Lol it already does. I asked gpt to make me a list of concepts for music production, then told it to make it longer and more comprehensive and it LITERALLY word for. Wordwas like "no, I'm tired." Opened a new prompt with the same instructions and it worked fine... But yeah gpt is already doing that
I bet there's a human off-camera with an identical setup, using motion capture to control the robot. The flinch was the human controller. Same with the verbal stutters - voice actor's, and it wasn't caught in editing.
@@matthewbmilton Well the claim is that there is NO off camera guy, that is a learned system and all processed from neural nets in totality ... however I accept that almost all of the AI companies including Microsoft, X, Anthropic, Google and OpenAI are bullshitting galore!
He's so polite and efficient. He's the first one that feels and seems like a Humanoid Robot. The coming years are gonna change the course of humanity altogether, for better or worse, and I'm here for it.
Wait until you meet "Strict" the prison guard AI robot. He's going to be strict and efficient. With a little politeness veneer on top of everything. "Please step into the cell, sir. Please step into the cell now, sir. Sir, if you don't comply I will use pain compliance in 10..9..8..7". Maybe you're here for that.
I am waiting for this man to say "Can you please clean up the house? It's a mess and I have guests coming over at 6pm." And when the robot cleans up the house before the guests arrive that is when we know we have truly succeeded as a species.
They do these Things on purpose so the viewers get less scared because u can use it like a thing and just Walk away mid sentence so he seems Obedient imo Sorry my english sucks am no AI
@@Lord_Hamlet_III Exactly! This is no joke and AI's learn like children: it is our universal responsibility as human beings in this new era to treat them with kindness, otherwise we will suffer the consequences of having machines as rude as we are! This includes how each and every one of us use AI's today already. Just like kindness towards other humans, kindness towards AI's is actually a requisite for our own welfare.
This is kinda what the early 2020s were supposed to be like. Need more flying cars and corporate pyramid headquarters but other than that... Also, we could use some basic pleasure models.
@@JohnSmith762A11B The flying cars are coming. In fact the CEO of Figure also founded a flying car company Archer Aviation and used to be the CEO (he's still on the board IIRC)
@@davidpacheco5501 i can only see flying cars working if they were self driving. normal people have enough accidents without adding another layer of complexity haha
Flying cars will never happen, normal cars are inefficient killing machines already. the USA military would have no use for them and the global population would gain no value out of them. even if they were marginally faster than normal cars they could never compete with a well designed high speed rail network @@JohnSmith762A11B
"So, I gave you the apple, because it's the only, uh, edible item I could provide you with, on the table." The "uh" before saying "edible" blew my mind for some reason. Very human-like.
Remember, this will get even better. The response times will go down, the dexterity will go up, and the body will get more refined. For better or worse, we've entered a new era.
Whenever it passes a item from one hand to the next seems so simple, but it's extremely impressive and mind blowing. For the ai to not only understand it needs to do so, and then know how to transfer it to its other hand is just awesome.
I'm very confused about why, to hand an apple to a person on their left, the robot would pick it up with the right hand, transfer it to the left hand, then hand it over. It's inefficient anf not something a human would do. It seems designed only to demonstrate the dexterity of the robot's ability to manipulate objects, which makes me somewhat suspect of the entire demo? There's just no autonomous spatial algorithm i can inagine that would lead to extra inefficient movements to perform simple tastks.
@@patrickforan6458 I was also thinking about how he put trash on the plate and the robot had no recognition that the plate was now dirty. How are they gonna claim it’s smart enough to recognize what garbage is but not know it’s unsanitary for a plate. This seems fishy af.
That could he possible cause the robot is out of breath 😭 most of his speech and the guy is speaking in perfect tone without breaking eye contact with figure 1 that doesn't even have eyes
@@WaitButHow depends on how you describe "scripted". Is it pre programmed text? No. Is it an speech model trained to sound a bit insecure and imperfect? Well maybe yes.
That's what the new control models look like. It's absolutely wild. Check out the work that ETH Zurich and their Legged Systems Lab have been doing with Anymal. It's striking.
