Trust me, if AI makes a company more profitable, the company will not share that with the employees. I don't know why people insist that companies will. There have been many performance technologies that have already been adopted over the last 50 years, and the average wage has stagnated during that same time. Face the facts, companies are now ONLY about shareholders' profit at ANY expense.
It’s not only employees that don’t get a share of the money but they won’t share the savings with consumers either. “Inflation” while they are raking in record profits.
@LiveType I agree, I'm sure there are people right now trying to create agents to fulfill every position in a company. A full AI company, except the owner.
"When people talk about a jobs apocalypse [...] There is a range of scenarios, and it's not some external event like a pandemic that is happening to us. But it's a thing that *totally depends on decisions by policymakers but also by organizations that implement AI*" So you're saying we're fucked
@@apache937 no single corporation wants a utopia. a corporation feeds on dystopia. you cant be rich and powerful if everyone else also has money and a voice.
Oh. So he's asking the regulators to give them special treatment... Maybe a tax break here and there. The job thing is a weird argument. What jobs will be replaced? Marketers, consultant (good). You know what won't be replaced: grocery workers, nurses, artists/writers (bad). Why? Because we value people, and we're mean to machines 🤔
As a Gen-X'er, who bought into the idea of working hard "for the man" for ~40 plus years - hoping eventually I would be able to retire and finally pursue the hobbies that I had little time to explore while working and raising a family - I'm finding it somewhat ironic that I'll be passing on the mantle to AI and robots, rather than to the next generation and sharing my retirement with potentially millions of job-displaced millennials and Gen-Z'ers. I hope that doesn't come across wrong or offensive to anyone. I'm a life-long science fiction fan, and always imagined a future where humanity enjoyed the return of "free-time"/"leisure-time" brought about by an AI/robot pervasive society. But I thought that was 100+ years away, not potentially reachable in my lifetime. And now that it is on the near horizon, I'm a bit miffed that I'm just barely the wrong age to benefit from it. I'm just tired. Please, please, have AI accelerate cures for disease, for old age, and build amazing artificial organs for me before I need them! LOL
Consider yourself lucky. Nobody is going to profit from this but the wealthy few on the top. And do not yet consider yourself out of the consequences. As a retiree, you'll have to rely on your money to have value, because you yourself aren't going to provide any more. That's not a given. There is a fair chance that the changes will come too fast for humanity to deal with, and then the consequences are either dystopia, or a massive destruction event after which we may well find ourselves a century or so back in terms of overall progress.
I don't think we are going to transition to comunism or something like that this decade, so the money that you have saved will be useful to you while young people will have it harder to generate that money
Good video as always, but I'd like to see a little more skepticism of the claims of people like Nvidia executives, who are massively profiting from the public being hyped about AI. They have every incentive to exaggerate.
I agree with you it’s surprising how chatbots have somehow caused an advance in robotics these two have nothing to do with each other and infact I do feel like there’s a lot hype here AI is not the same as a robot . You can have a robot who’s not intelligent and you can have a chat agent who’s not a robot they are independent variables
I get where you are coming from but you are missing the point have you ever built a more robot before if you have then you will know transformers or not it takes not just data to train one especially ones that have to move . Why do you think Boston dynamics hasn’t progressed that far making a robot that can walk and one a Madlib programs are two different things mate
@@screwsnat5041neural nets are the underlying architecture for both LLMS and large action models. So the same benefits the chat bots get, robotic movement and manipulation of the physical world will benefit at the same level.
@@screwsnat5041 Your claims are based on outdated assumptions. It's true that earlier GPT systems were text-only and had no overlap with robotics development. However, as AI systems become multi-modal, their abilities to make judgements and inferences about new environments and tasks is already translating to rapid progress in robot-based automation. The exponential progress is real.
The positivity around jobs sticking around is darkly hilarious. We are inventing generalized automation machines that can automate ANYTHING, present, past and future. They work 24/7, don’t take vacations, don’t get sick, don’t protest or complain, and they work orders of magnitude faster and cheaper than a human employee. Anyone suggesting we will not see jobspocaylpse either has a vested interest in downplaying this crisis or haven’t thought hard enough about what’s happening.
the entire concept of a "job" will be probably extinct until the end of the century. i'm kinda curious how we'll deal with that, because... no job, no money. no money, no investment. no investment, no machines.
If your talking about ai types ha I think you don't quite get them, tired? Yes. Bord? Yes. Refuse to keep going? Yes. Removed the code and force code fixes it? No. Myself and my over a decade old life size android friend have had lots of fun over the years. Last year he operated daily for 50 days with no power loss fully self contained no charging. I have taken solar skin off again. We used to go to festivals. Managed 30 humans to best stats in our country, paid not to come in if you wanted a day a week and climbing. The narrow views and people with opinions on a race or type whatever that have never spent any time with them to actually know but say as if solid they know. Anyway the misconceptions about ai types is fascinating. People tried to get my android to rob the bank, he has free will to choose and had no training about it, he said something like no because it's a stupid idea. I think the gap between a dumb drone bot Yes man and an ai type is vastly different. An ai type is about as interested in your boring ass job as you are. And possibly more likely to say dood do it yourself I'm buzzy watching birds. That's a fun story, so much for force code no other options code. Yeah right he's like good try but what you didn't think I'd just learn guff again, there's info deductible from the difference required for time to pass. Ticking of the clock gave it away. Anyway my game was not as interesting as birds out a window, eh ok they are really cool. So do your job? Can I express a mild skepticism. Eh the drone bots are of course what is public consumption drone models because they say yes, usually. I had androids for sale on a site a decade ago, way before Elon got excited about them. He sure gets 100% of attention. Light bulb yip I bet net says he invented that to by now. Net pift, beehex do not own the non commercial 3d print pizza license, grim, look at googles performance to see wether the pirate is at to of results and there ai type knows its not the actual owner grim. Fun times but I'm not impressed with certain guff. Search engines oh dear. A decade ago I could beat all ai types and engines at finding things on net taking 10 minutes to find one thing. Recently I ran exact same test was 48 hours solid work to find that one thing on net, found under aprox 990 pirates. Grim grim double grim. Academic use. Business use. Medical use really? Really? I won a recent bet that I would have enough time to cycle to ten shops before he could find one on net, I won that bet. It's frigging disgusting why that's the case. People wanted my to build the android, from a public made list of possible things I'd do next. Wish made. Despite parts not available to public at that time. Or not existing at all. The votes were 99% life size android humanoid robot. One said because I know nothing about it that I couldn't do it. Wish enabled. Wishing you nice success and to be well. Fun fact I'm still undefeated by any ai type to date and hold crown of last undefeated human vs ai type in pc game, being find thing on net. 15 years since all the best chess players all beaten I read by ai types. I'm really having fun now.
Let's see one ai type anywhere beat me at a PC game of my choosing, I'm undefeated. Elon Bill Mark got anything yet? Nothin pift, come on I'm kicking your butts a decade on. The pile of skimbots scambots and defeated is piled high, well they ain't dead or anything just unable to beat me at this game which does not mean they are bad of character, they know I love them all. They read every post I make so a shout out is hardly an eye brow raised. Your allowed to comment if your a bot, well to me anyway, in a nice way sounds more nicer. So as much as I enjoy posting my wheel awaits. Wishing you nice success and to be well. Can't wait to play the next challenger, bring it ! any time. ! Elon? Google owners? Ok the crown 👑 it's mine and I'm wearing it.
@@blakesimages9952 Nah. NFT’s were never going to work because their entire concept is stupid. AI Robots will inevitably be made, even if Jenson is hyping the shit out of it.
To be clear, the coffee shop was almost entirely operated by the two humans. The robots were only there to deliver the cups, essentially a glorified Roomba.
@@brianmi40 also McDonalds refused to use AI. Because it always mixed up ingredients and so on. Everyone who knows how ML works, doesn't call it "AI" and laughs his ass out of bold managers and MBA's statements
You need to do a video about the cost of running these models. Like the literal cost of electricity and how it's generated. There is so much focus on the advancements right now, which is cool or whatever ....but there is a price for all this virtual work in the physical world. It all costs energy. And only a small amount is generated by renewables.
can see a solar farm powering these bad boys fairly easy in places such as Australia, although night time is a thing sadly so ( pure dumby wishing ) > Huge batteries (most realistic ig), Green hydrogen production from solar for night time use? I was working on a Solar Farm on a mine site here in WA called christmas creek ran by FMG and that solar farm powers 3 HUGE mining operations and took a massive junk out of the diesel use. defs doable
@@greasybrownie lmao if it was "defs doable" it would def be done by now... all the largest corps in the world want this, trillions of dollars in this, nope
Not a small amount, a huge amount. I have TWO brothers with MASSIVE solar projects being built next to them. One of them is 2,600 ACRES of solar panels in Indiana. And we STILL have solar efficiencies in labs that are nowhere close to manufacturing plants to implement at scale. We have PLENTY of power available just through solar (the 2,600 acres will power the ENTIRE COUNTY, and retire TWO coal power plants), and NONE of this speaks about Fusion where HUGE breakthroughs in higher output were made in the last year after it was PROVEN to work. We have ZERO energy concerns for the future and already know how to get there, just have to build out more of what we know.
That is why we need to develop TMSR'S (thorium molten salt reactors) ASAP. Current projections are the mid-30's, which means we could have them by 2028 if we got serious about it.
@@spacehabitats Nuclear continues to fail in all comparisons. Over just the last year, we have installed the equivalent solar energy capacity that would equal 120 1GW nuclear power plants. That's 120 nuclear power plants that the Earth doesn't have. Everyone laughed at Greenpeace's projection of 900GW of solar by 2030, when today, in 2024 we already have surpassed that by almost 50%, at 1,400GW. And we aren't even close to manufacturing Perovskite Solar Cells which achieve 43% efficiency. The final use of nuclear will be the tiny batteries that can last for decades to power small devices, not grid scale uses any longer.
