My degree was joint honours AI and Philosophy, then MSc Cognitive Science, in which my thesis project was restricting procedural music generation (markov chains from Bach) by grammatical music rules, kinda analogous to the transformers at core of LLMs. Used and deployed all kinds of AI in games and theme park industry over past 20 years.

“We’ll just get make software that writes software!”, has been a clarion call for the past 50 years. There’s a reason any experienced software dev winces when someone uses the word “just”.
Good luck, have fun.

hunk wrote:

SpaceGazelle wrote:

The people at Davos aren't prepared. Capitalism in general isn't prepared. Society as a whole isn't prepared.

The internet made tech firms rich but this is something else. The sheer pace of it is nuts yet they're sitting on this and somebody somewhere will make a decision on what to deploy. Meta seem keen on making it open source which normally I'd agree with but it's also making me nervous. Too much power.

Big Tech (like any sector of Big Business) only thinks about short term profits. The long term effects of AI development..... In the end, AI might topple Big Business and society as we know it. For better of for worse.

The odd thing is it's more about the wow factor. Of course it wants the profits but the people working on it seem to be doing it out of astonishment. That will translate to profits but the danger seems to be human curiosity.

SpaceGazelle wrote:

The people at Davos aren't prepared. Capitalism in general isn't prepared. Society as a whole isn't prepared.

The internet made tech firms rich but this is something else. The sheer pace of it is nuts yet they're sitting on this and somebody somewhere will make a decision on what to deploy. Meta seem keen on making it open source which normally I'd agree with but it's also making me nervous. Too much power.

Big Tech (like any sector of Big Business) only thinks about short term profits. The long term effects of AI development..... In the end, AI might topple Big Business and society as we know it. For better of for worse.

The odd thing is it's more about the wow factor. Of course it wants the profits but the people working on it seem to be doing it out of astonishment. That will translate to profits but the danger seems to be human curiosity.

Potential fears? Can you describe them?

Also yeah, it's definitely a pandora's box thing. We can't put the genie back in the bottle.

Because it seems the sensible thing to do?

Waymo system is limited by referring to HD maps. Any deviation isn’t tolerated.

SpaceGazelle wrote:

Funkstain wrote:

But that amazing ability to almost instantly (within safe margin of time) assess countless variables effortlessly… show me self driving software anywhere near that.

But that's exactly what it does! And it does it without calculation using the trained model.

The only system I’m aware of doing this just like us (photons in, control out) is Tesla’s v12 self driving software. Video learning without human coding. They are feeding it millions of real world clips from their cars and the learning curve is exponential.

They all do the same, not just Tesla. It isn't even the best one.

They will be dropping a modified version into their Optimus bots later this year working alongside auto workers.



Dexterity is impressive given they’ve gone from a fella in a cloth suit to this in two years. Granted it’s not doing backflips on scaffolding but that’s probably unnecessary on an assembly line.

hunk wrote:

The people at Davos aren't prepared. Capitalism in general isn't prepared. Society as a whole isn't prepared.

The internet made tech firms rich but this is something else. The sheer pace of it is nuts yet they're sitting on this and somebody somewhere will make a decision on what to deploy. Meta seem keen on making it open source which normally I'd agree with but it's also making me nervous. Too much power.

Big Tech (like any sector of Big Business) only thinks about short term profits. The long term effects of AI development..... In the end, AI might topple Big Business and society as we know it. For better of for worse.

The odd thing is it's more about the wow factor. Of course it wants the profits but the people working on it seem to be doing it out of astonishment. That will translate to profits but the danger seems to be human curiosity.

Potential fears? Can you describe them? Also yeah, it's definitely a pandora's box thing. We can't put the genie back in the bottle.

The fear is losing control, and we will. We already are with critical systems like driving. I guess it's no different from humans losing control over things like the climate but this is willful loss of control to something we still consider cool.

I… I mean I understand how what I wrote was begging for the “but that’s exactly what it does” answer but there’s a level of understanding, or I’m not sure what the word is, but fine understanding of the data we get, the simple contextual awareness that that is a red T-shirt or those are dodgy road markings with such instant 100% correctness that easily make up for the speed with which machines can execute decisions, based on probabilities which are nowhere near complex enough to allow for real world conditions and variability

Will there be a data driven ai capable of making decisions across the variables that make it a better driver than humans eventually? Sure, I’m sure of this, but we’re not even close yet as far as I can tell - controlled demos are not exactly the same as real world driving conditions lol - and even then there’s not understanding just probability based output (and this is where it is explained to me that “understanding” is in fact just probability based output…)

Ok. And so we have cars better at driving than us which seems like a very focussed, quite complex but feasible product output. We know that ai can generate realistic images and videos of people, can help with better chatbots, can support mass, focussed, output - specific data crunching

What else are we looking at that I should be concerned about

Imagine a human folding a t shirt that slowly.
Sacked on the spot.

Hurry up you useless prick.

You should be concerned that we don't know the solution, the algo, but we deploy it anyway. The algo is just a matrix. It's both insanely complex and stupidly simple. it's just a bunch of numbers joined together and once it's trained it might fit easily in a few megabytes.

With driving we have some idea what it's doing but only a little bit. It'll brake to avoid collisions but it'll also see patterns all around that are complex but useful. Everything is insanely complex as the 3 body problem demonstrates.

As for what AI can do then it's probably easiest to try and think of what it can't do because the list is shorter. Empathy?

The patterns it can identify in nature is the real power in a complex universe. Once the designer chip bot is talking to the materials bot and the mining bot and the flying/driving bots it can all get out of hand pretty quickly. The automation is the scary part and industry and governments will use it. It'll work, and the better it gets the rate at which it improves is going to accelerate. It'll gather the data and train itself if we let it, and someone will.

You’d have to pay him for one of those.
Don’t worry though, I’m sure he’ll let you pay off your indentured servitude by earning company scrip toiling on his Mars base.

LivDiv wrote:

Imagine a human folding a t shirt that slowly.
Sacked on the spot.

Hurry up you useless prick.

‘By your command’

Okay. I understand the fear. I do. We use some form of self-improving AI to invent, say, room-temperature fusion power generators. And they work. And it changes our world for the better. But the snag is that we don’t know how the generators work. Only the AI knows. We literally look at it and go ‘nope, no idea, might as well accept that it’s magic’. And our entire civilisation comes to rely on this clean energy tech we don’t understand. Then the AI stops talking to us. Or we shut it down for whatever reason. And our fusion generators fall into disrepair and we’re fucked.

But … personally I don’t see that happening. I think climate change will have screwed us all long before such a story could come to pass.

Humanoid robots have always been a dumb idea, if you ask me. Humans are a fucking ridiculous design. Why on earth would you want robots to be the same shape? Make the robots better shapes for their given tasks.

The only reason to make a humanoid robot is to prove that you can. Because it’s really really difficult and obscenely expensive.