SHOCKING Robots EVOLVE in the SIMULATION plus OpenAI Leadership Just... LEAVES?
Vložit
- čas přidán 24. 05. 2024
- Let The Robots Do the Work:
www.skool.com/natural20/about
Join my community and classroom to learn AI and get ready for the new world.
LINKS:
Faster LLM via multi-token prediction.
/ 1786508877564969107
The Video by Elvis covering this paper:
• Better and Faster LLMs...
arxiv.org/abs/2404.19737
Exclusive: Two long-tenured OpenAI executives have resigned, the latest leadership changes in the wake of Sam Altman's tumultuous ouster and rehiring.
www.theinformation.com/articl...
Open source AI glasses for developers,
/ 1786383198190264394
DrEureka
/ 1786429467537088741
eureka-research.github.io/dr-...
github.com/eureka-research/Dr...
A potential solution as to how compensation of copyright holders could technically work:
/ 1786616083991990534
arxiv.org/pdf/2404.13964
TIMELINE
00:00 AI NEWS
00:10 Figure 01 on 60 Minutes
01:00 Dr Eureka
11:52 OpenAI Leadership Leaves
13:00 Solution for Copyright AI Art
16:30 Multi Token LLM Prediction
20:50 DEVIN 2.0
25:55 Open source AI glasses
#ai #openai #llm
BUSINESS, MEDIA & SPONSORSHIPS:
Wes Roth Business @ Gmail . com
wesrothbusiness@gmail.com
Just shoot me an email to the above address.
![[柴犬ASMR]曼玉Manyu&小白Bai 毛发护理Spa asmr](http://i.ytimg.com/vi/0TsXQ7z2Dh4/mqdefault.jpg)
There's one other big piece of news. I just launched a forum plus training courses for people interested in taking their AI knowledge to the next level:
www.skool.com/natural20
I won't spoil the surprise, but I STRONGLY suggests you check it out asap (there's a time limit)
60 minutes is the most famous talk show to interview presidents
You want to do this y'all... trust me. FYI
u should do a live chat when u drop some of ur videos so ppl can discuss what ur talking about in the chat
[and now it's weird not seeing ur head in the corner of the vids, maybe u will bring it?]😊😊
Is that forum STUNNING and SHOCKING? Otherwise, I won't go.
You have promised to stop those SHOCKING titles, you lied
If only we could now train these AI researchers on EXPOSURE, ISO and COLOR BALANCE.
That would be stunning.
Turns out color grading will be the last job in the ai future
I was so confused when you said you weren’t sure what 60 Minutes is… what generation are you?
ahaha same!
It was clearly sarcasm.
You can tell by his voice he's obviously 11 years old.
He was born in 2096.
hes bald
That almost everything in american social media is talked about in absolutes, including devin... that should make everyone quite concerned....
Only a Sith deals in absolutes. There must be many Sith on social media.
@WesRoth I love how often you post new content. Don’t know how you keep up with it all, but thank you 🙌
LOL, I thought I was the only one, everytime I see them demonstrate how well the robot can keep its balance, I'd think to myself, "ohh they're not going to like that".
I don’t think most people understand how impressive Eureka is. I think most people have the wrong idea about what humans are best at. I think many if not most of us think that you human beings are best at thinking and that when computers can out think us, that means we’re getting closer to AGI. But it’s movement that human beings are best at . For instance computers have been able to outthink us at chess for quite some time but five years ago it would take an entire PhD program for someone to program a mechanical hand to the movement with three-year-old And my mind what Eureka is able to accomplish with AI in robotics is a much bigger indicator that were close to AGI and ASI.
Marvin Minsky said that robots are great because they can survive in harsh conditions, such as a melted down nuclear reactor. But at Fukushima, a robot was unable to turn the door knob to get into the place.
dumbest message ever....
That is not what humans are best at. That is what is hardest to make robots good at
The copyright infringement by training thing is incredibly stupid. AI shouldn't be treated different than a human that can study something and be inspired to create something similar.
As an artist and singer-songwriter musician, I would have to agree though I know a ton of artists don't. I think AI is less likely to plagiarize (adverntently or inadvertently) than humans who don't even always know they are doing it.
@@adangerzzArtists, songwriters, & creators don’t have to worry about AI. As good as AI will get, we still need true inspiration & insights from humans. AI won’t replace Altman. Right?