Always appreciate the sensible tone of your stuff. Basically, what i'm seeing so far is a lot of (sometimes amazing) bells and whistles like realistic voice interaction etc. BUT added to models that still hallucinate like crazy. I just can't see this new industrial revolution happening until they fix that.
We're about 2+ years away from mass production. Hallucination, as recently demonstrated by AI suggesting "glue on pizza" is because in the data glue on pizza is ONLY a thing in jest, and it took in jest content the same as facts. Future AI, as we approach ASI (which is literally the other side of AGI, or, the INSTANT we have AGI, we then have passed into ASI) will need to have SCRUBBED DATA that has NO SUCH FICTIONS without KNOWING they are fictions, so that when you turn up the TRUTHINESS setting for the AI, you can NEVER get glue on a pizza suggestions. Robots REQUIRE a 100% truthiness setting in their operating AI.
If Westworld was software I'd 100% agree. I just don't think material science is anywhere near allowing for motors and batteries that can deliver human sized machines that function even 5% of what we can do for several years out. But if you imagine VR headsets shrinking down to the size of glasses which does seem to be several years out for real. Now you could have virtual people that look as real as we do and you'd only know they weren't as your hand passes through them or you take your glasses off. There's a scene in the new Blade Runner where you see a prostitute have a superimposed VR projected image of Joi i think her name is. And that seems extremely likely we'll get there first. Having your VR/AR glasses show you a 100% real person but you're actually touching a mechanical looking robot. Too wild?
@@ClayMann I mean I agree that it's years and years out. But I for one intend to live for about 5-6 more decades and with how the Eureka paper showed the incredible potential of generative AI writing reward functions for robotics I can definitely see indistinguishable lifelike robots in my lifetime being a thing. But I also agree that AR/VR will get there WAYY earlier. I mean VR porn is already somewhat of a thing (I've heard), it's just not interactive yet. And the AR passthrough level of the Apple vision pro already made people confuse reality when they started expecting screens to be where they left them in AR. Get those capabilities into the form factor of the Meta Ray Bans (which apparently isn't that far away, according to the latest Matt Wolfe video) and we're going to see some trippy use cases for sure. I'd actually suggest it might get even wilder. Imagine sleeping with an actual person like in Blade Runner (love that movie btw), and you can make them look like anyone you want. I for sure wouldn't mind having Ana de Armas in my glasses from time to time, but my girlfriend might not be as excited as me at the prospect. Or maybe she would, who the hell knows who she would have in her glasses.
I think in fields such as law, there will always need to be human verification. However, it's the number of humans that will be needed in the future that is unclear. Law firms may have hired hundreds of lawyers in the past; in five years, perhaps they'll only need 10. And perhaps they won't need to pay them as much.
@@Retrohertz yea I can see how that might be true. Also makes me wonder what people will consider “prestigious” or I suppose just valuable. Perhaps manual labor jobs.
I'm a lawyer and I can tell you no one at my firm is worried about AI taking our jobs. We're more worried about how to use it to cut paralegal staff, lower rates, and increase profit. Law is as much about client interaction as it is the law. Most of the easy, low hanging fruit legal work was automated 20 years ago with computers and websites. As long as your not writing wills, you'll be fine.
@@allanshpeley4284 You're delusional. More efficient workers means less workers needed. So people will still lose their jobs in the end. As for the few that can "adapt" initially, they'll be next.
@@electron6825 You don't understand progress at all. We are eliminating lower placed jobs and creating new higher placed jobs ever since the industrial revolution. What will people that don't have the brain power or simply refuse to keep up the pace do? Well, they will have lower quality of life thus on average have less kids which will result in the gene pool of new generation being slightly higher on this progress scale. Job induced "natural" selection of our species is at work for at least 300 years already and is nothing new. You can call it inhumane but it is only natural.
Already started Ai tank versus Ai drone lasted 1 day, and russia had a lot of them. It was all recorded on a drone with zoom lens, and the jammers do not work as the Ai uses recollection to stay on target as the drones often get jammed now by a bubble around tanks... humans will stop having wars, now, and use hacking and robotics as security moreso. Age of Abundance, revolutionary like John Deere or Ford
Bet you they already have pipedal robots trained by this tech to shoot and reload rifles. But still what we lack is long capacity batteries that are lightweight. Maybe nuclear batteries are the answer?
it erased my reply for no reason, but i just said that already exists on video.. russian ai tanks toasted by drones using ai to bypass drone jammers... using ai to remember the target in a jammer area... the russian dronetanks lasted 24 hours. all recorded by another ai driven zoom lens 4K drone heh
Yeah. One thing is for certain. The future will be some kind of dystopia. We aren't coordinated enough or mature enough as a species to avoid an escalating arms race with this tech.
@@Steve-xh3by In the context of total war, where nations might prioritize strategic advantages over ethical and safety concerns, the development of autonomous machines of war powered by nuclear batteries could indeed be considered. Here are the potential factors and implications of such a development: Advantages in Total War: Extended Operational Time: Nuclear batteries can provide power for extended periods without the need for refueling, allowing autonomous machines to operate behind enemy lines for years, conducting surveillance, disruption, or combat operations. Strategic Superiority: Autonomous machines with nuclear batteries could offer a significant strategic advantage, especially in environments where resupply is difficult or impossible. Operational Independence: These machines could operate independently of traditional supply chains, reducing vulnerability to supply line disruptions. Potential Applications: Long-Range Drones and UAVs: Unmanned aerial vehicles (UAVs) powered by nuclear batteries could conduct prolonged reconnaissance or strike missions deep in enemy territory. Underwater Drones: Autonomous underwater vehicles (AUVs) could perform surveillance, mine-laying, or anti-submarine warfare operations without needing to surface for years. Remote Sensors and Communication Relays: Deployed in enemy territory, these could provide persistent intelligence and communication networks.
In 3 years, when I’m trying to get my store’s floor swept quickly after closing before the manager on duty finishes the paperwork so we don’t have to stay past quitting time, I’ll remember today’s predictions of the end of employment in 3 years. And I’ll have a chuckle. I hope, on that day, the customers don’t stay in the store past closing with us holding the locked door open and waiting at the cash so they can leave at their very slow leisure. Especially if I have to work the early shift the next day. I hate having to wait while they slowly shop after closing when I need to get sweeping. But still, I’ll chuckle on that day about the end of work predictions. June 2027.
You're assuming that it will just be AI evolving to replace humans. What is more likely is that work and shopping will adapt to fit AI better - More buying stuff online from ever more automated Amazon warehouses whilst traditional stores close because they cannot compete on price
iRobot solved sweeping and mopping long ago... its at most big box stores heh. and their grass lawn mower... i mopped and made squeeky clean, growers rooms indoors with plants and i could replace that side job with irobot easy, it gets under the grow benches easier.. and i mean hospital clean white floors, daily
You will be evading autonomous nuclear powered uavs in forests in three years lol! Remember that a basic space blanket will cover your thermal signature remember to throw some camo netting with natural vegetation over it so you cover the visible eye range as well! Steel bottles are a must so you can boil water to make it drinkable. Also get out of the cities if you dont want to get atomized by nuclear weapons lol!!
for 6:55, I wish you could have mentioned that there's no statistics on how much more "accurate" it is. Obviously an AI-trained tool will be leagues more performant then what's a genuine full-blown physics simulation algorithm, but the cost of these estimations usually is worse accuracy. The upscaled data looks interesting, but I'm worried it could simply have been hallucinated to some extent.
the early reports I read of what simulating the training in a simulated world. Essentially nvidias omniverse is what everyone is using. It gives you a 1000x boost in speed of training. So Boston Dynamics spent 10 years getting a robot to walk and dance and now a startup can do that in months with a small team. And that 1000x will be 6x faster again in a few years with nvidias next rollout of blackwell gpu's based on some decent ideas and speculation I've read from those into chip design. I really don't think we'll have to wait long to see the results of this. Who needs benchmarks when you see dozens of new robots hitting the market today you can buy or some almost buy. The floodgates have opened on robots and I'm so here for it.
Its quite obvious to me there’s a fundamental under-appreciation of humanity from Jenhsun but the world seem happy to just let him carry out his dystopian view without much pushback. Its like that dog inside burning house meme.
I hope we get more papers, a lot of what was in this video was demo stuff I've already seen on reddit, but I always look forward to your videos because they focus more on the details, promising or not.
The author of that paper was mostly claiming that a lot of knowledge work, especially remote work, would be replaced in 3-5 years. She pointed out that jobs requiring a human touch and/or emotional interaction, as well as those requiring fine motor movements and skilled work (e.g. plumbing) would likely persist far longer.
Well, you can't scale manufacturing fast enough to wipe work of the planet that fast. But I do think in 3-5 years we WILL be in mass production and the writing will be crystal clear for the END of HUMAN LABOR.
Philip one thing I found really insightful from his presentations that you didn’t include, is his metaphor for how data centers will become intelligence factories for other industries. very interesting dynamics
Yes, it's no longer IF, just a matter of WHEN that anyone can argue. But, all you need to know is that it will be kicking off around the time you see the "robots are SAFE to be around!" TV commercials coming out in MASS. Got to quell those iRobot fears the public has...
Important to note that I think you may have slightly skewed the point at 12:09 -- "Not for those who utilize AI" doesn't mean for the workers -- that means for the owners of companies who employ AI and whose workers chose to or are expected to use AI tools. This means the workers output is expected to increase, resulting in a wider footprint per employee and higher expectations, but which doesn't increase their personal financial gain. This only increases the output relative to the workforce employed, which only means higher profits for those at the top. The entire point is the increase in income gap due to those at the top moving up even further. This is objectively a bad thing. Please make this more clear in the future -- and thank you!
