OpenAI loses another senior figure, disperses safety research team he led OpenAI has lost another senior staffer, and on his way out the door this one warned the company – and all other AI shops – are just not ready for artificial general intelligence. The departing exec is Miles Brundage who on Friday will cease working as senior advisor for AGI readiness. AGI – artificial general intelligence – is …
On one hand, corporations will care just as much about AGI safety as they did about LLM safety, which is not at all - or at least just doing little enough that they can pretend they care.
On the other hand, you're not getting AGI from LLM no matter how large you scale it, so we're safe from that till at least another fundamental breakthrough. Yay?
In my opinion that next fundamental breakthrough will be when they will successfully wire a computer to a human brain and that the computer will be capable of using the living neuronal network as a means of advancing it's own learning process. That's when the real scary stuff will begin.
We already know that Neuralink has already succeeded in establishing a basic wiring stage. From here on in they, Neurlink and the other research labs, will start to quickly advance this kind of technology as I imagine that the equivalent of a Moore's law will apply here too.
Objectively, we're going in that direction. We have machines to do the hardest parts of manual labor (aka mining, tunneling, farming, etc). These machines need human supervision, but humans do not need to do the hard work. We have robots for many aspects of manufacturing. Cars are made mostly by robots, with humans just checking things out.
As a species, we are working towards having robots do everything. We'll have a robot car, no need to drive. We'll have a robot butler, no need to clean the house. We'll have robot road makers, house builders, farmers, etc. We're going there.
When we get there, what will we do ? We'll sit back and watch all that stuff work for us. Meanwhile, we'll be watching cat videos on Youtube. That means we'll need the means to have a computer, a connection, and the snacks we eat in front of said screen. All of that will need to be provided for, because we won't be working anymore.
This will be a sea change in human society. The sci-fi novel The Expanse touched on that subject, with Earth population being able to decide between Basic, where they would be fed/cared for for nothing, and /Not Basic/ (don't quite remember), where you had a career (mainly in politics, apparently) and a salary that could allow you some benefits beyond just having food, clothing, shelter and health care.
In the long run, we're going to have to come to grips with a society where humans no longer need to work to provide for themselves.
But it's in the long run.
The idea that automation will remove the need to work has been around as long as technology. So has the notion of technology benefiting all of humanity.
I don't think there's any reason to suppose this particular technology will be different to any other in those respects.
Plus, it ignores the huge cost of the care and feeding of the beast. They're already talking about having to build nuclear power plants - not for the benefit of humanity but for the benefit of the (owners of the) machines. And it's an endless process: there will always be something more that can be learned and ingested and the training and retraining must inevitably continue. And it's the poor bloody humans that will be excavating new material in the content mines.
But above all, these fantastic visions of an idealistic future require us to believe that, perhaps with a little light-touch external guidance, Utopia will joyously be delivered by the likes of Microsoft and Google. That seems like the ultimate AI hallucination.
"In the long run, we're going to have to come to grips with a society where humans no longer need to work to provide for themselves."
The *extremely* long run. Capitalists do not want a world in which we no longer need to work to survive. Not only is that leaving money on the table, from their perspective, but if we're not working to exhaustion, then we have the time to sit and think about how shitty the status quo is, and how to go about knocking them off their perch. "The purpose of a system is what is does", ie: it funnels public money to the already ultra-rich and powerful, and it protects them from the consequences of their action. That is not a boat which they are going to want rocked.
I mean, so much of the debate around AI and its abilities is already hand-wavy bullshit to that end. "Look this way, at the existential terrors that a non-existant AGI will bring! Plz don't look behind the curtain at how we really just use LLMs to automate our biases at an industrial scale, push creatives out of work, and hoover up the entire internet."
The fact that OpenAI *still exist*, and haven't been sued for trillions in copyright infringements kinda bears that out. Remember when the RIAA sued the little people for trillions just for a few MP3s?
" Capitalists do not want a world in which we no longer need to work to survive"
Capitalism would no longer exist if people did not have to work, as the things that you need to survive would no longer have any intrinsic value that could be exploited by the banks/capitalists..
No. Stuff you need to survive has intrinsic value because it's stuff you need to survive.
Take away someone's food/water/land/living space and find out how much intrinsic value they attach to it.