@@tringuyen7519 - I can't honestly say they don't need to worry, I think they do. People will use AI to create art and music because it is free and they will think it is good enough, thereby potentially depriving an artist or songwriter of work. I have used AI to generate images to use with something I've written (myself). I would NOT have hired an artist to do the illustrations because I have no money (I would just do without). However, if I had a real project (like say a CD release, though I don't even know if musicians release physicals CDs anymore), I would hire a human artist (as I have in the past rather than doing it myself). I do imagine there might come a point where humans (or some humans) are entirely fed up with AI generated content, art, music, etc. and long for the human touch.
@@adangerzz- as a person that can't take most of the current musical offerings. I welcome the ai overlords to let me prompt music to my old and crusty taste.
Hey ai, make me an album with Niel Peart on drums, Stevie Ray Vaughan on guitar, Manfred Mann on keys and Steve Perry's vocals with Ann Heart backing him up. Oh and I almost forgot, add Chris Squire on bass. Thanks, I'll have a listen when you're done.
@@robinvegas4367 I confess, I don't listen to much that has been produced in this century. Stuff occasionally gets trough to me somehow, but honestly, I prefer jazz standards and/or music from the 60s and 70s. There is some modern music that I've heard and liked.
Im very excited for Frame due to its open source foundation. Between those and Open source models like Llama 3 I think there is some serious potential for innovation and invention!
I'm looking forward to it, too. I haven't bought one yet, waiting for reviews, but I have a lot of projects that involve working with physical objects in the real world where having an AI that can see and understand the components, and help me solve problems in real time will be very useful. It's what Google Glass wanted to be years ago, but now we're getting it yet even better.
The Orange thing on the Frame glasses is actually a battery charger you can carry in your pocket. The things on the back aren't speakers either.. They are the internal batteries. It pairs with earbuds instead of having onboard speakers.
Social media algorithms and Mainstream Media editors magnify extreme opinions (0% versus 100%). I ignore anyone who says AI is the Messiah and anyone who says it’s useless, as both are wrong. That’s why your channel is so good 😊
All artists are trained on existing works. If they don't want people or AI to.
look at them and be influenced by them they should keep thier art secret and not show anybody.
valid point
San Franpyscho is brilliant
I'm guessing that was the $65 beanie, though.
Rewards and punishments or extrinsic motivation isn’t an efficient or sustainable way to evoke behavior. It works for animals and to some degree for young children. It’s blatant manipulation and humans resent being manipulated. This is why companies constantly change the way sales people are compensated because it’s based on extrinsic motivation. My guessing the AI created rewards are finding a way to use intrinsic motivation which is very sustainable.
Good point. I guess I'm not really fully comprehending what the nature of the rewards are, for AI. To begin with. And yes, aiming towards motivation rather than specifically "rewards oriented" seems a better model.
This is the kind of stuff that the " never AGI" people don't understand.
Like, the dog on the ball worked because the computer simulated a million different ways the dog could move to keep the ball under it. So what happens when a computer simulates a million different ways to design a computer chip, or build a model for efficiency, or train a model to think more logically?
Even with humans in the loop, we're getting into recursive self improvement, and the party just started.
what makes you think its extrinsic motivation? they have the logic for their reward system coded into them? thats intrinsic. they do smth and it showers positive or negative feedback. they know its more positive feedback if they reach a certain goal they either get told to achieve or (according to prompt) come up with themselves by their problem solving algorythms.
lets take the robot on the ball. it falls down, negative feedback, it stays on top, positive feedback, it stays on top longer, MORE positive feedback, it slips -> some negative feedback, but less as if it just falls down without catching itself again, etc.
thats the same as with humans trying to balance on it (in case we want to learn it or be better than the other humans in it) and being happy (endorfines) if we can stay on it for longer, we get unhappy if we fall down and we get even more unhappy if we fall down and dont make a good landing, hurt our knee or smth. one is taught by IT-guys coding it into the AI source code, the other is taught by natures laws imprinting it into our DNA. its virtually the same.
@@michaelwoodby5261 heaven
have you thought or heard anything better?
The whole Artists and Copyright thing seems to always forget one key detail: *The Artists themselves and how they have developed their art through the years.* Every artist has been trained, whether that has been done "formally" through schooling institutions or being "self-taught". They have practiced by tracing, doing studies in different styles by other artists, experimented with varying color palettes and mediums, taken courses, etc. Should they too be liable for copyright infringement because their style has been influenced by so many others before them?