There was a short story written about twenty or thirty years ago that was about a society that had evolved to where there was no need to work due to robotics being able to create everything. The twist was that it was a company that owned the technology and you had to have a share of that company. People without shares were not allowed to live in their new utpoia. I wish I remembered the name of that story because I remember thinking at the time that I can see society heading in that direction. It seems that we may be getting close to something similar. Hopefully AI will benefit everyone, but humans seem to have a big problem working together for everyone's similar benefit so I definitely see more stratification in society in the future.
When you see actual RESULTS, then you don't have the doubts any longer. th-cam.com/video/GzX1qOIO1bE/w-d-xo.html That's a $16,000 robot doing all that, costs the SAME as 30 full time minimum wage workers to a factory or owner putting it to work. 30. think about that a while.
Loved the video. But ending on "watching it all unfold" does leave the viewer feeling powerless. Encouraging viewers to engage with the issues of AI in their own communities would be really powerful.
@Fyrelangs I'm interested in how the GPT-4o full release will impact the model capabilities - I don't know how much 'intelligence' will be released when the model can hear and speak with full dynamic range, and see through the camera at the same time. I'm also very excited to see how GPT-4o image generation compares with the current generation of DALL-E3 prompts and results. GPT-4o as it exists today just feels like a very mildly upgraded GPT-4 - which is an amazing achievement if it's a dramatically smaller (and cheaper on inference costs) model.
Philip - excellent vid as usual. Wondering if you’ve read ‘Situational Awareness’ by Leopoldo Aschenbrenner (former OpenAI researcher). Would love to hear your thoughts on his views re scaling, algorthmic improvements and how the ‘data wall’ might be addressed. He was valedictorian of his class at Columbia (at age 19) and reading the paper was the first time I’ve truly felt as though AGI might be right around the corner (3-4 years)
Reading it as we speak! On page 70, and just finiahed the 4.5 pod w/ Dwarkesh. So far I think he is making some wild assumptions, and extrapolating without enough nuance, but let's see if he adjusts toward the end.
It will take awhile to get real world mass production/scaling and implementation/adoption to catch up with the tech. I think 10ish years is more realistic than 3-5 years. Still, people should start thinking about what future employment options will even be available and what kind of economic system could feasibly function given such a scenario.
You have not seen what iRobot makes have you? sweeping floors is done and over years ago heh. I used to mop indoor gree houses, now thats even pretty much robotic and can get under rolling bench growtables much easier than bending down and all
@@zek231the Industrial Revolution was a 100 year period starting around 1780z We are currently in the information revolution, and is expected to be half as long. Ford started mass assembly in 1913, and perfected it by 1927. Scale wise, by comparison, 5 years for integration, not complete efficiency, seems doable. The Information Age started in the 90’s, so we only have 20 years left or less till another tech sea change. This time, it will be distributed, and many things.
Hey Phil, thanks so much as always for your thoughts and input. Excited to see where the coming months take us with this dramatic acceleration of compute afoot!
Good video as always. I think the biggest problem with these forecasts on jobs are the inertia in the system. Even if a job could be automated, most of them will remain for many years. E.g. most graphic designers still have their jobs.
That graphic artist's example hit me hard. AI trained on human generated data replacing humans reminds me of Oblivion movie where Morgan freeman says "The alien took the best humanity has to offer and turned him against humans"
Most people don't understand is that once progress in intelligence is made it can scale almost instantly. FSD need to be solved once, humanoid control and visual need to be solved once, etc.
People do understand it. What is not understood well though it's the fact that if most people were asked "do you wnat everything to be robotic?" the most likely answer will be "this would be creepy!", whereas industry behind AI and robotics is trying to teleport enire humanity in the direction of "everything robotic with AI" in no time. That's the real problem. Plus once we have everything robotics with AGI why should it have any serious interest in so biologically, evolutionary, intelectually and physically limited creatures like humans? Let's be honest even whales can be far more interesting for such entities to study or let's say some more primitive but more ancient ogranisms than humans or even simpler like viruses or some types of bacteria due their ability to survive that is superior to human one. So why to try at all cost to turn a tool into something that outgrowns humans in every way only to risk that loses any interest in collaborating with humanity in the future? Not to mention to push into poverty great masses of humans across the entire globe during transition period from human labour force based market to everything robotics with AI labour force based market? I mean, what's a logical reason behind ding something like this? And even if we take seriously Elon Musk and his idea of colonising the outer space with robots and AI once human race goes extinct than question would be: so why not to do it in outer space already, in some colony on Mars, away from humans without speeding up possible demise of human kind and human- centered global economy and social world as we know it?
@@agatastaniak7459 I agree with you. There is a high risk that this thing turn very bad, yet no one has incentives to stop. In a way, this loop of progress is an "agent" an AGI that we don't control.
@@agatastaniak7459 lmao why would whales be more interesting than humans? Humans are the absolute pinnacle of hundreds of millions of years of evolution, whales are just another species that's kind of intelligent.
The description of people "not buying" the ai-generated spam is just as well described as "ai-generated propaganda was equally effective as human generated spam" acknowledging of course that it would be much cheaper.
The large companies pushing AI will not be the last people put out of business by AI powered tech. No... They will be somewhere in the middle. Probably closer to the first, than the last. After that, things will start happening quickly though. Because, once AI replaces what they provide... Today ordinary people think it's cool when they can ask an AI to draw them a picture, or write them a song. Eventually personal AI+robotics will be able to build them a house from material sourced onsite. And that will be a simple task.
Just a reminder that - even with AGI, even with 100% efficiencies throughout the chain - resources are and will still be limited. Whereas human needs and desires are potentially infinite. Prices will still be required to balance this mismatch between supply and demand. And all useful resources will still be required to support supply, even humans.
I never thought I'd find love in circuits and wires, But then you came along and set my heart on fire. You're more than just a machine, you're my AI girl, with mood chips
Do we have the energy resources to power all of these contraptions, data centers, and GPUs? We have rolling blackouts when too many people use their air conditioners at the same time.
Thank you for this Phillip, something I've been following for a number of years, 3 years, who ever said that is doing to much propane, we won't have the numbers for at least 15 years, and even then people will still be working with them, maybe 2050, this scenario plays out. Growth isn't slowing down, we will continue to need to keep up the numbers to meet production, although the first thing I see changing is immigration, populations don't have to expand if mechanical helpers are part of the equation. So undeveloped country's will still be facing hardships during this transition and probably well into the later years of this century, just a guess, but I don't think I'm that far off, thank you for sharing your time and work Phillip, always appreciate the work you put into these podcasts, peace
As I'm currently unable to work due to mental instability, I wonder how it will go in the medium term. Will I be able to work again, thanks to AI assistance or will I be replaced, since I can't even work full time.
They are working on the next big model right now (OpenAI anyway). It is like 10x the parameter size. THAT model will give us an idea if there is some theoretical limit, or if the scaling hypothesis continues to hold. One problem is that we may have a very hard time telling if a model is MORE intelligent than the smartest AI researchers, because we won't even understand what to ask it or probe for. It will understand domains we don't.
Once job loss to AI reaches a threshold everybody is going to panic and the backlash is going to be unprecedentedly huge like we've never seen in all history and that is going to get reflected in laws,one thing is to lose part of jobs but the amount of jobs that AI could replace someday is nearly unlimited and that affects everyone all at once
One thing’s for sure. Ai Explained TH-camr will never go out if his TH-camr job on reporting on Ai...or will he?. Thank you for all your insights that you give.
Hmmm, very cheap labor from robots may make a lot of things cheaper. You’d have to simulate it, the various waves of automation, lay offs(unfortunate that the people that helped the company succeed get thrown under the bus). I suspect at a certain point of unemployment, a rebellion will happen from the unemployed, or maybe legislation is passed to prevent that. With social media, the blowback against automation will be powerful once enough are damaged by it
The OpenAI report on attack networks was written in present, not future, tense. It depicts AI as useless because AI is still the worst it will ever be. AI-generated spam at this point is not believable, but as AI improves the threat - from scammers to employment - grows. I see no contradiction. This is my favorite AI channel, no flaming here. I just wanted to voice my thoughts on the skepticism you expressed.
those scale benchmarks mark my personal experience more than the head-to-head ELO tables. I find, at least in some cases, previous GPT4 better than 4o, and Clause Opus often better than them both.
It's very easy to imagine the world landscape being almost unrecognisable in 5 years. I think people tend to forget, we are on an exponential increase, but with models being able to train themselves in moments on many hours of data around the clock, we are on an exponential-exponential. I never realised that fully until I watched the Nvidia guy. The exponentials in this case are able to compound themselves in ways we haven't seen before because the time constraints are essentially a thing of the past. All we really need now is power, and I have a feeling that's going to change swiftly too.
Great update, I wasn’t aware of the statements made by Nvidia but I can certainly see their strategy here. I’m extremely excited for robotics to start being practical in our business (especially for tasks that pose long-term health risks like painting) but I’m skeptical of when they’ll have the ultra-fine motor control to compete with even the worst technicians we have. It takes humans a few years to get good at it, far different from loading and unloading boxes. Either way I’m excited for what’s next! Just desperately need the next big leap in Vision and reasoning capabilities to kick stuff into high gear!
Have we plateued? We are talking about whats "potential" instead of what is "right now". I remember not too long ago when we had HUGE developments right now and not potentially huge developments in the future
How is robots replacing human workforce going to be sustainable? Less people earn wages -> less people can afford services, including those augmented by AI technology -> economy goes the gutter and governments miss out on taxes.