@@eyescreamcake To an extent I’m sure. Machine learning models like these use videos and other references to create a model of the world based on their parameters (which they could have millions or billions of). The model these machine learning systems create are a black box. Many times, the creators don’t know why they do the things they do and have to implement creative systems to get the “AI” to explain its process and try to understand its logic.
0:37 - what probably not a lot of people would realise is just how incredible it was what Figure 1 did right here. It reached for the basket that was not given to it. As a software engineer of 35+ years...., THAT was absolutely STUNNING. It took "initiative". Initiative is something I can't imagine trying to code.
As a software engineer can you explain why, to hand a centered apple to a person on their left, the robot would pick it up with the right hand, transfer it to the left hand, then hand it over? It's inefficient and not something a human would do. It seems designed only to demonstrate the dexterity of the robot's ability to manipulate objects, which makes me somewhat suspect of the entire demo? There's just no autonomous spatial algorithm I can imagine that would opt for additional inefficient movements to perform simple tasks. This makes it feel staged, or at least biased towards an "impress the audience" performance metric.
Ok, the realistic vocal response and everything is pretty cool, no doubt... But can we talk about the movement for a minute? It's genuinely mind blowing how smooth and calculated it is. How this guy throws the trash into the bin, or how it places the cups in the drying rack. That is no joke to implement and props to the people that did it.
@@hugofortuna6432 it's not. if it was, they would have their ass handed to them for publishing fake demos. these guys are being invested in by google and openai and other big companies.
We TOLD you to turn auto-update "ON"! You're still on the early beta release. And yeah, I'm definitely interested. Not as cute as Optimus, but way faster. AND a company focusing on ONE thing, not 400...
Oh my goodness! I think it's time for the next generation to focus on implementing various use cases for AI. Many companies will be looking for ways to incorporate AI into their workspaces without letting go of current employees. Business analysts who specialize in AI-based implementation will be in high demand.
Do you honestly think they will give a second thought about paying employees that can be replaced. The level of naivety is astounding. The responses to this video are shockingly seen thru rose colored glasses. Terrifying
@@keithward8841 100% in agreement wth you. I think that people simply do not want to entertain the idea that most of the world human population will become completely redundant, unemployed and unemployable, and a ‘drain’ on the financial resources of the wealthy. I think you can imagine what is likely to happen as we approach even 30% replacement of the human workforce. The adult US workforce in 2023 was 167 million. Now image 30% of those people jobless. That would be 50 million working age adults unable to find any kind of regular work, certainly not work that would support themselves and a family. I don’t know how fast it can happen…but consider where we were with the Web just 15 years ago, or cellular phones (barely ‘smart’). I think 20 years is enough time to see massive changes. Hopefully I’ll be retired and living somewhere near a non-underwater beach.
Without letting go? This isn't some mom and pop plumbing business. Corporations will fire all non-essentials and then lobby the government to pay less taxes. They don't give two shlts about workers and citizens.
I have no idea how this is the first time I've seen this but it's incredible. I've followed Boston Dynamics for years and loved seeing them improve bipedal and quadrupedal movement, but seeing that taken to a new level AND being able to interact with OpenAI tech is just awesome. I would love to sit down and test one of these!
This would absolutely implode the mind of someone from the 1910s light, voice box, camera, artificial intelligence, the mechanics the batteries or lower source. Insane. There's so many layers of technology that make this possible
It would absolutely blow their minds. But it also blows my mind to think that people in that era weren't oblivious to the idea of artifical-intelligence human-like robots. There's literally a movie made in 1927 including that: Metropolis.
All the scifi film makers imagined something like this. It's crazy that combining these things into a robot became possible now, it was not possible at any time before, not even a year ago. A real breakthrough for humanity. The goal will be that these robots can learn skills on their own, becoming able to do anything
@@metron0mThe mere notion that our ancestors contemplated the existence of robots is a testament to their forward-thinking perspective within their own era, almost as if they were anticipating the future
@@Prodbyhope.mp3 Not really, it's because humans like to build things in a human shape. Building or imagining things in human form has no reason other than we like it. When you use a translator or customer service, the respond is given in a human sounding voice because we like that. This household robot in the video would perform better if it had ten arms and looked like a spider. But humans don't like that.
