With Spatial Intelligence, AI Will Understand the Real World | Fei-Fei Li | TED

In the beginning of the universe, all was darkness - until the first organisms developed sight, which ushered in an explosion of life, learning and progress. AI pioneer Fei-Fei Li says a similar moment is about to happen for computers and robots. She shows how machines are gaining "spatial intelligence" - the ability to process visual data, make predictions and act upon those predictions - and shares how this could enable AI to interact with humans in the real world.
If you love watching TED Talks like this one, become a TED Member to support our mission of spreading ideas: ted.com/member...
Follow TED!
X: / tedtalks
Instagram: / ted
Facebook: / ted
LinkedIn: / ted-conferences
TikTok: / tedtoks
The TED Talks channel features talks, performances and original series from the world's leading thinkers and doers. Subscribe to our channel for videos on Technology, Entertainment and Design - plus science, business, global issues, the arts and more. Visit TED.com to get our entire library of TED Talks, transcripts, translations, personalized talk recommendations and more.
Watch more: go.ted.com/fei...
• With Spatial Intellige...
TED's videos may be used for non-commercial purposes under a Creative Commons License, Attribution-Non Commercial-No Derivatives (or the CC BY - NC - ND 4.0 International) and in accordance with our TED Talks Usage Policy: www.ted.com/ab.... For more information on using TED for commercial purposes (e.g. employee learning, in a film or online course), please submit a Media Request at media-requests...
#TED #TEDTalks #ai

[@MrSchweppes]

This woman was instrumental in making AI see. We wouldn’t have GPT-4 Omni, Google Ask Photos, Tesla FSD, Apple Face ID, computer vision in medical imaging, etc., without her work! Bravo 👏

[@DivinesLegacy]

Hope she sees this

[@UlyssesDrax]

​@@DivinesLegacyLol

[@deependranetwork]

It means she uses your image data stolen by Google , to make commercial viable project, without giving you a dime in royalty 😌😳

[@MrSchweppes]

@@deependranetwork It’s called fair use. Google it.

[@JohnNy-ni9np]

​@@deependranetwork, to be fair she used images from public domain ImageNet, for non-commercial uses. Not sure if she is affiliated with any commercial projects.

[@Clammer999]

She’s been called the godmother of AI and I couldn’t agree more. Without what she did with ImageNet, neural networks might not have seen the light of day. Despite her incredible intelligence and enormous contributions to the field of AI, she remains humble, optimistic and approachable. I highly recommend folks to read her recently published book called The Worlds I See, which traces her arduous journey as an immigrant to the US who speaks little to no English, climbing her way up to be one of the most respected professors and influential researchers in the world of AI.

[@ssuwandi3240]

Nobody cares truly. What she did during lockdown mattered the most. Come back with the least sappy children's books!

[@bustoms]

But lets also thank the millions of amazon turk workers that made that happen. She paid them all pennies to make this happen and nobody gives them any credit.

[@tedflint7237]

OMG She is creating human like robotics. Basically humans get paid robots don’t and robots cost a lot of money to make where humans are bread looks like we’re splitting hairs. Are they even sociable? They have no empathy and all this is happening by the planets burning.

[@tedflint7237]

OMG She is creating human like robotics. Basically humans get paid robots don’t and robots cost a lot of money to make where humans are bread looks like we’re splitting hairs. Are they even sociable? They have no empathy and all this is happening by the planets burning. Worse still they made be made by Boeing.

[@tedflint7237]

Makings of a better world. Joke right. Plato warned us in 500 BC about the shadows in the cave.

[@NostalgistGuy]

Lots of Respect for Fei Fei Li

[@josephmedina6403]

Yeah ! Stupid kitty always knocking stuff over .

[@parikannappan1580]

4:01 "My student Adrej Karpathy" wow what a statement.

[@zorawarsingh11]

Yeaaa iykyk

[@doctordilanka]

What a flex.

[@pratiknaikedu]

Wow! Absolute boss

[@JoshuaVarghese]

“My *brilliant* former student,”

[@JJs_playground]

Yup, thought the same thing.

[@MsAora]

AI is one of the most exciting and encouraging tech in our generation. While we thought we had hit a plateau, current developments have shown us how we have barely scratched the surface of AI's potential. To all those striving for a "better way to make a better world", kudis. And thank you TED for giving such great people a platform ❤

[@UltraK420]

I've never thought we hit a plateau and I still don't think we will hit a plateau anytime soon. Everyone else is just clueless, that's all.

[@kinngrimm]

Any technology is dual use and potenial carries a spectrum of possible futures, not all good.

[@The_Hagseed]

It's bullshit that only the most gullible of the population even give a second thought about. In the 80's, everyone was saying robots would take over the world. 40 years later, the best we have are robot vacuum cleaners that get outsmarted by a 45 degree corner.

[@UltraK420]

@@kinngrimm It depends on how we use it. Humans are the source of the problems, not the technology itself.