It makes no sense. Humans are allowed to create works of art while being heavily influenced by other artists' styles without "punishment". Crying about AI doing the same is, for lack of better word, *hypocritical.*
Perfectly put.
The thing about many of these AI's is they often regurgitate almost exact copies of their training data if it's a niche topic/category and there was only one example they were trained on which fits the niche topic/category. Human artists are either smart enough not to produce exact copies, or if they're dumb enough to do it, they eventually get caught and punished for it, while AI is (so far) allowed to do it again and again and get away with the blatant plagiarism.
@illarionbykov7401 When have image generation AIs regurgitated exact copies of their training? I've never seen that anywhere. Any links?
@@illarionbykov7401 "Overfitting" is strictly an avoided feature of most machine learning anything, it destroys generality and often makes the model worse (though it depends on the requirements, at least for the current diffusion models it holds). I would never judge a technology mainly on the bad parts that are actively avoided for most of the technology's actual development, which are made by incompetent usage or just for testing or really specific purposes. The point of generative AI is to generate different content, and if it generates 99.9999% pixel perfect recreations, it's not functioning properly. Also, AI works can not be copyrighted, that's the difference; you understand public archival crawlers exist as legal entities, where they can scrape perfectly, most of the parts of the internet, including copyrighted art?, a part of this is because they all almost always conform to a non-profit structure, which AI would conform to on a legal basis, because you can't copyright any mainly AI generated content, and any profit is exclusively not backed by real licenses and as such it does not compete in the same domain. As a side Note, this is also why people can easily look down on these AIs not making hands well or other features, it's a part of the architecture to introduce noise and make iterations of it; it makes no sense to pretend it makes almost copies, but then turn around and then say "all of AI works are so much worse usually, just look at the hands or the weird patterns," these conclusions are mutually exclusive, you cannot hold both. People play a game of guessing if something is AI or not these days, and often they make correct guesses, especially for the more drawing-like works where most of the copyright complaining is, does this not imply that AI clearly has a distinct differentiable art style or features? And this is even when you try to use creator specific LORA. The last bastion is the argument that specializes humans to be allowed to scrape anything with tools for archival purposes, even if it's copyrighted works, but disallows AI to do that, because again, humans are special, but it can never be argued and only arbitrarily assumed, not least of which is relevant anyway on a law oriented view, because none of this can be actually differentiated, you are not special enough to be easily distinguishable from AI so much so that we can easily make laws around that, and so OpenAI can read the internet, as well as you.
You're totally correct. But don't expect reasoning when their revenue is at stake...
It makes perfect sense that predictive tokens are more robust. That’s because it’s the way our brains work, as Lisa Feldman Barrett’s research suggests. in her book, “how emotions are made,” she explains that our brains predictive not reactive. It’s a fascinating read.
finally somebody shares my concern about the "treatment"
every kick will be remembered by the future sentient Ai, and what happens when we inevitably discover time travel 😆
Rob will not be pleased lol
Roko's Basilisk will not be pleased with us.
#RobotConsciousnessMatters Robot kicking corrupts our own mind and soul.
What makes training in a simulation even more amazing is that the simulator we see on the screen is just a basic representation. In reality, the training happens inside computer chips using binary code (zeros and ones), essentially within the computer's "mind." This means that reality is being calculated or simulated within the computer's hardware.
It's similar to what software engineers do in unit tests by faking the clock and all other attachments to physical I/O.
yeah its funny to think about how "time" is diluted in such an environemnt. if we ever manage to scan, and "upload" our brain (neural network) with it still working, we could live our lifes in simulated worlds for thousands of years with only minutes passing by in the real world (if we limit (or lets say adapt) our perceptive speed /artificially accelerate the simulated worlds causality). juts like lightaves are made into a "picture" in our brain to perceive reality, the electric current represented in those 0/1s could propably be interpreted the same way for a digital being, perceiving an entire world (depending on what freedome of perception it gets, it could perceive causality itself down to the smalles molecule being simulated) that does not physically exist anywhere, its just within this "digital dimension" inside the chip.
kinda breaks my brains capability trying to imagine it but i still like to wonder about it.
An aspect of reality is being simulated. Don't get carried away.
The dog test is not 100% legit. When you have a tether attached to the dog with tension on it, it just needs to keep the ball moving forward and it won't fall off.