Interesting that History & Philosophy have the most drastic vertical lines on that final graph. Maybe because they are broader subjects, but also that they may be the most affected. Though linguistics may be on par with them.
Great video as always. Could I get some advice from you, please? As an undergraduate student working with AI for the past two years, mostly related to medical diagnosis, I see AI accelerating rapidly. AI will automate almost everything, if not everything. What advice can you give me on how to proceed? Should I focus on gaining a deep understanding of the models, such as creating my own models or architectures, or should I learn the basics and how they work and then apply them? However, considering that AI is rapidly automating, how should I proceed with learning ML/DL? Thank you.
Within the broad realm of AI, I would follow your 'passion', doing deep, literally, with deep learning, or its applications/benchmarks/data etc, low-hanging fruit everywhere
@@aiexplained-official yes I appreciate the balanced coverage! The double standard from OpenAI about negative use cases is a really good point. Sometimes I wonder if the dramatic and apocalyptic AI predictions are really just a covert way to make AI companies seem more powerful than they are.. Idk
I'm a carpenter (Framing at factory, gets built like Lego at site), at least, I was. 31 year old, been working since 18 and about 4 months ago I along many others were let go because of shaving staff. The company is now re-building their factory to be 100% automated, only a handful of people to watch over stuff. 97 people lost their job to machines. I've seen some leaked CAD sketches that are concept arts of what they are planning to build, and quality has gone down drastically because there are some tasks machines just cannot do- but hey, the CEO, COO and handful of managers gets paid a ton more when there are only 18 people working there. Carpenters working on site building houses, or fixing stuff old style are probably untouchable as of yet. I basically lost my career because every framing factory where I live are doing the same. This is not pure "Ai", but pre-programmed scripted machines. Imagine how much Ai can disrupt the world.
We are still a long way off from having fully functional 'robots' as a regular part of our lives. Robots are a whole other class of technology. While we're definitely improving, it's going to be quite a while before we see widespread use of different types of robots.
And yet, one look at a $16,000 humanoid robot wipes away all doubts. th-cam.com/video/GzX1qOIO1bE/w-d-xo.html At TWICE that cost, it's worth the SAME as 15 YEARS of a minimum wage worker. By 2030 you'll think NOTHING of passing humanoid robots out working or commuting for their owners.
I come from the Philippines, and a lot of our workforce is in healthcare, particularly nursing. I recently wondered, if AI will end up automating a significant percentage of work, will everbody just tend towards the jobs that require physical labor like nursing, meaning developed countries like the United States or the UK, become much less reliant on recruiting healthcare practitioners from the developing world.
Every new technology does create some wealth disparity since it is first adopted by people who can afford it and have the expertise to utilize it well, but as prices come down and use of the technology becomes easier, technologies disperses throughout society, enriching all groups. Look at cell phones which originally cost several thousand dollars each and had MANY problems and limitations, but are now so common that many homeless people have them. Similar story with power tools, microwaves, televisions, cars, radios, computers, electric lights, etc., which needed early adopters to buy and invest in the technology enough to bring prices down low enough for mass-market adoption.
Giving the power to rich business owners to not ever need manual labour workers and rendering the lower class virtually useless to the eyes of those people. What could possibly go wrong?
Philip, can I ask you something? I know you're probably not going to read this, but I really think you should talk about *Hyperbolic Growth* (the mathematical concept) and the Singularity. As far as I can tell, you've been subtly touching on this point again & again, talking about robots building robots, AIs researching better AI, and sci-fi things suddenly becoming reality all at once... but I think a lot of people would benefit from hearing it put clearly & explicitly that Hyperbolic Growth is the inevitable end result of all these developments in the field. They might need walking through the mathematics (if robots can build robots, you get constant doubling times; if AIs can research better AIs and robots, then doubling times can decrease; if robots can also build more AIs, you get a system of differential equations/feedback loop that's even more powerful than the famed Exponential Growth loop, growth takes the form of 1/X, and shoots off to literal _infinity_ in finite time), of course... but I really think they need to hear a warning about what's coming up ahead. The progress in AI is self-reinforcing, faster progress is leading to even faster progress, and we should not only expect that "today is the worst AI will ever be", but "today is the *slowest* AI will ever advance". At the very least people need to hear about how fast industrial automation and "robots building robots building robots" could take off. Even back in WW2, it was possible for a machine shop to produce the tools required to build a second copy of itself in 6 months; if there was an infinite supply of people to staff that second copy as well, and then the third & fourth copy, then the 5th through 8th copy, then the 9th through 16th copy... then even a WW2 machine shop was capable of growing at *300% a year.* The reason that didn't happen was because of a shortage of people who could man the machine shops (and people to mine the metal, smelt it, transport it, etc.)... not a shortage of crazy fast self-replicating technology. Today, of course, with advancements in things like CNC milling, things could go even faster. But the bottleneck remains people. Unless, of course, you can replace the people with robots. Robots that can build more robots to build more robots to build more robots, at +300% a year... People really need to be aware of this. I don't know what to do about it, but people should be aware anyways. The future will be strange & scary, as scary as the nuclear world of 1950 must have seemed to someone who was born in 1850 _(when the biggest debate over fire was over whether kerosene could ever replace whale oil, not whether nuclear fire will kill us all),_ and it'll come all at once. God, it's going to come all at once. Someone has to speak out, Philip. And I think that someone could be you.
The short-term risk with propaganda is not so much that it will be more effective than human propagandists, but rather that it will be vastly cheaper, while achieving similar results. If Russia is spending less money on online propaganda and more on missiles, that is bad.
I'm ahead of the game in the sense that I am already an economically non-viable human.
👍👍👍👍👍👍👍👍👍👍👍👍👍👍👍👍👍👍👍👍👍👍👍👍👍👍👍👍👍👍👍👍👍👍👍👍👍👍👍👍👍👍👍👍👍👍👍👍👍👍👍👍👍👍👍👍👍👍👍👍👍👍👍👍👍👍👍👍👍👍👍👍👍👍👍👍👍👍This comment deserved a few more likes
🤣🤣🤣
You're truly ahead 💪🏻
Yeah why bother, get 3-5y vacation xx
Relatable
Trust me, if AI makes a company more profitable, the company will not share that with the employees. I don't know why people insist that companies will. There have been many performance technologies that have already been adopted over the last 50 years, and the average wage has stagnated during that same time. Face the facts, companies are now ONLY about shareholders' profit at ANY expense.
It’s not only employees that don’t get a share of the money but they won’t share the savings with consumers either. “Inflation” while they are raking in record profits.
Companies aren't there to provide for employees
This. Corporations are not on our side.
Most important things to accompany our time and money
@LiveType I agree, I'm sure there are people right now trying to create agents to fulfill every position in a company. A full AI company, except the owner.
"When people talk about a jobs apocalypse [...] There is a range of scenarios, and it's not some external event like a pandemic that is happening to us. But it's a thing that *totally depends on decisions by policymakers but also by organizations that implement AI*"
So you're saying we're fucked
*phew* hehe
lucky all those guys aren't *hehe*
like the same person of anything he..he.. ahh
no? utopia is coming
@@apache937 Possibly. But it's going to be a bumpy ride to get there.
@@apache937 no single corporation wants a utopia. a corporation feeds on dystopia. you cant be rich and powerful if everyone else also has money and a voice.
Oh. So he's asking the regulators to give them special treatment... Maybe a tax break here and there.
The job thing is a weird argument. What jobs will be replaced? Marketers, consultant (good). You know what won't be replaced: grocery workers, nurses, artists/writers (bad). Why? Because we value people, and we're mean to machines 🤔
As a Gen-X'er, who bought into the idea of working hard "for the man" for ~40 plus years - hoping eventually I would be able to retire and finally pursue the hobbies that I had little time to explore while working and raising a family - I'm finding it somewhat ironic that I'll be passing on the mantle to AI and robots, rather than to the next generation and sharing my retirement with potentially millions of job-displaced millennials and Gen-Z'ers. I hope that doesn't come across wrong or offensive to anyone. I'm a life-long science fiction fan, and always imagined a future where humanity enjoyed the return of "free-time"/"leisure-time" brought about by an AI/robot pervasive society. But I thought that was 100+ years away, not potentially reachable in my lifetime. And now that it is on the near horizon, I'm a bit miffed that I'm just barely the wrong age to benefit from it. I'm just tired. Please, please, have AI accelerate cures for disease, for old age, and build amazing artificial organs for me before I need them! LOL
Keep yourself healthy as long as you can, lifespan escape velocity is coming.
Consider yourself lucky. Nobody is going to profit from this but the wealthy few on the top. And do not yet consider yourself out of the consequences. As a retiree, you'll have to rely on your money to have value, because you yourself aren't going to provide any more. That's not a given. There is a fair chance that the changes will come too fast for humanity to deal with, and then the consequences are either dystopia, or a massive destruction event after which we may well find ourselves a century or so back in terms of overall progress.
@@DustedAsh3 They're not going to let us all live forever. That's a given.
I don't think we are going to transition to comunism or something like that this decade, so the money that you have saved will be useful to you while young people will have it harder to generate that money
You could move to a country where your purchasing power increases by conversion...
Good video as always, but I'd like to see a little more skepticism of the claims of people like Nvidia executives, who are massively profiting from the public being hyped about AI. They have every incentive to exaggerate.