You'll always need to swap your thing you made/work you did for someone else's thing they made/work you need them to do, or for basics like food and clean water or a place to live.
Resources will never be infinite.
Whether you do this by exchanging money or clubbing them over the head until they comply is up to you.
Well presumably it's more of a not giving it to them until they do something you want first. In the theoretical society where everything physical can be created by a machine, including another make-everything machine, the need for people to do something to get that is mostly eliminated*. However, the discussions of what we will do in that condition often skip over the part where no such machine exists. We've had this post-scarcity future in science fiction for a long time. How long has the idea of a robot butler existed? It's longer than we've had the word robot, which is why the idea still exists in a world where there aren't that many butlers. Yet, there are no robot butlers, no matter how wealthy you are. At best, there are machines that can do a small subset of butlerian tasks only when someone has already fulfilled the specific requirements of the machine. If I was sufficiently wealthy and wanted to, I could buy a machine that can make breakfast, as long as it's a specific breakfast, and something to deliver it to me, as long as I've built the routes the machine will take to do so, but the machine will need the ingredients of breakfast in a specific place rather than acting like a human would. If we can't make that, maybe the prospects of a future where there are no physical needs are a bit lower than science fiction would suggest.
* Even if we got make-everything machines, there would still be limits on some things. The one that jumps to mind is land. People want houses and they mostly don't want them in Antarctica. A machine will not make it possible to have as much space as you can use. Maybe the population will fall and we'll end up having enough space for everyone, but that's far from guaranteed. Unless we have an omnipotence machine, chances are that there are things that people want that can't be fulfilled automatically.
This post has been deleted by its author
Only because of the monetary systems currently in use. If "AI" and robotics becomes as widespread as some predict, the monetary system would collapse completely, because there flat out wouldn't be enough jobs for anywhere near the billions of humans that populate this little planet. You'd have to, at a minimum, set up a "guaranteed minimum income" system which effectively guarantees individuals what they need to exist and thrive, ala "Star Trek."
Sure there might be perks and bonuses you could "earn" by doing something that can't be automated, but your essentials would have to be taken care of gratis by the now-world government.
" Capitalists do not want a world in which we no longer need to work to survive"
Capitalism would no longer exist if people did not have to work, as the things that you need to survive would no longer have any intrinsic value that could be exploited by the banks/capitalists..
As Vetinari and Lady Margolotta separately observed stripping away the facades, it all really comes down to power and control. In Margolotta's case raw power was much more intoxicating than puting ones fangs into the jugular (and cuts out a messy intermediate step.)
I cannot see the assorted and various megalomiacs that have always afflicted our world, being satisfied with playing an elaborate game of Simon Says with their legions of AGI automata.
You your want real fear, pain, suffering and absolute obeissance there is no substitute for real flesh and blood.
"When we get there, what will we do ?"
For a lot of people, die of malnutrition / pollution / entirely preventable illnesses...
Automation is not designed to benefit the masses. It is there to make companies less expensive to run, with an arguably more compliant "workforce", in order to concentrate more money into the pockets of fewer people.
History hasn't gone that way. We made computers, which could automate lots of tasks. Not only did that require a lot of people to build and program the computers, but most of the jobs it eliminated were replaced by something different that wouldn't have existed without the computer available. Whenever something becomes cheap as chips so you can have a large quantity of it for relatively little, we've found ways to want more stuff. I don't see a reason why that would change. I also don't see a reason to expect that the automation of so many tasks will speed up. Yes, it will continue to happen, but I wouldn't bet on that happening quickly. Unless it happens over a short time period, society will probably find a way to absorb it the way it has all the other times.
"Objectively, we're going in that direction. We have machines to do the hardest parts of manual labor (aka mining, tunneling, farming, etc). These machines need human supervision, but humans do not need to do the hard work. We have robots for many aspects of manufacturing. Cars are made mostly by robots, with humans just checking things out."
Maybe someday. Currently, I think not as much as we're led to believe.
My memory says that Elon tried to build Model-3 Tesla's that way a few years ago -- with decidedly mixed results.
I could be wrong. I hope I am. But my impression is that much as I would like all boring, tedious, repetitive and/or outright dangerous jobs to be done by machines,it's going to take many, many decades for that to come about.