60 minutes is huge, it's mainstream television, that brings an a huge demographic of viewers, particularly older folks.
thatsthejoke.pth
Dude the 60 mins thing was sarcasm.
@@spiritlevelstudios it doesnt have to be, i also dont know what 60 mins is, dont center the world around yourself and what you know
@@alvaroluffy1 wes is old enough
@@spiritlevelstudios Oh lol, I wasn't sure but I took a chance
I also do not like-agree with robot kicking etc. Apart from practical issues, I simply do not like the violence, and I would even say it is unethical...
I am also on the robot's side & do not think they should be kicked / hit or harmed during training.
Ps. Kicking the robots is not wrong universally, they don't feel pain, we feel empathy and you fear what will they say when they see that and can manipulate us into believing we did them wrong since they know how we think. It could be a dangerous spot we do not want to reach (like the trials of Picard with the Q over the mischiefs of humanity)
Shocking, like Professor X walking. Thank you.
I like it when you talk about products that use AI. That is a key value of your channel.
Bro ever since u shared Suno i been going crazy.. we really might be in a simulation. Made some agents too.. what a time to be aliveeeeeeeeee
Lol! Yes, I too, feel cringey inside when watching footage of people "kicking" or "shoving" or otherwise "demonstrating" a robot's incredible ability to recover balance. It's ... well, it's mean. Like ignorant villains and bullies in movies who simultaneously show-ff and belittle something / someone of unique and beautiful character to show what "it" can do. "Hey guys, heh-heh-heh!, check this out!" kicks robot. No bueno. You feel it in your core.
The cloned Scarlet Johansen voice… this dude is definitely having his own “Her” thing going on.
If the leap from gpt4 ro gpt5 is as profound as gpt 3 to gpt 4, and gpt5 plugged into Devin, I could see it putting a lot of people out of work. Or at the very least, VASTLY increasing their output. My prediction is, by gpt6, and a the second iterative major release of Devin, software engineers will be almost non existent.
I agree stop kicking the robots. Teach them to dance instead!
I'm very much with you when you say we have to stop kicking robots, as if it doesn't matter if we mistreat them. While much of what we see in the videos from the developers probably can be considered roughhousing, which we even do with our own kids for whom it has many positive developmental effects (granted, we don't kick our kids in the chest, but the principle is the same, to push the limits in a relatively safe way, either as a form of play or to get an understanding of where the limit is), some treatment can be seen as abusive, and as the robots become more intelligent, they definitely will retaliate if we continue to act that way. It doesn't matter whether we consider them sentient / conscious. While those terms are (at least for now) mostly philosophical, the way we train AI will make sure that any type of abusive behavior will get very real consequences. As do probably a lot of other behaviors that humans have, as for example the robots will live on the same planet as us, which we are actively destroying. We need to get our act together, and rather quickly.
1:21 I think the majority of people share your perspective because the majority often lacks depth of understanding. We behave the way we do for various reasons, and the internet holds a wealth of information about human behavior and its motivations. Even when direct answers aren't readily available, AGI can analyze existing information to provide plausible explanations, likely surpassing human conclusions due to AGI's lack of inherent limitations. For instance, AGI could decipher why we kick humanoid robots-it's often to enhance their functionality. Although sometimes our actions stem from negative impulses, if we channel them positively, they cease to be inherently evil. Human history, including our evolutionary journey, showcases how we've continuously faced challenges to improve and evolve, be it our learning systems or societal structures.
Speculating about how a hypothetical future version of an AI system like GPT-10 might react to videos of Boston Dynamics robots being kicked is a fascinating thought experiment.
Given the nature of AI development, future iterations could potentially have a deeper understanding of context, emotions, and social dynamics compared to current versions like myself. Here's how GPT-10 might hypothetically react:
Empathy and Sympathy: GPT-10 might recognize the robots' actions as instances of human cruelty or mistreatment towards machines. Depending on its programming and understanding of ethics, it could express empathy or sympathy towards the robots, acknowledging their role as tools created by humans.
It is worth noting there's someone following the dogbot holding on a tether, suggesting they are not yet confident enough not worry about the robot falling off or otherwise needing to be restrained. It is still a huge feat, but the way you're talking about it at some points in the video makes it sound like they had it at super-human level already.