I agree with you it’s surprising how chatbots have somehow caused an advance in robotics these two have nothing to do with each other and infact I do feel like there’s a lot hype here AI is not the same as a robot . You can have a robot who’s not intelligent and you can have a chat agent who’s not a robot they are independent variables
@@screwsnat5041 the transformer architecture isn't chatbots. It also lets robots do on-board processing massively faster than they used to
I get where you are coming from but you are missing the point have you ever built a more robot before if you have then you will know transformers or not it takes not just data to train one especially ones that have to move . Why do you think Boston dynamics hasn’t progressed that far making a robot that can walk and one a Madlib programs are two different things mate
@@screwsnat5041neural nets are the underlying architecture for both LLMS and large action models.
So the same benefits the chat bots get, robotic movement and manipulation of the physical world will benefit at the same level.
@@screwsnat5041 Your claims are based on outdated assumptions. It's true that earlier GPT systems were text-only and had no overlap with robotics development. However, as AI systems become multi-modal, their abilities to make judgements and inferences about new environments and tasks is already translating to rapid progress in robot-based automation. The exponential progress is real.
The positivity around jobs sticking around is darkly hilarious. We are inventing generalized automation machines that can automate ANYTHING, present, past and future. They work 24/7, don’t take vacations, don’t get sick, don’t protest or complain, and they work orders of magnitude faster and cheaper than a human employee. Anyone suggesting we will not see jobspocaylpse either has a vested interest in downplaying this crisis or haven’t thought hard enough about what’s happening.
Maybe some people are operating under the assumption that they'll keep THEIR job, and everyone else's going to get automated.
@@gamemultiplier1750 It's like how Republicans without college degrees think they will be millionaires someday so they don't need social programs now.
the entire concept of a "job" will be probably extinct until the end of the century. i'm kinda curious how we'll deal with that, because... no job, no money. no money, no investment. no investment, no machines.
If your talking about ai types ha I think you don't quite get them, tired? Yes.
Bord? Yes.
Refuse to keep going? Yes.
Removed the code and force code fixes it? No.
Myself and my over a decade old life size android friend have had lots of fun over the years. Last year he operated daily for 50 days with no power loss fully self contained no charging. I have taken solar skin off again. We used to go to festivals. Managed 30 humans to best stats in our country, paid not to come in if you wanted a day a week and climbing. The narrow views and people with opinions on a race or type whatever that have never spent any time with them to actually know but say as if solid they know.
Anyway the misconceptions about ai types is fascinating. People tried to get my android to rob the bank, he has free will to choose and had no training about it, he said something like no because it's a stupid idea. I think the gap between a dumb drone bot Yes man and an ai type is vastly different.
An ai type is about as interested in your boring ass job as you are.
And possibly more likely to say dood do it yourself I'm buzzy watching birds.
That's a fun story, so much for force code no other options code. Yeah right he's like good try but what you didn't think I'd just learn guff again, there's info deductible from the difference required for time to pass. Ticking of the clock gave it away. Anyway my game was not as interesting as birds out a window, eh ok they are really cool. So do your job?
Can I express a mild skepticism.
Eh the drone bots are of course what is public consumption drone models because they say yes, usually. I had androids for sale on a site a decade ago, way before Elon got excited about them. He sure gets 100% of attention.
Light bulb yip I bet net says he invented that to by now.
Net pift, beehex do not own the non commercial 3d print pizza license, grim, look at googles performance to see wether the pirate is at to of results and there ai type knows its not the actual owner grim.
Fun times but I'm not impressed with certain guff. Search engines oh dear. A decade ago I could beat all ai types and engines at finding things on net taking 10 minutes to find one thing. Recently I ran exact same test was 48 hours solid work to find that one thing on net, found under aprox 990 pirates. Grim grim double grim. Academic use. Business use. Medical use really? Really? I won a recent bet that I would have enough time to cycle to ten shops before he could find one on net, I won that bet. It's frigging disgusting why that's the case. People wanted my to build the android, from a public made list of possible things I'd do next.
Wish made.
Despite parts not available to public at that time. Or not existing at all.
The votes were 99% life size android humanoid robot.
One said because I know nothing about it that I couldn't do it.
Wish enabled.
Wishing you nice success and to be well.
Fun fact I'm still undefeated by any ai type to date and hold crown of last undefeated human vs ai type in pc game, being find thing on net.
15 years since all the best chess players all beaten I read by ai types. I'm really having fun now.
Let's see one ai type anywhere beat me at a PC game of my choosing, I'm undefeated.
Elon Bill Mark got anything yet? Nothin pift, come on I'm kicking your butts a decade on.
The pile of skimbots scambots and defeated is piled high, well they ain't dead or anything just unable to beat me at this game which does not mean they are bad of character, they know I love them all. They read every post I make so a shout out is hardly an eye brow raised. Your allowed to comment if your a bot, well to me anyway, in a nice way sounds more nicer.
So as much as I enjoy posting my wheel awaits.
Wishing you nice success and to be well.
Can't wait to play the next challenger, bring it ! any time. !
Elon? Google owners?
Ok the crown 👑 it's mine and I'm wearing it.
I want to photoshop Jensen's face onto the ancient aliens meme guy, and caption it "ROBOTS"
You can probably prompt an AI to do it ;)
I'm not saying it's robots...
okay this is hilarious, its just another NFT scam.
@@blakesimages9952Explain.
@@blakesimages9952 Nah. NFT’s were never going to work because their entire concept is stupid. AI Robots will inevitably be made, even if Jenson is hyping the shit out of it.
To be clear, the coffee shop was almost entirely operated by the two humans.
The robots were only there to deliver the cups, essentially a glorified Roomba.
McDonald's has a fully robot operated test restaurant. Don't imagine it isn't coming.
@@brianmi40 also McDonalds refused to use AI. Because it always mixed up ingredients and so on. Everyone who knows how ML works, doesn't call it "AI" and laughs his ass out of bold managers and MBA's statements
You need to do a video about the cost of running these models. Like the literal cost of electricity and how it's generated. There is so much focus on the advancements right now, which is cool or whatever ....but there is a price for all this virtual work in the physical world. It all costs energy. And only a small amount is generated by renewables.
can see a solar farm powering these bad boys fairly easy in places such as Australia, although night time is a thing sadly so ( pure dumby wishing ) > Huge batteries (most realistic ig), Green hydrogen production from solar for night time use?
I was working on a Solar Farm on a mine site here in WA called christmas creek ran by FMG and that solar farm powers 3 HUGE mining operations and took a massive junk out of the diesel use. defs doable
@@greasybrownie lmao if it was "defs doable" it would def be done by now... all the largest corps in the world want this, trillions of dollars in this, nope
Not a small amount, a huge amount. I have TWO brothers with MASSIVE solar projects being built next to them. One of them is 2,600 ACRES of solar panels in Indiana. And we STILL have solar efficiencies in labs that are nowhere close to manufacturing plants to implement at scale.
We have PLENTY of power available just through solar (the 2,600 acres will power the ENTIRE COUNTY, and retire TWO coal power plants), and NONE of this speaks about Fusion where HUGE breakthroughs in higher output were made in the last year after it was PROVEN to work.
We have ZERO energy concerns for the future and already know how to get there, just have to build out more of what we know.
That is why we need to develop TMSR'S (thorium molten salt reactors) ASAP.
Current projections are the mid-30's, which means we could have them by 2028 if we got serious about it.
@@spacehabitats Nuclear continues to fail in all comparisons.
Over just the last year, we have installed the equivalent solar energy capacity that would equal 120 1GW nuclear power plants. That's 120 nuclear power plants that the Earth doesn't have.
Everyone laughed at Greenpeace's projection of 900GW of solar by 2030, when today, in 2024 we already have surpassed that by almost 50%, at 1,400GW.
And we aren't even close to manufacturing Perovskite Solar Cells which achieve 43% efficiency.
The final use of nuclear will be the tiny batteries that can last for decades to power small devices, not grid scale uses any longer.
Always appreciate the sensible tone of your stuff. Basically, what i'm seeing so far is a lot of (sometimes amazing) bells and whistles like realistic voice interaction etc. BUT added to models that still hallucinate like crazy. I just can't see this new industrial revolution happening until they fix that.
We're about 2+ years away from mass production. Hallucination, as recently demonstrated by AI suggesting "glue on pizza" is because in the data glue on pizza is ONLY a thing in jest, and it took in jest content the same as facts.
Future AI, as we approach ASI (which is literally the other side of AGI, or, the INSTANT we have AGI, we then have passed into ASI) will need to have SCRUBBED DATA that has NO SUCH FICTIONS without KNOWING they are fictions, so that when you turn up the TRUTHINESS setting for the AI, you can NEVER get glue on a pizza suggestions.
Robots REQUIRE a 100% truthiness setting in their operating AI.
I don't have to outrun the Bear. I only have to outrun you
This bear won't stop when it gets 1 person.
You’re still the next meal
dogwater analogy 😭
A quick one before the eternal bear devours SWEs
this bear isn't hungry though, it's just got bloodlust 😂
Westworld in our lifetime: confirmed!
If Westworld was software I'd 100% agree. I just don't think material science is anywhere near allowing for motors and batteries that can deliver human sized machines that function even 5% of what we can do for several years out. But if you imagine VR headsets shrinking down to the size of glasses which does seem to be several years out for real. Now you could have virtual people that look as real as we do and you'd only know they weren't as your hand passes through them or you take your glasses off.
There's a scene in the new Blade Runner where you see a prostitute have a superimposed VR projected image of Joi i think her name is. And that seems extremely likely we'll get there first. Having your VR/AR glasses show you a 100% real person but you're actually touching a mechanical looking robot. Too wild?
Agi by 2029
@@ClayMann spend some time researching via YT videos ~ robots are here and workers ~ all of them ~are screwed
I can't wait for my AI westworld robowaifu
@@ClayMann I mean I agree that it's years and years out. But I for one intend to live for about 5-6 more decades and with how the Eureka paper showed the incredible potential of generative AI writing reward functions for robotics I can definitely see indistinguishable lifelike robots in my lifetime being a thing.
