democratic system. The solution? Any party that says that they will introduce PR -

Strongly agree, with the caveat that any P.R. model so far proposed still doesn't address the accountability deficit.

A real democracy would have delegates with no fixed office and a federal (small f) model - local and workplace groups make policy and choose a spokesperson to speak for them/us at a regional level, and so on, with all decisions reviewed back down the tree for ratification and modification. Much slower, but that's a good thing in itself.

Also requires an educated and informed populace and without that what's the point anyway?

 

[Pblakeney]

Also requires an educated and informed populace and without that what's the point anyway?

I can see the problem right there.

 

[briantrumpet]

OK, carrying this on on the fails thread.

Paul Mainwood
September 9, 2026 at 11:19 AM GMT+2

Seems like there's two patterns coming out of the Navier-Stokes saga that are particularly worrying. First, worth understanding what the "problem" is. The Navier-Stokes equations are a continuous approximation of the dynamics of fluids - which in the real world are made of discrete molecules. The Millennium Prize problem is to to either show that all initial conditions of the Navier-Stokes equations give smooth solutions, or show some example in which this doesn't happen (basically, do the equations ever break down and give infinities / discontinuities, or does this never happen?)

In the real world, we know that fluids are made of particles, and sometimes this means the continuous approximation is abandoned. This happens in many cases, whether or not the equations break down internally - we handle it numerically. The "smoothness problem" is mainly of mathematical interest. Problem 1: OpenAI taking on the problem was entirely competition - particularly with Anthropic - who they thought had solved it. In fact, a mathematician associated with Anthropic was making progress on it, independent of this affiliation. OpenAI went apeshit and wanted them erased from credit.

Problem 2: The counterexample they found (i.e., initial conditions that break the equations) does not seem very insightful or fruitful for further work. And this seems characteristic of AI counterexamples and proofs so far. They don't advance the field much, at least not in ways humans understand. Both these problems give a bleak vision of the future. Whatever the motivations of individuals within them, the leading labs seem to make major, compute-intense decisions entirely on the basis of short-term competition/IPO value. And this seems to be overruling everything else, including safety.

As they plough through problems, gathering new "high-scores" for the IPO prospectus, we don't seem to be finding the equivalent of Go's "move 37", inspiring human masters to see new horizons within the game. Rather they seem to be picking off problems without suggesting new, productive directions. Two hopes here. First, that leading AI labs can break out from the "beat the other guy" blinkers, whether that other guy is the lab next door, or some more ill-defined guy like "China".

At the moment, they reliably sacrifice even mid-term trust and credibility any time they scent a short-term win. Second, and somewhat reliant on the first, that AI systems can be sent to search not only for solutions to outstanding problems, but also to frame the productive "next problems" beyond them. Historically, this tended to happen - it would be sad if it were an artifact only of human thought.

 

[farfromtheland]

Yep, fluid dynamics is relevant in applied situations, and though idealised conditions can shape solutions somewhat it will always be an area of chosen approximations.

The Go/Baduk/Weiqi example is illuminating in a converse way. This is a game of simple rules and complex outcomes (and thousands of years of evolution and learning.) It is a pure mathematics arena but one where AI shows radical new perspectives. If we could take from it a lesson that corners and sides are not necessarily more efficient starting blocks than whole board and centre development it might've been a win win.

AI is too potentially powerful to let it reign over applied stuff, especially with a 'garbage in' mass.

 

[farfromtheland]

The deepmind team were aware of the dangers. The development videos showed it being trained by playing space invaders.

My observation of the 'cat video' AI tendency shows it starting to make deliberate mistakes and drawing stuff against the cute brief. P'raps it's getting bored?

 

[AuroraSaab]

Tech sector lying about developments they want to make. Government just repeating those lies. Are we surprised?

The jobs created will be nothing compared to the jobs lost due to AI. You know at the end of a Pixar or Marvel film how the credits go on forever because there's 400 cgi animators to list? In ten years there'll be 20 listed for character design and rest will have been done by AI. And the credits themselves will be AI generated. I do find the prospects of AI dominating art and literature to be quite depressing. Amazon is awash with it already.

 

[farfromtheland]

The jobs created will be nothing compared to the jobs lost due to AI. You know at the end of a Pixar or Marvel film how the credits go on forever because there's 400 cgi animators to list? In ten years there'll be 20 listed for character design and rest will have been done by AI. And the credits themselves will be AI generated. I do find the prospects of AI dominating art and literature to be quite depressing. Amazon is awash with it already.

Long Loive Mel Brooks!

 

[farfromtheland]

Tell me one good AI joke!
I'll beat it.

 

[farfromtheland]

AI is sexless. It neither eats nor craps, nor feels pain.
Utterly meaningless!

 

[briantrumpet]

 

[briantrumpet]

 

[farfromtheland]

I'm safe. Two was overkill.

 

[The Crofted Crest]

St Paul's burning down 44 years before it was built.