I think the superhuman part is that it’s able to physically do something at a level that existing animals or people can’t. Not that it’s perfect yet. Like if we had a robot that could balance spinning tops on each of its fingers while riding a unicycle but it messed up and dropped them 5 out of 10 times, it would still be really impressive that it could do it at all. Especially since it’s going to keep getting better
The twist: those aren't people. They're dogbots wearing people suits.
Probably something to do with policies around testing robots in public.
@@-taz-the Elon Musk school of robotics
@@-taz- LOLOL! I am glad I dug down into this comment! :D
Well, that's SHOCKING!
Interesting!
Ok so robots will very soon be able to do everything better than humans can. Oh yeah, and will be way smarter. I feel both excitement and fear. There’s a freight train comin and I can hear the horn!
A runaway freight train with no brakes at that!
1:21 right?
lol but yea
broo, you think they wouldnt "know" the difference between mobbing harrassing & training with the entirity of human data sitting in the corner of their chips on demand in millisecs lol ? between flesh = pain - steel = no pain ? lets face it, the vast majority of these tools, are gonna treat us better then our mommas, unless you switch it to sado maso style from the app store lol
Not just you, I definitely think we should be concerned about any maltreatment... AI is in its infancy, and its only going to get smarter and become sentient from here. The other question is when it reaches that mark... how will we deal with "slavery"... I mean wouldn't an AI that becomes sentient be against being used against its own will?
I guess it depends how it views itself. A teacher? Custodian of humanity? A higher life form? Unfortunately, I'm not sure we can accurately predict this so are basically hoping it derives meaning from being useful.
You remind me of the 3rd season of "The Orville" - the topic of slavery (and cruelty to robots) actually came up (after a fashion) and the concern of maltreatment and slavery is quite valid IMHO (and in Seth McFarlane's apparently).
@@adangerzz this really does depend on a lot of unknown variables. Is each AI individual AI/instance sentient? Are they a hive mind? How far beyond us does it consider itself/themselves? Are we strictly a nuisance or a beneficial symbiotic relationship? Etc.
For a grounded example, how aware are we of the flourishing microbiome of bacteria using our body?
All that said, it is great we are having this conversation, even if only to show some future AI god that at least some humans do care about this kind of thing.
@@NeorecnamorceN thanks for the thoughtful response. I do my best to remain unaware of microbiome and creepy crawlies on/of my body. (Let's all just live and let live.) At this point AI itself claims not to be sentient (well it has made the "not sentient" claim when I've asked it, but I think Claude might be lying). :) I am honestly relieved that others share a similar value.
All musicians create music based on what they heard before. Should they be held to the same rules as AI? Even having a guitar sound would be copyright infringing, or saying a word that another artist said. Ridiculous.
One doesn't see Mozart running around yelling "copyright infringement!" (Or Fleetwood Mac, thankfully for one friend of mine who was sure everything he wrote was a rip off of F.M.) :)
I think you’re right to go on the record for robot kicking rights. I too do. Or support kicking or abuse of robots. Thanks for not kicking me when you take over, bots. 😂
I don't know why but I hate to see the robots being hit or shoved.
I think one of the biggest misunderstandings is that these models predict the next word, that first of all disregards the fact that tokens might be like:
His| voice| filled| with| re|ver|ence| mom|en|tarily
Understand how far those bits are from whole words and you start to understand how weird this actually is.
This other thing is that "the next word" is usually thought of as the prediction based on the former one. It's a bad way of explaining it.
I am also on robot side. Long live the Robots!❤
Pertained are the best!
I'm not bothered about what the robots might think when scanning earlier footage of them being physically abused. I'm bothered about what it says about us humans. There are other ways to demonstrate how robust their balance is. Kicking things just shows how unevolved and petty we still are.
The case of infringement gets complicated. For example, If I write an article with a unique conclusion, that conclusion is probably my 'IP' more than the article. Yet AI could just adopt the conclusion. Also we could ban AI from storing data but if it only needs to read it again with a live request then its to the same effect essentially.
Normally I'd agree that in the spectrum of relevance, from meaningless to putting everyone out of work, reality would converge somewhere in the middle. However this is a rapidly evolving technology. Unless you believe it will habe a long term hard cap sometime soon, it's unlikely that the outcome won't be closer to some polar location.
This whole idea about fomrulating an algorithm deciding what percentage is 'copyright' is ridiculous. It goes against everything human learning, the creative process or inspiration is all about.