I think you're missing a piece of the puzzle.
Tax.
Today we as countries derive income from people - income tax and consumption tax and some capital gains.
If no-one is working, then that tax disappears.
There is no proven model through which we've been able to effectively tax multi-nationals and prevent them extracting profits to tax havens. As these corporations (or whatever name you'd like to call this new power structure), they will gain power and become even more difficult to govern.
As a result, the only outcome that is foreseeable is that these same companies that already extract all their profits will do so in future, so all that money that was flowing through a country's economy will be extracted.
The net result is that the only structure able to afford to pay UBI would be the AI companies. But they will have no interest in paying, hence why they extracted the cash in the first place.
This also ignores the societal problems of living without purpose. There's a very good reason that in the UK the average period for drawing down a pension is 8 years. It's not because we retire too late to enjoy it, it's that the process of retirement removes purpose removing the desire to live. Watching cat videos is cannot give anyone purpose. It is a time filler between having purpose and death.
>AI could enable sufficient economic growth that an early retirement at a high standard of living is easily achievable<
It almost looks like the venture capital money invested in AI already allowed Miles to do just that.
For some reason i seriously doubt AI will allow people working in other fields like building construction work or taking care of the pluming in your home to retire early.
It could to a great extent by de-skilling the job.
Instead of having to know the building regulations, be a certified electrician, etc, you can rely on the tool to tell you how to diagnose a problem, what to do to fix it and what final checks need to be done to validate the result.
There is a troubling and interesting challenge with this sort of technology, which started with the arrival of the Internet. With more and more info on-line, the qualifications and regulations to be in a trade have been updated more and more frequently. This helps to protect the industry to some extent from the DIY'er to looks at a video and does the work for themselves. However, this is a stalling exercise since once the regulations are converted into a tool that is able to process the regulation and 'see' the context of the need to apply it in a given circumstance, you're not able to do that any more.
Put more generally, there's a question that needs to be addressed in future about what it means to know a subject and to be a specialist. What will it mean to know 'law' or to be a 'lawyer' if you have a sufficiently 'useful' AI tool that can process the problem and provide a way to resolve it within the normal bounds of the application of that discipline.
I don't know the answer to that, but it feels like we're going to need to come up with something fairly quickly.
Not that old chestnut again! Somebody's been earning a living by spouting this nonsense at least once a decade since the 1950s, before they have it pointed out to them that manufacturing all the robots that can do every conceivable job now and in the future will need even more complicated robots, and the robots that make all those robot-making robots will have to be manufactured by even more complicated robots than those....
There is also a question of economics. Yes, it may be technically possible to have robots do everything for us, but will it be economically possible? That's why robot harvesting systems, for examples, have been so slow to develop: yes, machines can pick strawberries more efficiently than a bus load of modern slaves from Albania (poor sods) but that machine is going to be sitting put of use for eleven months a year while the poor sods from Albania can be used to do other things.
I read that string of statements and conclude the staff must hallucinate as much as their creations, given everything from the vague possibility of them creating AGI through to the utopian ideas of the consequences are pulled from nothing connected to reality.
Not unusual for the AI space but a particularly strong example of the nonsense that emerges - guess that's why they were on the big bucks.
"...early retirement at a high standard of living."
For the CEO's and upper management of the AI companies, yeah, but for everyone else, we're looking at abject poverty and destitution while the greedy trough-feeders at the top of the pyramid grab ever larger shares of the national budget.
This is I think what we're seeing in action:
1, Invent something that sounds incredible, trained on data you've stolen
2, Launch it and accept it's a bit rubbish, but expect that with more development and data you can make it better
3, Do enormous levels of marketing to make everyone feel like they're an idiot if they're not using it (actually training it whilst making themselves more stupid since they're becoming dependent)
4, Use the input from the users to train the system
5, Start replacing the lower level people who have been using it as you now have sufficient training to replace them
6, By showing that you're improved the system, you encourage more people to use it as the quality has improved
7, Start eating away at the next level of people
8, Repeat until you have a pretty decent control of most jobs
9, Put your prices up since the people doing those jobs are now long gone and the systems and processes are forgotten, so now you're irreplaceable
10, Keep going as the governments will be too terrified to stop you as you are now national critical infrastructure
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