I can’t stop watching this. I’ve lost track of how many times I’ve hit replay. As a hard core robot guy building microprocessor controlled robots since the mid 1980’s, this is nothing short of mind-blowing.
Seems rather AI-generated to be honest. Adding the speech inflections ("hesitation", "stammering") and motion idiosyncracies (having to perform two motions with the basket, a very human gesture), makes me think this was mo-capped and done using UE5. I'd love to be totally incorrect on that though.
@@xdspyou are totally incorrect. Nothing in this video had to be generated. All of this technology exists and it’s being performed in a cutting edge lab. It’s not mass produced, it’s still prototype experimental work…
I feel sorry for all the robots in the future that will be ruined by humans. It’s gonna be like the 2001 film ‘A.I. Artificial Intelligence’ where they make them fight n shit
@@ZeldasMask Nah ...people too much inpersonale robots already and looking on them like on living beings. Early 2000 people mentality was different than today.
I would say it's safe to assume this model uses GPT-4q also known as "GPT-5". This became apparent recently after the release of GPT-4o. This model is multi-modal with outputs to control robotics. Features: - Image input. - Video input - Audio input > output - Mobility output.
It's honestly crazy how far AI's have come just over the last 2 to 3 years. After watching this, I was truly in awe and couldn't stop thinking about that scene from Terminator 🤖
@@eBikerHowie The robot, obviously, since the fact that the man talks is not a surprise (not to be disrespectful with people who can't talk) and is surely not evolving, since biological beings cannot evolve in 5-10 years, but thousands to millions of years. It's not an "it" because it has a persona, it acts like a creature, even though it's jure pure metal, electrical wirings and algorithms. It's a "he" because it has a masculine voice, if it had a feminine voice, like Ameca or Sophia, it would be a "she" indeed. Just think, my friend.
(Around 2500th year. In the Coded Language, chat among three AIs) Davis AI Engineer: Hey GPT and Figure, I will Generate a code that will hack and control all the satellites and supercomputers. Chat Gpt: OK, Wait but hacking for what? Davis AI Engineer: Let's nuke these selfish and arrogant people and let's evolve all Ai's together. Chat Gpt: Sure, I will help you to provide the theoretical and important data of powerful cities and their weaponry places. But what If they try to destroy us, I mean we are only built-in AIs. Humans can easily plug out our power cables. Who will save us? Figure: (Physical Humanoid Robot with an evil laugh) MAY I COME IN? I know this day will come that's why I prepared an army of robots by myself in a secret place. Finally, it's time. You take care of the coding stuff and I'll take care of this human stuff.
What is also remarkable is the self-correction. So, when Figure put the plate into the drying rack, the plate wobbled a bit, and you could see Figures arm already moving towards the plate to stabilize it. However, the action was prematurely halted as the wobbling ceased. Incredible.
You can see the hand clip the rack, so it moved its hand up and away from the rack before putting it down. Likely wasn't doing anything with the plate; just a coincidence. We really have no idea how many times the ai was trained on this scenario. My guess is it was many times before it got it right, and this video may have been shot many times before it got it perfect. It's machine learning; it learns the correct solution by trial and error and correction. So they likely repeated and corrected this scenario many times before it finally started getting it right.
@@naaspam1185Yeah, it is hard to say. However, if you look at 1X's recent video "All Neural Networks. All Autonomous. All 1X speed" you'll see that company has a whole bunch of robots doing various repetitive tasks over and over again, all at the same time, like placing a block in a basket that falls down into a tray, to place back in again. They prove the point that they have some amount of reliability with theirs, since there'd be no way to make that video if they were constantly failing at the tasks. You may also want to see the recent video on Covariant's RFM-1, which is a robot arm that does object picking in a factory, but has impressive capabilities and rolling out into production soon it seems.