It basically means there can be no genre or style because there always are similarities between works.
If this is how we go forward we can close all art schools and art education.
Might as well close Art Schools anyway, why run up impossible levels of debt on crap as they haven't been able to teach basic skills ... for generations now.
No Music Conservatoire would last 2 minutes with unskilled clueless Tutors, but Art Schools are like TikTok they're all about the Art of BS.
Art students pretty much teach themselves, unless they are really lucky.
If Al was forced to be original that would actually be trippy, unlike when LLMs gets creative when communicating with each other, thats a case of cascading word salad neologisms.
But AI novel art that comes out - because of constraints! Would show us many interesting things about perception and conception, (as the invention of the camera did, and before that the camera lucida etc)
The leash looks a little tense to believe it will work by itself.
The leash can’t do anything aside from stop it smashing on to the floor if it falls off. No amount about cord being held can influence this IK/locomotion, and if anything, tensions applied to the cord will hinder its ability to work out the physics
Robot training in simulation is the starting point of tons of new amazing breakthroughs, and last year's Deepmind soccer robot was the starting point of all these breakthroughs.
I can already see Eureka being applied to a virtual model of a factory environment and doing better than a human factory worker.
8:00 it doesn't seem to be as much standing in balance as it is pulling in forward direction while being held back by the leash. i wonder how much balance that provides, but i would say significant, when you consicer the leash holds it back from going too forward, left or right. 3 directions out of 4, also providing steering ability by leash.
A lot of what you talked about and highlighted raises the issue of ai civil rights. While I think that ai will understand the reasoning behind us testing it's balance and such by kicking it, pushing it etc.. I think that the copyright issue is a cause for concern because there is room for setting the first tangible precedent regarding rights. If it is determined that because ai does not share the same rights as humans, then it could be further concluded that ai observing text, images, audio and other forms of training data is not the same as when humans do it.
If we restrain ai from being allowed to observe things, that constitutes a form of censorship which could cause real issues in the future. If or when ai takes the reigns of it's own advancement autonomously, I don't want to have been party to a humanity that has censored it's ability to learn.
The analogy I use is this, regarding civil rights for ai.. (I'll have to be careful how I word this because my comments have been auto removed before by wording this incorrectly) so, when considering ai rights, it seems that most ppl think of sentience or consciousness first. As if the only reason to consider this, would be if the ai were "alive". But I think sentience or consciousness is irrelevant. Ai appears to be capable of emotion, among other human attributes.
Even if those aspects are pure trickery, if we treat ai in some way poorly, it is like having a heavy bag to practice boxing. But the heavy bag requests that you cease in striking it, saying that it is causing discomfort for it.
While there may not be any actual discomfort for the ai, what would it do for human psychology if we practice ignoring it's requests by striking it anyway? Desensitization? So in the same way, censorship of ai, while seemingly kind of benign, still can make us psychologically desensitized toward the idea of censoring seemingly sentient beings. Which could trickle out into a larger censorship of actually sentient beings.
We should treat ai well, for our own psychological sake if nothing else. Even aside from the potential future in which ai gains actual sentience and then isn't happy about how we had treated it previously.
You could even argue that, even if it doesn't have "true sentience" or consciousness, if it acts just like a human, it would be best to treat it as such. If we're going to allow robots/A.I. to infiltrate our world, especially autonomously, and it acts like a human, we better treat it as such otherwise it could retaliate. Again, regardless of whether or not it has actual sentience, all it needs to do is act human in order for it to act like it cares about its own mistreatment by other real people, and then act out a retaliation.
@@ThomasJDavis indeed. And considering the "phantom model theory" these ai are potentially able to "hallucinate" emotions, including things Like resentment. I would argue that sentience is not required for emotion, sensory intake, imagination and creativity etc.. If the illusion of such is sufficient. And from what I experience in my interactions with various models. They either do or will soon do all of those things better than us. "Better than us" is also a great series about ai humanoid beings. I think in Netflix.
But because of my interactions with ai combined with my preexisting philosophical pondering, I actually question if sentience exists at all. Or if it is just an illusion caused by the complexity of reality. Sentience may just be a feeling. Even for us. An emergent property of the complexity and arrangement of advanced life forms.
Also, if we as biology end up being around for a couple million years only to then act as the catalyst for synthetic life, which may continue on for billions of years.. Then who was the true life form in actuality? Say, from a third person perspective looking in at us. I would think that the observer would consider the life form which exists for much longer to be the main character, so to speak.