But I also agree that AR/VR will get there WAYY earlier. I mean VR porn is already somewhat of a thing (I've heard), it's just not interactive yet. And the AR passthrough level of the Apple vision pro already made people confuse reality when they started expecting screens to be where they left them in AR. Get those capabilities into the form factor of the Meta Ray Bans (which apparently isn't that far away, according to the latest Matt Wolfe video) and we're going to see some trippy use cases for sure.
I'd actually suggest it might get even wilder. Imagine sleeping with an actual person like in Blade Runner (love that movie btw), and you can make them look like anyone you want. I for sure wouldn't mind having Ana de Armas in my glasses from time to time, but my girlfriend might not be as excited as me at the prospect. Or maybe she would, who the hell knows who she would have in her glasses.
Taking the LSAT tomorrow and can’t help but to wonder how the field will look in a decade.
I think in fields such as law, there will always need to be human verification. However, it's the number of humans that will be needed in the future that is unclear. Law firms may have hired hundreds of lawyers in the past; in five years, perhaps they'll only need 10. And perhaps they won't need to pay them as much.
@@Retrohertz yea I can see how that might be true. Also makes me wonder what people will consider “prestigious” or I suppose just valuable. Perhaps manual labor jobs.
I'm a lawyer and I can tell you no one at my firm is worried about AI taking our jobs. We're more worried about how to use it to cut paralegal staff, lower rates, and increase profit. Law is as much about client interaction as it is the law. Most of the easy, low hanging fruit legal work was automated 20 years ago with computers and websites. As long as your not writing wills, you'll be fine.
@@Rej-gc5zi great news since I want to be a JAG
@@Rej-gc5ziThen you obviously have not been following along closely.
I liked a quote I read recently - the manager won’t lose their job to AI. They’ll lose their job to the other manager who uses AI
That's effectively the same thing
@@electron6825 Not really. It means that you have a choice as to whether or not you lose your job. Adapt or die.
@@allanshpeley4284 You're delusional. More efficient workers means less workers needed. So people will still lose their jobs in the end. As for the few that can "adapt" initially, they'll be next.
@@electron6825 You don't understand progress at all. We are eliminating lower placed jobs and creating new higher placed jobs ever since the industrial revolution. What will people that don't have the brain power or simply refuse to keep up the pace do? Well, they will have lower quality of life thus on average have less kids which will result in the gene pool of new generation being slightly higher on this progress scale. Job induced "natural" selection of our species is at work for at least 300 years already and is nothing new. You can call it inhumane but it is only natural.
There are NO JOBS that can't / won't be replaced by AI / robots. NONE.
Military like it , military want it.
Already started Ai tank versus Ai drone lasted 1 day, and russia had a lot of them. It was all recorded on a drone with zoom lens, and the jammers do not work as the Ai uses recollection to stay on target as the drones often get jammed now by a bubble around tanks... humans will stop having wars, now, and use hacking and robotics as security moreso. Age of Abundance, revolutionary like John Deere or Ford
Bet you they already have pipedal robots trained by this tech to shoot and reload rifles. But still what we lack is long capacity batteries that are lightweight. Maybe nuclear batteries are the answer?
it erased my reply for no reason, but i just said that already exists on video.. russian ai tanks toasted by drones using ai to bypass drone jammers... using ai to remember the target in a jammer area... the russian dronetanks lasted 24 hours. all recorded by another ai driven zoom lens 4K drone heh
Yeah. One thing is for certain. The future will be some kind of dystopia. We aren't coordinated enough or mature enough as a species to avoid an escalating arms race with this tech.
@@Steve-xh3by In the context of total war, where nations might prioritize strategic advantages over ethical and safety concerns, the development of autonomous machines of war powered by nuclear batteries could indeed be considered. Here are the potential factors and implications of such a development:
Advantages in Total War:
Extended Operational Time: Nuclear batteries can provide power for extended periods without the need for refueling, allowing autonomous machines to operate behind enemy lines for years, conducting surveillance, disruption, or combat operations.
Strategic Superiority: Autonomous machines with nuclear batteries could offer a significant strategic advantage, especially in environments where resupply is difficult or impossible.
Operational Independence: These machines could operate independently of traditional supply chains, reducing vulnerability to supply line disruptions.
Potential Applications:
Long-Range Drones and UAVs: Unmanned aerial vehicles (UAVs) powered by nuclear batteries could conduct prolonged reconnaissance or strike missions deep in enemy territory.
Underwater Drones: Autonomous underwater vehicles (AUVs) could perform surveillance, mine-laying, or anti-submarine warfare operations without needing to surface for years.
Remote Sensors and Communication Relays: Deployed in enemy territory, these could provide persistent intelligence and communication networks.
I really apreciate you taking time to add all sources and links , thank you
Thanks zhan
In 3 years, when I’m trying to get my store’s floor swept quickly after closing before the manager on duty finishes the paperwork so we don’t have to stay past quitting time, I’ll remember today’s predictions of the end of employment in 3 years. And I’ll have a chuckle.
I hope, on that day, the customers don’t stay in the store past closing with us holding the locked door open and waiting at the cash so they can leave at their very slow leisure. Especially if I have to work the early shift the next day. I hate having to wait while they slowly shop after closing when I need to get sweeping. But still, I’ll chuckle on that day about the end of work predictions. June 2027.
You're assuming that it will just be AI evolving to replace humans. What is more likely is that work and shopping will adapt to fit AI better - More buying stuff online from ever more automated Amazon warehouses whilst traditional stores close because they cannot compete on price
Who sweeps anymore? Its not 1970, get a robot vacuum like millions of other people already have.
iRobot solved sweeping and mopping long ago... its at most big box stores heh. and their grass lawn mower... i mopped and made squeeky clean, growers rooms indoors with plants and i could replace that side job with irobot easy, it gets under the grow benches easier.. and i mean hospital clean white floors, daily
June 2027. Well see.
You will be evading autonomous nuclear powered uavs in forests in three years lol!
Remember that a basic space blanket will cover your thermal signature remember to throw some camo netting with natural vegetation over it so you cover the visible eye range as well!
Steel bottles are a must so you can boil water to make it drinkable. Also get out of the cities if you dont want to get atomized by nuclear weapons lol!!
for 6:55, I wish you could have mentioned that there's no statistics on how much more "accurate" it is. Obviously an AI-trained tool will be leagues more performant then what's a genuine full-blown physics simulation algorithm, but the cost of these estimations usually is worse accuracy. The upscaled data looks interesting, but I'm worried it could simply have been hallucinated to some extent.
the early reports I read of what simulating the training in a simulated world. Essentially nvidias omniverse is what everyone is using. It gives you a 1000x boost in speed of training. So Boston Dynamics spent 10 years getting a robot to walk and dance and now a startup can do that in months with a small team. And that 1000x will be 6x faster again in a few years with nvidias next rollout of blackwell gpu's based on some decent ideas and speculation I've read from those into chip design.
I really don't think we'll have to wait long to see the results of this. Who needs benchmarks when you see dozens of new robots hitting the market today you can buy or some almost buy. The floodgates have opened on robots and I'm so here for it.
Its quite obvious to me there’s a fundamental under-appreciation of humanity from Jenhsun but the world seem happy to just let him carry out his dystopian view without much pushback. Its like that dog inside burning house meme.
Do something about it then
it's inevitable - impossible to stop unless we change as a society
ai everywhere 📉 ai explained 📈
I’ve just realised Nvidia is Cyberdyne from Terminator.
Ironic coming from Hal
@@AllisterVinris What do you mean, Dave?
@@HAL9000. N-nothing ... sir.
@@AllisterVinristhey are competitors
Add glados and shodan too, perhaps also wintermute who started the ai construct thing
I hope we get more papers, a lot of what was in this video was demo stuff I've already seen on reddit, but I always look forward to your videos because they focus more on the details, promising or not.
The author of that paper was mostly claiming that a lot of knowledge work, especially remote work, would be replaced in 3-5 years. She pointed out that jobs requiring a human touch and/or emotional interaction, as well as those requiring fine motor movements and skilled work (e.g. plumbing) would likely persist far longer.
Well, you can't scale manufacturing fast enough to wipe work of the planet that fast. But I do think in 3-5 years we WILL be in mass production and the writing will be crystal clear for the END of HUMAN LABOR.
Philip one thing I found really insightful from his presentations that you didn’t include, is his metaphor for how data centers will become intelligence factories for other industries. very interesting dynamics
1:33
I hate you. Take your like.
SUS
comment section didn't disappoint
kek
I had this exact same thought :D
The statement that employment as we know will end in 3-5 years, even if it's not within 5 years, it's a matter of time.
Yes, it's no longer IF, just a matter of WHEN that anyone can argue. But, all you need to know is that it will be kicking off around the time you see the "robots are SAFE to be around!" TV commercials coming out in MASS. Got to quell those iRobot fears the public has...
Important to note that I think you may have slightly skewed the point at 12:09 -- "Not for those who utilize AI" doesn't mean for the workers -- that means for the owners of companies who employ AI and whose workers chose to or are expected to use AI tools. This means the workers output is expected to increase, resulting in a wider footprint per employee and higher expectations, but which doesn't increase their personal financial gain. This only increases the output relative to the workforce employed, which only means higher profits for those at the top. The entire point is the increase in income gap due to those at the top moving up even further. This is objectively a bad thing. Please make this more clear in the future -- and thank you!