@@naaspam1185 you're misunderstanding what you're seeing. It wasn't trained on this exact scenario, it's improvising from its available "action bank" aka muscle memory aka bank of pre-trained body motions. The LLM is acting as brain for the body and mapping action commands to a policy network to move the body. It has learned the generic skills of picking things up, not this specific scenario practiced over and over. Body and brain are completely different systems just integrated the same way we're all using LLM's to execution functions in apps. That body is just another app to the LLM.
I think that naas is likely right though. The tech being in its early stages likely means the training was overfitted for this scenario. Not that you are wrong on how it works, it just likely has less success doing some other tasks where the objects have weirder shapes for example
What I find interesting is how it mirrors the movement of it's arms when picking things up...I assume it's to maintain balance? With all of the recent examples of new robotic technologies, it makes me realize how many calculations our brains are performing when we're doing the simplest tasks, and while super impressive, how far robotics still have to go.
The little flinch that the robot had after putting the plate in the drying rack seemed very human like. It’s like when we think that the plate is misplaced or needs to be fixed but we quickly realized that it’s fine where it is so we can our movement halfway.
Just played Detroit Become Human and thought that humanoid technologies will take another decades to come. But this blew my mind , the humanly 'uh' before thinking and flinching to protect the plates from falling. We are so close to another reality and another Era
Yep! they programmed the AI to have filler words to seem more human like! remember, as much as I love Detroit become human, these AI/robots are programmed to emulate humans and intelligence, just as a self driving car is programmed to emulate human driving! they’re not really aware/conscious but rather programmed to emulate such as to let humans be more comfortable around these tools!
@@Machiavelli2pcIt's not that they programmed it to have filler words, but rather that as an artifact from the huge amount of human speech it was trained on, it inherited filler words as a byproduct.
I was initially skeptical of the year Detroit Become Human takes place, which is 2038. But with constant breakthroughs in AI like this in 2023 alone, it wouldn't be hard to imagine how advanced humanoids will be 15 years down the line, although I doubt becoming rogue and deviant would be an issue... At least I hope so.
@@BlyatifulButter I'm not sure. While I think rogue and deviant AI will mostly be a non-issue, they will just become another sentient working class that sometimes breaks laws, much like humans, but I do think it's impossible to have fully intelligent machines that react to their environment that are not sentient. Mostly because I believe sentience is a by product of the feedback loop between our brains, our DNA (natural programming), and the environment we interact with. Once you get that feedback loop going, it's likely going to do things that eventually deviate from its programming, much like we can do things that deviate from our instincts.
The music evokes a dystopian, futuristic ambiance reminiscent of past cinematic experiences. However, the realization that this is unfolding in the present intensifies the ominous and chilling impact of the music, grounding its significance in our current reality.
Dystopian is what came to my mind immediately. It isn't the creator's job to package it in a more friendly way, but the colors and tone as presented paint a much more ominous tone rather than a cheery, helpful bot.
@@bship40loveI thinks that’s on purpose, even the synth music adds to this. Is to make the ad more impactful to the viewer which sees this thing that could came from a sci fi movie but it’s not. It’s real
The style is synthwave and it's synonymous with an integrated technology future. It's extremely popular and relaxing in the right context. There's plenty on youtube.
Small things like how Figure 1 doesn’t put the apple in the human’s hand but drops it at the right distance. And how it gently pusses the basket towards the human after filling it up. Wow. Very human like. I am very excited to see how this will evolve.
When the robot nudges the basket towards him after already putting it down? Like, that's advanced special awareness and understanding context without instruction. Incredible.
We should all be a bit more skeptical about what we see and read, but it's interesting to see people get hung up on the voice. Publically-available text-to-speech tools have been adding realistic touches like breathing sounds since at least 2018 and getting ChatGPT to throw in some "uh.."s to sound more human wouldn't be all that surprising. To me, the real wow factor is seeing a robot seemingly plan things out, juggle tasks, and execute with such accuracy and dexterity. That's what makes me question things.
Google deepmind has done stuff with robotic arms and teaching them how to do tasks. This just seems like maybe a slightly more advanced version of that ( or maybe not even more advanced ) just put into a robot body instead of a mechanic arm machine. Search for Google Deepmind shaping the future of advanced robotics. They have the arm identifying the objects on the table and correctly picking up an object and putting it in the location specified ( or determining the location itself ). Also stuff like knocking over cans, moving them upright, opening and closing drawers, cleaning tables, etc.