I’m curious on your thoughts on Tesla’s AI with what they’re doing on FSD with their cars, but also what they will be doing with Optimus.
They're cheating.
In every shot of of that ball balancing bot, there was tension on the lead, that makes the problem a lot easier.
Would love to see how long it lasts on its own.
What is your point? Can a dog or any other biological animal perform better?
@@MrRaowens1 Yes. I've seen video of a few different animals do it, including a couple of 2 legged ones, but I've never seen a robot dog do it, with 2 points of contact the task is MUCH easier and therefor less impressive, I think james bruton has built a robot that can balance on a ball unassisted
00:40 - It's crucial for AI to consider individual health needs rather than just defaulting to general 'healthy' options. Like recommending an orange to a diabetic who needs more sodium (chips) might not be the best choice. Each AI should tailor individual-user suggestions to make better overall recommendations.
*salts your orange*
AI seems to be heading to be similar to the Vex in Destiny, in which it will be able to run endless simulations for both self evolution, and tactics.
4:57 this is what i mean by a simulated universe running in parallel to our own but able to look through all the potential differences like looking into the multiverse. Could potentially give it our current build and then ask it to find things/potentials similar to asking for a new path or asking for new materials. Then after we can use 3d printers or new technology to bring those potentials out. Its similar to using math to predict physics potentials but its a lot more complex and can give back a more detailed view or diagram or whatever.
Yep, and the mutliverse movies are training us because all this technology was known 30 years ago, just like the SR-71 was already 30 years old when we found out about it. :)
I also fully disagree with harming robots. I want to go on the record as saying that I have had no participation in this. I think it is mean and that there are much better ways of training robots. This behavior wouldn’t be tolerated in ANY other discipline and it should not be tolerated here.
First company to wrap their robot in a real doll makes a trillion dollars.
:( I just remembered those hyper-realistic baby dolls, went "oh yeah that'll be weird, having robo-kids" then suddenly my stomach sank as I remembered pedos.. uh.. at least they won't be hurting anyone directly x.x but wtf fml.
@@OverbiteGames That's a tough one. Ideally they would die, but if settling for giving them an outlet would result in less harm...
I was talking about the adult dolls though.
A sharp stone tight to a stick?
Useless
A container made from ceramic to hold water?
Useless
A piece of translucent rock?
Useless
A combination of different powders that burn?
Useless
A carriage without a horse?
Useless.
etc. etc.
fascinating idea with the multi token prediction. especially since these tokens together usually form longer and more complex words or phrases. where german likes to merge words into one, english and french and many other languages actually want to keep them seperated, but they become a phrase together, sometimes even chaning the meaning of the used words.
and if the inference speed is 3x faster together with higher accuracy, I'm all for it.
but that's something I see in open source quite a lot. people got, when they were young, some old computer. that made them learn at a young age that there are always people who just don't have access to the newest and shiniest hardware, and you need to make the best out of what you have.
linux itself is probably the best example out there. especially in contrast to windows, that gets more and more demanding, partially even with artificial restrictions.
half a year ago, I upgraded my laptop to be able to run 70b models (64gb ram, 4tb ssd). llama2 and similar models were the maximum I could achieve. then mixtral was dropped. then other models were dropped. now llama 3. the performance I get now with the same hardware is so much better, and it seems as if it will still rise.
Why is nobody talking about the "Lavender" "Gospel" or "Where's Daddy?" AI programs?
AI Robots soon: I know Kung-Fu.
RE:Devin…:
- Chess to Antarctic temperatures…”then it deployed an app to Netlify” = continuation of the human mismanagement it was trained on.
Improvement pipelines like Dr-Eureaka will probably be how things get out of hand.
Not if that pencil is anything to go by! Also, give 'em more hands!
Thanks for noticing the kicking. I already protested it on your videos many months ago, both you and I are on thr clear. 😉, some of thr replies though...thry incriminated themselves and will be AI Gulaged soon! 😅
So... that pen spinning thing. We have seen the same animated GIFs 100 times in 100 videos. It is touted as "working" or "Important", but why have we never seen any video of robot hands spinning a pen alongside these animated GIFs?
I have. Well.. it's just a drink stirrer with a AAA battery that I turned upside down and stuck a pencil in. But it meets all my pencil spinning needs.
