This is what I first though of after I saw the drawing tool use demo. Right now we have specific AI that can generate images and drive cars, but tool use shows a process oriented AI that can use tools to drive a car, like we use our eyes/hands/feet. It's process, not task, that matters for AGI. And I think for artistic collaboration you just want something that knows your process deeply.
I also think the big data approach will never quite reach their goal. There will always be a wall infront of your goal that you will never be able to cross even with infinite data. An LLM isn't good because it generates a really kick ass token, it's good because it generates a lot of good tokens, and the system around it manages the context well enough so it doesn't get confused.
Don't be so quick to downplay process oriented AI due to latency
Why would that be front page news? You can literally buy tracking data for soldiers in the field, not from hacks from Google, study their movement patterns as much as you want. This has already been done and resulted in attacks.
It's just moving the goalposts, this happens every time an AI solves a problem, doesn't matter if the goalposts were there for 26 years.
What's interesting is that there are a set of people who are "in charge" and can as they wish arbitrarily set the goalposts to the thing that they happen to be best at. While this might be satisfying for an Humanity vs AI narrative, it's concerning for an us vs them one. Are these people really special? or do they just change the rules of the game so that outsiders (human or AI) can't win.
If it was you or I that solved this problem our would our rewards stop at $1M? Or would we get authority? If the achievement earns that for an insider but becomes 'just a solved problem' when an outsider does it, what exactly is being rewarded?
It took writers 2400+ years to solve Socrates' UX complaint, we can now talk to previously written text and it's pretty close to the original context. lets hope AI is a bit quicker in solving their own UX complaints.
And so it is with written words; you might think they spoke as if they had intelligence, but if you question them, wishing to know about their sayings, they always say only one and the same thing.
And now imagine that the only one and same thing the LLMs say is not only fixed, but intentionally steered as the biggest and most inflexible propaganda mechanism we have ever seen.
Especially given the lack of competition in the space, and people mostly using models from 2 or 3 providers.
It's quite literally a impossible for there to be a world without work.
You can't have fully realistic VR
You can't have AI/Robots that replace all work. A bulldozer makes a bad livestreamer. You'd need synths, indistinguishable from humans that both supply and can't receive(!) emotional labor, clearly an impossible task. Maybe most people will get all their needs from semi-realistic VR and AI-tiktok, but some won't.
Even now some people are hospitalized due to talking to AI too much. The thing currently needs a lot of labor from humans
>Work: Using physical or mental effort to do a task
This used to be
>Work means using physical effort to do a task
In the future it will be (not saying LLMs will bring us there)
>Work means using physical, mental or emotional effort to do a task
Women are well ahead on this since men effectively automate the physical and mental labor for them. There was a big advantage for men, then it sorta equalized but still in men's favor, but during the AI age it will truly be a woman's world. However you route it it's going to be a very long walk from "no more coding" to "we are all babbling toddler with synth mommies".
Isn't the problem just that open source, by it's very nature, is low quality since it comes from the unix-like ecosystem where everything is built atop everything else. So inevitably when the skid comes to you with a bug you can say "ah that's a bug in a specific build of your shell with these environment variables, myyyy code is clean as a whistle, minimal, and follows best practices" (if we ignore that in practice it depends on a billion different programs versions all of which are incorrect)
Complaining about the inevitable toxic noise when the same boneheaded development decisions are made over and over like everyone has gone insane. You don't have to build a death trap on the land your cat and children roam.
If the town feels that arresting children is a good use of public money they that's their right. At some point the citizens have to actually care about how their town operates if they want it to get better. Most townspeople care about UX as much as f2p mobile shovelware developers.
If they can "interpolate" existing data to that level then saying "just don't hack" doesn't make any sense, hacking is derived from knowledge of software systems. It's far harder to solve navier-stokes than creating a program that replicates and abuses computer resources.
You could as models improve continue to remove more and more training data, what happens when there is no more data left to remove but a running system still outperforms humans? I think you grossly overvalue data.
I also think the big data approach will never quite reach their goal. There will always be a wall infront of your goal that you will never be able to cross even with infinite data. An LLM isn't good because it generates a really kick ass token, it's good because it generates a lot of good tokens, and the system around it manages the context well enough so it doesn't get confused.
Don't be so quick to downplay process oriented AI due to latency
reply