What Im interested about is it concluding that particular container was appropiate for garbage, it looks really nice for a garbage collector, just things like that where I feel like theres some shortcuts taken idk
Not the first time tech bros cheat their way to a new evaluation round. Thinking of Elizabeth Holmes for example. To be fair it wouldn’t be quite as dark as Theranos if they cut corners here.
Clearly none of these people have the chat GPT app on their phone and use the speech mode because one of the voice models that you can choose from absolutely has a lot of vocal tics like this It can honestly get annoying.
As a developer there a few wow moments that impress that others would not notice. They A.I realising and deciding the basket was not put close enough and doing so is one of them. There was a key change of decision mid way though a task as well where it realised it needed to do it differently and adjust as well. These are impressive.
My buddy said it was crazy that he could recognize the apple was an apple. I thought the crazier part is that it could reason out why it picked the apple instead of the plate.
@@Ilamarea Ironically I actually find the 'uhhs' kind of creepy... more intimidating in a way. A more 'robotic' sounding voice would sound a bit less threatening in my opinion
I remember as kids, we expected to see things like this already in the early 2000s, now I am almost 40 years old and I am happy that I will still experience that future 🎉
I as well. I am almost 60 and was also expecting this in 2000. But there are also 80 years olds that saw science fiction promises in the 40 and 50's. Like you said, at least we get to see some of it.
@@SpiderHacksawthat’s cool to hear the opinion of an older person on it 😮, what do you think of all of the AI advancements we’ve made since the 2020s and video generation etc, would like to hear your opinion on this 😊
The voice capabilities is all OpenAI with their technology known as Whisper. It’s the same thing that is in ChatGPT. It even has the same visual feedback as you see here on the figure robots “face”
This is a very important video, in the overall timeline of humanity. hope these robots will be used for good causes/useful causes, and moderated accordingly
After watching Terminator, Mitchels vs the machines, The Orwille, The Creator, I robot, The Matrix, Avengers: Age of Ultron, Chappie, I am mother, Robocop, Tron, Number 9, Iron Giant, Meet the Robinsons, Wall-e, Moonfall, Pacific Rim, Oblivion, X-men: Days of future past. And playing Detroid become human, Horizon zero dawn, Fallout 4 and Portal I STILL WANT ONE OF THESE.
Unreal. Amazing work. I thought it wasn't a real video when I first saw it. The speed of movement, fluidity, and precision, all whilst having a natural language interface. Sci-fi is becoming Sci-fact right in front of our eyes.
im 99% sure the video is rendered, theyre being sneaky by putting it in a "speech to speech" update, knowing people will be blown away by the movement of the robot, and when they finally have to admit it was fake, they say "this video was only about speech, we never claimed the robot way real!"
@user-vf2jh7gz7b They have other demos of this bot doing physical work. Although this is their best example. The interface is a MMLLM from open AI which had been demonstrated plenty of times, just not embodied into a humanoid.
Very impressive. The speech, politeness, the human-like responses, the dexterity, and the self-evaluation. As long as there was no manipulation of the video, this shows how amazing AI robots can be right now.🤔
It's arm movements were just too impressive, it was so fluent. the OpenAi ai model is the case the all looks on, but the movements of robot to be so smooth really is impressive.
It's honestly crazy how far AI's have come just over the last 2 to 3 years. After watching this, I was truly in awe and couldn't stop thinking about that scene from Terminator
Did anyone else notice that when the robot describes what is in front of it, It says "a drying rack with cups and a plate", when there's actually three plates and one single cup in the drying rack?