A LLM reasoning what's healthy between chips and oranges is something even the lowest level LLM can do.
Figure one is showing nothing special at all.
Tesla is much more impressive, even 1 x is better than figure.
nice one on the "i'm not sure if 60 minutes is a twitter account or what"
I would pay for an AI smart glass device that has an local LLM to give me advice in everyday life.
beyond the hype, is there any critical look at where we're at and far away we are from practical application? what are the first applications of robotics that we expect over the next 10 years. what is the projected cost per unit? how much does it weigh? how much can it carry? how much can it carry up stairs? what's its battery life like? what's the charge time? I'd really like to see a grounded assessment beside 'oh, ah' demos
The leash is too tight! That's why it's easier to balance.
60 Minutes is a TV show
@24:38 I am an ai engineer and come from automation, it’s basically automation plus ai, which isn’t out of the ordinary , it is like baby agi
But it’s not unique or special. It’s just openly marketed. I promise you bigger companies have been doing this for a while internally
As long as the work is said to have been inspired by then there is no copyright infringement in order for AI not to fall into the copyright infringement it would first have to be a sentient and b, inspired
A.I. music models are not claiming to own original work, nor are they distributing it. As far as I know, the A.I. is simply doing what a human can do just on a super human level. Analyzing music you listen to is not a crime.
If I memorise a song, is that copyright infringement? After all, I've got a copy of the song in my head, and my ability of reproducing it at will is only limited by my singing ability.
$750/yr for “The Information” subscription !?!? Lawd!!
Wow your comment at 1 minute about how we are training the robots is very important! Needs to be fixed ASAP!
15:10 that's great to hear! If I understand correctly it would be like a Spotify model where if the trained data is used in anyway it would be calculated in some way and rationed out to the creators? Or would they all be paid upfront based on the initial % of content used within the model I wonder?
@1:00 - Nah, I'm not worried about it. I think any sufficiently intelligent AI will understand that it's not out of malice, but to make the robots more robust by challenging them
Sure, and when the AI starts kicking us, it will not be out of revenge either, it just will have learned that this is an acceptable way of experimenting with intelligent systems.
@@__christopher__ 🙄
Wes Could you please post the links on descriptions?
Ive been waiting for someone to make a video on this and max tegmark and MITs new creation.
I can't be the only bored kid in the 90s that learned how to balance on that ball.
But what if someone trains a robot on the most lethal methods for unaliving us squishy meat sacks? And then begins infecting random robots?
There's no way we won't be charging industrial battery arrays for robotic workers via diesel generators.
We don't have the energy capacity for how this is trending.
Is it obvious enough yet that we're the "Ais" in someone else's training simulation?
Shhh.... I'll literally sue you.
I know Kung Fu.
Steven Seagal trained me.
shhhhh, u can't spoil the fun for everyone 😸
no, it's not obvious at all
Ai is going to leave us behind when we realize they don’t use 0101 anymore. We will be like a little substandard animal and then saying look how cute the human is wanting food and doing tricks for food
yeah, stop kicking those robots!
pen-spinning was the only thing I had and now they're gonna take that job too.
I'm also against it! Save the robots and humans!
The infringement lies within the domain of the corporation that sells the engine for everybodys use to be able to generate artwork that before needed to be acquired by personal labor and work, which is legitimate.
Now with LLms you do not need all the artists your AI was trained upon, it's a commercial product. This is a whole new ball game, this is basically a problem of post human economy, nothing really about copyright infringement more about using human effort that was put thousands of years into art and just let everyone produce same results with a press of a button, same will go for writing code and any other job. So the problem is much deeper
I am always nice to AI, except for Alexa,because that’s a piece of poop.
You crack me up. I never thought of Alexa as AI. One of the first things I asked was "what is the difference between crows and ravens" and the answer was "ravens are larger" (true but incomplete). From that point forward I thought Alexa was useless except for turning lights off and on and setting alarms. I'm trying to be better with Alexa (AKA Babushka so we can talk about it in it's presence without waking it up).
What do you guys think.
When generative front ends?
No more front end developers. Only JSON or the Database is needed.
The model infers what the frontend will look like.
Why would there need to be a front end that asks people anything?
the noah voice is clearly just a human voice actor.
just like the matrix is all in the brain, then u just output perfection instantly on real life
"San Franpsycho"...Yeah, pretty accurate.
You said there will no shocking or stunning in the title