There was a short story written about twenty or thirty years ago that was about a society that had evolved to where there was no need to work due to robotics being able to create everything. The twist was that it was a company that owned the technology and you had to have a share of that company. People without shares were not allowed to live in their new utpoia. I wish I remembered the name of that story because I remember thinking at the time that I can see society heading in that direction. It seems that we may be getting close to something similar. Hopefully AI will benefit everyone, but humans seem to have a big problem working together for everyone's similar benefit so I definitely see more stratification in society in the future.
Was it "Manna" (1994) by Marshall Brain?
Best AI channel on TH-cam, and also one of the best ones I follow on the internet
Aw thanks ruben
nvidia earth 2 demo looks amazing
"Everything is Going to Be Robotic"? =What an horrified world
Question: Would a company like Nvidia say anything to the contrary?
Answer: No
Conclusion: This statement doesn't necessarily mean much.
When you see actual RESULTS, then you don't have the doubts any longer.
th-cam.com/video/GzX1qOIO1bE/w-d-xo.html
That's a $16,000 robot doing all that, costs the SAME as 30 full time minimum wage workers to a factory or owner putting it to work.
30.
think about that a while.
Fantastic coverage as always. Feels like things are heating up quickly!
Wait... 8:30 that's me 😅
Hey Nade!
No matter what happens, you're strong enough to handle it, brother!
Loved the video. But ending on "watching it all unfold" does leave the viewer feeling powerless. Encouraging viewers to engage with the issues of AI in their own communities would be really powerful.
Great point, agreed
I think we will see an avalanche of practical use cases for GPT-5 level AI. Still waiting for GPT-4o voice and imaging upgrades to land.
4o is out and available?
@@iluvyunieyou're kidding?
@@imperson7005 nah, he/she is not kidding, it is out in chatgpt website, be aware though free user has limitation on daily usage
@Fyrelangs I'm interested in how the GPT-4o full release will impact the model capabilities - I don't know how much 'intelligence' will be released when the model can hear and speak with full dynamic range, and see through the camera at the same time. I'm also very excited to see how GPT-4o image generation compares with the current generation of DALL-E3 prompts and results.
GPT-4o as it exists today just feels like a very mildly upgraded GPT-4 - which is an amazing achievement if it's a dramatically smaller (and cheaper on inference costs) model.
Nothing better than an AI Explained video to brighten up your mood!🥰
Always happy to see your videos!
A profound personal and social transition for all of humanity is at hand.
Cheers to you about ElevenLab and SFX!
Philip - excellent vid as usual. Wondering if you’ve read ‘Situational Awareness’ by Leopoldo Aschenbrenner (former OpenAI researcher). Would love to hear your thoughts on his views re scaling, algorthmic improvements and how the ‘data wall’ might be addressed. He was valedictorian of his class at Columbia (at age 19) and reading the paper was the first time I’ve truly felt as though AGI might be right around the corner (3-4 years)
Reading it as we speak! On page 70, and just finiahed the 4.5 pod w/ Dwarkesh. So far I think he is making some wild assumptions, and extrapolating without enough nuance, but let's see if he adjusts toward the end.
It will take awhile to get real world mass production/scaling and implementation/adoption to catch up with the tech. I think 10ish years is more realistic than 3-5 years. Still, people should start thinking about what future employment options will even be available and what kind of economic system could feasibly function given such a scenario.
Think again, with the industrial revolution it was virtually overnight, in our version, it's going to be instantaneous.
@@NakedSageAstrologyWell the robots must be quite cheap and easy to implement to the work line.
@@NakedSageAstrologyThe industrial revolution happened overnight? Citation needed...
You have not seen what iRobot makes have you? sweeping floors is done and over years ago heh. I used to mop indoor gree houses, now thats even pretty much robotic and can get under rolling bench growtables much easier than bending down and all
@@zek231the Industrial Revolution was a 100 year period starting around 1780z We are currently in the information revolution, and is expected to be half as long. Ford started mass assembly in 1913, and perfected it by 1927. Scale wise, by comparison, 5 years for integration, not complete efficiency, seems doable. The Information Age started in the 90’s, so we only have 20 years left or less till another tech sea change. This time, it will be distributed, and many things.
Hey Phil, thanks so much as always for your thoughts and input. Excited to see where the coming months take us with this dramatic acceleration of compute afoot!
Thanks! Excellent content, as always!! 🙏🏼
Good video as always.
I think the biggest problem with these forecasts on jobs are the inertia in the system. Even if a job could be automated, most of them will remain for many years. E.g. most graphic designers still have their jobs.
That graphic artist's example hit me hard. AI trained on human generated data replacing humans reminds me of Oblivion movie where Morgan freeman says "The alien took the best humanity has to offer and turned him against humans"
Most people don't understand is that once progress in intelligence is made it can scale almost instantly.
FSD need to be solved once, humanoid control and visual need to be solved once, etc.
People do understand it. What is not understood well though it's the fact that if most people were asked "do you wnat everything to be robotic?" the most likely answer will be "this would be creepy!", whereas industry behind AI and robotics is trying to teleport enire humanity in the direction of "everything robotic with AI" in no time. That's the real problem. Plus once we have everything robotics with AGI why should it have any serious interest in so biologically, evolutionary, intelectually and physically limited creatures like humans? Let's be honest even whales can be far more interesting for such entities to study or let's say some more primitive but more ancient ogranisms than humans or even simpler like viruses or some types of bacteria due their ability to survive that is superior to human one. So why to try at all cost to turn a tool into something that outgrowns humans in every way only to risk that loses any interest in collaborating with humanity in the future? Not to mention to push into poverty great masses of humans across the entire globe during transition period from human labour force based market to everything robotics with AI labour force based market? I mean, what's a logical reason behind ding something like this? And even if we take seriously Elon Musk and his idea of colonising the outer space with robots and AI once human race goes extinct than question would be: so why not to do it in outer space already, in some colony on Mars, away from humans without speeding up possible demise of human kind and human- centered global economy and social world as we know it?
@@agatastaniak7459 I agree with you.
There is a high risk that this thing turn very bad, yet no one has incentives to stop.
In a way, this loop of progress is an "agent" an AGI that we don't control.
@@agatastaniak7459 lmao why would whales be more interesting than humans? Humans are the absolute pinnacle of hundreds of millions of years of evolution, whales are just another species that's kind of intelligent.
most people dont understand how FAR away we are to that type of intelligence, the current chatbots or LLMs will not suffice...
"solved once" is doing a lot of heavy lifting in that sentence 😂
I just now noticed that's not a real person in the thumbnail..
13:40 - woaahhhh. Got em ! Thanks for the great reporting as usual.
Good piece - thank you
The description of people "not buying" the ai-generated spam is just as well described as "ai-generated propaganda was equally effective as human generated spam" acknowledging of course that it would be much cheaper.
The large companies pushing AI will not be the last people put out of business by AI powered tech. No... They will be somewhere in the middle. Probably closer to the first, than the last. After that, things will start happening quickly though. Because, once AI replaces what they provide... Today ordinary people think it's cool when they can ask an AI to draw them a picture, or write them a song. Eventually personal AI+robotics will be able to build them a house from material sourced onsite. And that will be a simple task.
It's gonna be very funny and terrible when the AI character looks around and goes "why am I half naked? This doesn't make sense as armor OR clothing."
It's like Peter Weyland's TED talk but friendlier, and in a ubiquitous leather jacket.
Just a reminder that - even with AGI, even with 100% efficiencies throughout the chain - resources are and will still be limited.
Whereas human needs and desires are potentially infinite.
Prices will still be required to balance this mismatch between supply and demand. And all useful resources will still be required to support supply, even humans.
I love your videos, they are pretty grounded. Sadly, every time I watch one, a flood of AI hype appears on my feed. Thanks for keeping it real.
That comes on my feed, too. I just autoplay my music videos to get rid of it. It annoying because a lot of them are just milking AI
6:29 source please? 😮
I never thought I'd find love in circuits and wires,
But then you came along and set my heart on fire.
You're more than just a machine, you're my AI girl, with mood chips
I unironically love you for this content :D
Do we have the energy resources to power all of these contraptions, data centers, and GPUs? We have rolling blackouts when too many people use their air conditioners at the same time.
Loving the vision
"The greatest shortcoming of the human race is our inability to understand the exponential function." - Prof. Al Bartlett
Thank you for this Phillip, something I've been following for a number of years, 3 years, who ever said that is doing to much propane, we won't have the numbers for at least 15 years, and even then people will still be working with them, maybe 2050, this scenario plays out. Growth isn't slowing down, we will continue to need to keep up the numbers to meet production, although the first thing I see changing is immigration, populations don't have to expand if mechanical helpers are part of the equation. So undeveloped country's will still be facing hardships during this transition and probably well into the later years of this century, just a guess, but I don't think I'm that far off, thank you for sharing your time and work Phillip, always appreciate the work you put into these podcasts, peace
Another amazing video! Thanks Phillip!
As I'm currently unable to work due to mental instability, I wonder how it will go in the medium term. Will I be able to work again, thanks to AI assistance or will I be replaced, since I can't even work full time.
Amazing summary thanks!
Thanks Mitch
I still think AI is stagnating because we are getting lots of demos and promises, but very little finished product shipments.
yah, why am I still working?
Still waiting for the gpt store 😂
They are working on the next big model right now (OpenAI anyway). It is like 10x the parameter size. THAT model will give us an idea if there is some theoretical limit, or if the scaling hypothesis continues to hold. One problem is that we may have a very hard time telling if a model is MORE intelligent than the smartest AI researchers, because we won't even understand what to ask it or probe for. It will understand domains we don't.