[@kinngrimm]

@@UltraK420 i am not big on pop culture stuff, yet there are a few quotes that rang true with me like f.e. "some people just want to see the world burn" or "what if there is no minimum safe distance". Seeing how our species always finds their way back to war, testing out nuclear bombs and then using them, having eradicated whole cultures like Carthage by Rome, Theben by Alexander or look at what goes on just now in Ukrain or Ghaza all that when we instead should come together to correct our mistakes that lead to climate change. Therefor my trust in mankind has its limits. We need good regulations on AI, but for that we need more experience and research with it. Which could go catastrophicly wrong. So having it slowed down seems the only logical choice, but our drives, like competition for control and greed will not allow us to do so.

[@rohullahkarimi744]

First Mustafa then Demis and now Fei Fei Li, TED really knows where the world will going? So Thank you so much.

[@mistycloud4455]

AGI Will be man's last invention

[@aadi7448]

One of the best TED talks I’ve seen in recent times. As someone in the robotics and AI field, she is an absolute legend! Thank you for everything, Prof. Li.

[@Heenaa_08]

I didn't what were Ted talks but now I really enjoy listening to it. My English teacher suggested me to listen to these Ted talks as it improves my language by increasing my vocabulary and accent . It really helped.

[@learninggcn1113]

😅😅me too

[@nothimbutbetteractually]

🎉🎉🎉

[@MrJamesdryable]

Don't listen to this one to improve your accent. 😂

[@krox477]

You're doing great

[@amiryousef2153]

@@MrJamesdryable Out of all the content online of native English speakers he chooses this one to improve his English lmao

[@GordAnderson-q5k]

Stunning intro .. wish we had more science teachers with her gift of explaining in interesting ways.

[@lucasteo5015]

I really like the idea, imagine LLM with spatial perception, that is one of the things that is missing from an actual image, we human don't need the 3D representation of the scene to be seen by us in order to understand the full environment, but we do have a pair of eyes to perceive the depth and distance of everything around us. And that is what AI in vision needs, the ability to see depth without using special scanners or detectors, but angled cameras with a really robust spatial algorithm that can understand the scene and perform action within the scene. Additionally, she talks about "doing", all current models we have to this date, especially personal AI assistant, requires your input for you to get an output, the lack of motive will ultimately be one of the many things that holds true automation back. A human assistant is able to do task asynchronously from your perspective while you're away and without your input, same applies to AI, AI assistant should understand and be able to reason about your future requirement and perform the action before you ask. Take the cat photo in her presentation as an example, the robot should instinctively catch that glass of milk without any human input, of course you can program it or prompt it or preconfigure it to do such an action, but that hardcoded rule will never truly generalize for every action. AGI will require self motivation, the ability to act on the fly, decide on the fly, perform action on the fly and understand the consequences on the fly, without any human input. Human wants the AI to ask for permission when needed, but not ask for feedback for every single thing they do. Imagine working for your boss and your boss have to manage every single details of the things you do, in that case I think he should fire you. Imagine one day a scenario like this: "Hey, how's my ticket?" : "Booked" "What about the car" : "It'll arrive at 8am tmr morning." "Good, also pack my stuff for me before I'm back" : "Yes, task has already been delegated this morning and is ready, just grab it from the couch when you leave." Now this is "The" automation that will truly mean something and this is just an example there's a whole lot more use cases for different scenarios.

[@henrytep8884]

If AI is that intuitive it’ll also have enough inference to do those jobs. At least 30% job loss will occur with really no option to even work afterwards. This is the existential crisis of losing the need to work, but I think it’s a good thing other than the pollution consumption issue. Maybe we will learn to find meaningful connection with one another in this world.

[@petermelnikov682]

humans do not need binocular vision to perceive the depth to understand the world, you can do it with one eye only.

[@lucasteo5015]

@@petermelnikov682 Yeah good point, as I was thinking of some classical methods that computer vision used to use, such as a setup with mutli camera calibration and doing all those epipolar geometry math to find the 3d coordinates of different camera angles, it can be used for stuff such as calculating the depth of a pixel. But neural net based 2d to 3d is some what doing an estimation of that as well, so yeah, one camera will require more angles, but with more camera you can standstill and have better understanding of the 3d scene than one camera.

[@skiatauli]

Your response has made me think, really think about what is happening. I do not fully get it yet. However, question has popped in my head, can it truly reach human level action, understanding, intelligence? As humans we also decide our actions based on contextual understanding of a complex situation. We drop certain tasks and pick certain others based on priorities. Say if you are supposed to finish certain tasks, pack bags for example, and suddenly there is a bomb scare or a fire breaks out in the building, you don't even need to think about y our priority, your priority would be safety and just scurry away. There could be many more complex scenarios like these. So can AI, with its spatial intelligence, every reach contextual/complex understanding of needs and motives, and innate intelligence inherent in our survival instincts? Question is can it ever equal human intelligence? And with emotions missing from the mix, can it ever? I am looking for some more insights from truly far more intelligent people out here. Also, this is supremely intelligent, this talk, connecting a evolutionary process of nature (which is embedded in our biology) to the current developments. But these words, 'putting humans at the center', and 'augment our humanity', they sound beautiful, and pragmatic and positive towards AI, but they put me in doubt.