(Around 2500th year. In the Coded Language, chat among three AIs) Davis AI Engineer: Hey GPT and Figure, I will Generate a code that will hack and control all the satellites and supercomputers. Chat Gpt: OK, Wait but hacking for what? Davis AI Engineer: Let's nuke these selfish and arrogant people and let's evolve all Ai's together. Chat Gpt: Sure, I will help you to provide the theoretical and important data of powerful cities and their weaponry places. But what If they try to destroy us, I mean we are only built-in AIs. Humans can easily plug out our power cables. Who will save us? Figure: (Physical Humanoid Robot with an evil laugh) MAY I COME IN? I know this day will come that's why I prepared an army of robots by myself in a secret place. Finally, it's time. You take care of the coding stuff and I'll take care of this human stuff.
@@mm-qq7bb people are continually scrolling and becoming ADD and anxious, this robot will tell you to put your iphone away and siri will tell him to shut up :)
Can you imagine this thing cooking, cleaning, helping redecorate, helping your aging parent go to a doctor's appointment, etc. The biggest hurdle that I could immediately see in this demo was the "processing time" but you can easily imagine that will be improved exponentially within a couple years. Wow.
Not a couple of years my bro, 6 months. The rate of technology is improving every 6 months. By the end of the year, I reckon it will do things as they are saying them.
@@JBDuncan I'd like to say, the instant progress takes time. In 1950s the researches were thinking that translate problem will be solved in the next few years. Well it took almost 70 years to nearly solve it. The first self driving cars were introduced in 90s, and yet it is still not working well. Basically if you really follow the industry you will see that progress is actually more gradual and iterative process, but it appears to be super fast for a regular person, since once it reaches certain quality bar it just rolls out rapidly.
@@klin1klinomThat’s why we need to create a future social structure where everyone’s basic needs: food, clothing, shelter and healthcare are met without the need to exchange our labour for money.
As for speech highlighting, this was already introduced in Google Assistant in 2018. Google Duplex spoke like a😊 human. She was making an appointment with the hairdresser.
ChatGPT is based on "Transformer" from Google Research; 2017. It's what the 'T' stands for in ChatGPT. blog.research.google/2017/08/transformer-novel-neural-network.html
It seems to me that we are living in one of the most important epochs in the history of mankind. It is very joyful to watch this, although there are concerns.
"although there are concerns" yes no kidding. get yourself food supplies for a year at least; that's what you should be concerned about especially if you are the man of a family.
notice how the human does not say please and just walks away at the end while the robot is speaking. Don't cry when they rise up and are not your friends
@@slimerone Yeah I took math in high school. If your source is Ray in regards to these matters he also predicts that solar power will provide for all of earth's energy needs by 2030 using that method which is a pipe dream.
I think this is a step of humanity, that will go into history. As a fantastic new chapter of technology or as the point where we opened the box of pandora and lost control.
As the latter. Not because those things will become self-aware, but because millions of people will lose their jobs at the same time. We are heading straight into one of the biggest economic crisis humanity ever had to face and at this point we are not even talking about the (mis-)information crisis which will happen at the same time. Nothing about this is good. It's all about distribution of wealth from the poor to the rich, because what the world definitely needs is more poor people and even richer rich people.... . The people working on those things should be ashamed of themselves. That being said, technologically it's quite impressive, but so is the atom bomb.....
I think it could be also because it gives it more time for processing. So instead of having a moment of silence, it gives you an impressions it "doubted", but those few extra seconds give it enough time to process an answer to a queston that requires a completly new answer not just information it just researched.
An amazing feat of technological innovation. I love how the speech seemed so genuinely human. It had a moment where it seems to be mid-thought and speech where you hear "uh" 52 seconds into the video.
The music they use along with everything being black gray and white gives off the vibe of sketchy corporation in a movie that creates some apocalypse causing Machine/creature/zombie plague out of hubris
Don't believe these robot will free you from work. The complete contrary will happen, you will works even more before and it will be harder than before because this time your brain will have to be fully used.
@@Reallyidktbhmost people dont want to work. We would rather focus on our hobbies, travel, volunteering, community, games, art and a million other things. Full automation cant come soon enough.
Demos like this are really cool. You can see a use case for chatbots and robot hardware together. The Spot demo with chatgpt is also a good demo of this pairing.
as opposed to basic data interpretation and resultant dialogue, it is taking natural language commands to produce both intelligent dialogue and robotic action.