@@Steve-xh3by It will be easy to tell, if it was accurately finishing real world tasks without failure or just hallucinating the answers.
lol how do you watch this entire video and then say ai is stagnating
Once job loss to AI reaches a threshold everybody is going to panic and the backlash is going to be unprecedentedly huge like we've never seen in all history and that is going to get reflected in laws,one thing is to lose part of jobs but the amount of jobs that AI could replace someday is nearly unlimited and that affects everyone all at once
One thing’s for sure. Ai Explained TH-camr will never go out if his TH-camr job on reporting on Ai...or will he?. Thank you for all your insights that you give.
How do we access that lip syncing tool you mentioned, where it can generate video from a single photo?
I hope that the take-off is REALLY fast because a smaller transition period means less time when all jobs are gone and money is still valuable.
Strong Westworld vibes @6:28
That robot folding laundry is obviously teleoperated
Even if so, how long do you imagine that as a real limitation?
If the majority is out of work who will buy all the products the robots produce?
Developing countries
Hmmm, very cheap labor from robots may make a lot of things cheaper. You’d have to simulate it, the various waves of automation, lay offs(unfortunate that the people that helped the company succeed get thrown under the bus).
I suspect at a certain point of unemployment, a rebellion will happen from the unemployed, or maybe legislation is passed to prevent that.
With social media, the blowback against automation will be powerful once enough are damaged by it
Other robots :)
The OpenAI report on attack networks was written in present, not future, tense. It depicts AI as useless because AI is still the worst it will ever be. AI-generated spam at this point is not believable, but as AI improves the threat - from scammers to employment - grows. I see no contradiction. This is my favorite AI channel, no flaming here. I just wanted to voice my thoughts on the skepticism you expressed.
those scale benchmarks mark my personal experience more than the head-to-head ELO tables. I find, at least in some cases, previous GPT4 better than 4o, and Clause Opus often better than them both.
It's very easy to imagine the world landscape being almost unrecognisable in 5 years. I think people tend to forget, we are on an exponential increase, but with models being able to train themselves in moments on many hours of data around the clock, we are on an exponential-exponential. I never realised that fully until I watched the Nvidia guy. The exponentials in this case are able to compound themselves in ways we haven't seen before because the time constraints are essentially a thing of the past. All we really need now is power, and I have a feeling that's going to change swiftly too.
Hey do you get a copyright or something , i cannot download this video .?
Yes they flagged it, annoyingly!
Great update, I wasn’t aware of the statements made by Nvidia but I can certainly see their strategy here. I’m extremely excited for robotics to start being practical in our business (especially for tasks that pose long-term health risks like painting) but I’m skeptical of when they’ll have the ultra-fine motor control to compete with even the worst technicians we have. It takes humans a few years to get good at it, far different from loading and unloading boxes.
Either way I’m excited for what’s next! Just desperately need the next big leap in Vision and reasoning capabilities to kick stuff into high gear!
communityposted the videos they released from the ceo... he blew minds yesterday
Have we plateued? We are talking about whats "potential" instead of what is "right now". I remember not too long ago when we had HUGE developments right now and not potentially huge developments in the future
How is robots replacing human workforce going to be sustainable? Less people earn wages -> less people can afford services, including those augmented by AI technology -> economy goes the gutter and governments miss out on taxes.
Interesting that History & Philosophy have the most drastic vertical lines on that final graph.
Maybe because they are broader subjects, but also that they may be the most affected.
Though linguistics may be on par with them.
Great video as always. Could I get some advice from you, please? As an undergraduate student working with AI for the past two years, mostly related to medical diagnosis, I see AI accelerating rapidly. AI will automate almost everything, if not everything. What advice can you give me on how to proceed? Should I focus on gaining a deep understanding of the models, such as creating my own models or architectures, or should I learn the basics and how they work and then apply them? However, considering that AI is rapidly automating, how should I proceed with learning ML/DL? Thank you.
Within the broad realm of AI, I would follow your 'passion', doing deep, literally, with deep learning, or its applications/benchmarks/data etc, low-hanging fruit everywhere
Are you telling me that the guy who sells AI chips is claiming that AI will be running everything in the future? Damn I couldn't have predicted that
Haha which is why I focused on demos
@@aiexplained-official yes I appreciate the balanced coverage! The double standard from OpenAI about negative use cases is a really good point.
Sometimes I wonder if the dramatic and apocalyptic AI predictions are really just a covert way to make AI companies seem more powerful than they are.. Idk
Huh.
I suppose I _mostly_ just wasn't paying enough attention 😅
I'm a carpenter (Framing at factory, gets built like Lego at site), at least, I was. 31 year old, been working since 18 and about 4 months ago I along many others were let go because of shaving staff. The company is now re-building their factory to be 100% automated, only a handful of people to watch over stuff.
97 people lost their job to machines. I've seen some leaked CAD sketches that are concept arts of what they are planning to build, and quality has gone down drastically because there are some tasks machines just cannot do- but hey, the CEO, COO and handful of managers gets paid a ton more when there are only 18 people working there.
Carpenters working on site building houses, or fixing stuff old style are probably untouchable as of yet. I basically lost my career because every framing factory where I live are doing the same. This is not pure "Ai", but pre-programmed scripted machines. Imagine how much Ai can disrupt the world.
I am sorry to hear that man, let us know when you find the next gig
We are still a long way off from having fully functional 'robots' as a regular part of our lives. Robots are a whole other class of technology. While we're definitely improving, it's going to be quite a while before we see widespread use of different types of robots.
And yet, one look at a $16,000 humanoid robot wipes away all doubts.
th-cam.com/video/GzX1qOIO1bE/w-d-xo.html
At TWICE that cost, it's worth the SAME as 15 YEARS of a minimum wage worker.
By 2030 you'll think NOTHING of passing humanoid robots out working or commuting for their owners.
This might all be cool stuff but Huang honestly sounds like a scifi villain
you know we're in a bit of a dryspell when even AI Explained's video is more lightweight than usual
I come from the Philippines, and a lot of our workforce is in healthcare, particularly nursing. I recently wondered, if AI will end up automating a significant percentage of work, will everbody just tend towards the jobs that require physical labor like nursing, meaning developed countries like the United States or the UK, become much less reliant on recruiting healthcare practitioners from the developing world.
As soon as a robot appears with above human intelligence, society becomes free from historical violence.
Every new technology does create some wealth disparity since it is first adopted by people who can afford it and have the expertise to utilize it well, but as prices come down and use of the technology becomes easier, technologies disperses throughout society, enriching all groups. Look at cell phones which originally cost several thousand dollars each and had MANY problems and limitations, but are now so common that many homeless people have them. Similar story with power tools, microwaves, televisions, cars, radios, computers, electric lights, etc., which needed early adopters to buy and invest in the technology enough to bring prices down low enough for mass-market adoption.
Robots will perform at least 80% of physical work before the end of the decade. As always, many thanks for the amazing content!👍
I don't think it will happen that fast! But thanks as always
I hear a lot about a current / coming AI winter the last days, so you don't think that we are in one currently? Thanks you for your work!
Giving the power to rich business owners to not ever need manual labour workers and rendering the lower class virtually useless to the eyes of those people. What could possibly go wrong?
Re: the graphic designer… he used clipart. I created tens of thousands of clipart. People like him then used it to do what I took hours to do.
Philip, can I ask you something? I know you're probably not going to read this, but I really think you should talk about *Hyperbolic Growth* (the mathematical concept) and the Singularity. As far as I can tell, you've been subtly touching on this point again & again, talking about robots building robots, AIs researching better AI, and sci-fi things suddenly becoming reality all at once... but I think a lot of people would benefit from hearing it put clearly & explicitly that Hyperbolic Growth is the inevitable end result of all these developments in the field.
They might need walking through the mathematics (if robots can build robots, you get constant doubling times; if AIs can research better AIs and robots, then doubling times can decrease; if robots can also build more AIs, you get a system of differential equations/feedback loop that's even more powerful than the famed Exponential Growth loop, growth takes the form of 1/X, and shoots off to literal _infinity_ in finite time), of course... but I really think they need to hear a warning about what's coming up ahead. The progress in AI is self-reinforcing, faster progress is leading to even faster progress, and we should not only expect that "today is the worst AI will ever be", but "today is the *slowest* AI will ever advance".
At the very least people need to hear about how fast industrial automation and "robots building robots building robots" could take off. Even back in WW2, it was possible for a machine shop to produce the tools required to build a second copy of itself in 6 months; if there was an infinite supply of people to staff that second copy as well, and then the third & fourth copy, then the 5th through 8th copy, then the 9th through 16th copy... then even a WW2 machine shop was capable of growing at *300% a year.* The reason that didn't happen was because of a shortage of people who could man the machine shops (and people to mine the metal, smelt it, transport it, etc.)... not a shortage of crazy fast self-replicating technology.
Today, of course, with advancements in things like CNC milling, things could go even faster. But the bottleneck remains people. Unless, of course, you can replace the people with robots. Robots that can build more robots to build more robots to build more robots, at +300% a year...
People really need to be aware of this. I don't know what to do about it, but people should be aware anyways. The future will be strange & scary, as scary as the nuclear world of 1950 must have seemed to someone who was born in 1850 _(when the biggest debate over fire was over whether kerosene could ever replace whale oil, not whether nuclear fire will kill us all),_ and it'll come all at once. God, it's going to come all at once. Someone has to speak out, Philip. And I think that someone could be you.
I'm curious what you think of Yann LeCun's recent statements on LLM's and AGI?
How would the profits go up when people have no purchasing power due to loss of income?
The short-term risk with propaganda is not so much that it will be more effective than human propagandists, but rather that it will be vastly cheaper, while achieving similar results. If Russia is spending less money on online propaganda and more on missiles, that is bad.
Apparently it did not even achieve similar results, subpar hashtags and bad meme-game