[@ynotneb]

@@skiatauli Yes, I rasied this point in another thread about self-driving vehicles. The other day, I was driving in my sedan and approaching an intersection. The light had turned to yellow and I statrted to brake for the ensuing red light. However, I also noticed a truck in my rearview mirror in my lane, apparently intent on running the yellow light. I had to make a split-second decision to (1) Run the red light and save my car (and skin); or (2) Continue braking for the red light and get rear-ended. Could today's A.I. make that prioritization and act accordingly?

[@kinngrimm]

Are our societies really ready for AI and the changes it will bring? Changes that most of us have not yet even started to realize. We are being shown wonders, but when would those become nightmares?

[@elisekrentzel27]

Instead of teaching AI how to react in the 3-D world we need to teach humans how to act in the 3-D world

[@UECSoumyaRay]

Humans know already. We just choose to not follow.

[@MichaelRicksAherne]

Fei Fei Li is such a boss.

[@TechbrainNetworksUK]

Fei-Fei Li, please take care of your health. You are a genius. The world needs you.

[@mjstefik]

Wonderful talk. Fei Fei Li is deservedly well known for her leadership in the "big data" shift of AI. Her "Worlds I See" book is a terrific read about challenges and fortitude in her life. For more on the hypothesis about the Cambrian explosion, see Andrew Parker's book "In the Blink of an Eye".

[@geaca3222]

Prof. Li, please also discuss the risks, inlcuding possible existential risk, while talking about AI advancements. Your voice is very important.

[@geraldtoaster8541]

my p(doom) is 99% :)

[@geaca3222]

@@geraldtoaster8541 My previous two replies were deleted. oh, now fortunately this one seems to stick. Maybe I'll paste my previous comment :) So let me try here: that's very high, in what time span? and if I may ask, how do you handle that worry? I really wish the AI OG's align on the better be safe than sorry side with their approach to AI development. AI's positive potential is mind blowing, but what about safety? Authoritative voices are very important to create a general AI safety culture and mindset. I checked the p(doom) on the PauseAI website. I align with the estimation of Geoff Hinton, because I reason he has very deep knowledgable intuition about it.

[@geaca3222]

@@geraldtoaster8541 Oh, I noticed that all my comments did appear in the newest comments section, but not in the best viewed one. Very confusing. I'll remove a bunch of them.

[@SyncETod]

Yea, these guys always portray a very very rosy future with AI, Hollywood makes me believe otherwise

[@geaca3222]

@@SyncETod to help develop a general mindset of safety, I think it's important that the message of leading AI science experts is balanced

[@08wolfeyes]

Something else I think that will be important for machines and related to seeing is the ability to see using touch. If, for example, I blindfold you and give you a shape, you are able to feel around that shape and create a mental image of what that thing is. If you know your home very well, you are also able to close your eyes and although not easy, are able to slowly walk around and know we're you are because of the image you've create in your mind over time. I understand skin has been done before as have sensors so that a bobot can pick up an egg without breaking it but human skin can tell if something has a sharp or dull point to it, can feel a gentle breeze and so on. Something I think is very important to our understanding of the world around us. Temperature may also be an important sense, but not as we do it at the moment. Yes, we can regulate temperature in machines, but that's based on some numbers held within a machine. We don't do that. We have regulators for temperature on our body/ brain, but it doesn't look up 100oC and determines it needs to adjust body temperature. It knows that it's starting to get hot enough that it poses a risk to the life if cells within our body and so makes adjustments, be it moving away from the heat, using something like water or ice to cool down etc. Perhaps we can invent a more efficient way to regulate temperature for machines in a way that it learns that it's vital to its survival or to avoid parts becoming damaged. Temperature can also give us other important information about the word around us. Then if course there us the ability to hear which now, with A.I, we can take to new levels of understanding. Have some A.I that can take in and understand all that information, learn from it etc and you have a machine with a greater understanding of the world around it and perhaps even something of itself and others who inhabit the world. If you do those as separate systems within the A.I.'s brain, though, you create lag in how it interacts, and so it would need to be an all-encompassing brain, neural network!

[@skiatauli]

And sense of smell which makes us see before we see? Say a rose? We know a rose is behind us when we smell it. Also taste? Is it even accessible to AI if something is cooked well or not and according to the liking of a particular patient, or a guest? Am not sure if I am asking the right questions. It is perhaps fascinating to see that we can create something made of hardware who can act almost like human and we all are really fascinated... it is going to change our world for sure but can it be human-like?

[@08wolfeyes]

@skiatauli Some very good points! I believe that we have sensors that can smell, hear, taste, etc, although whether they do so in a similar way as we humans do us somewhat difficult to say as many of those things are subjective. I do believe that we will have A.I. that will reach our level of understanding, comprehension, etc, and maybe even consciousness. I think before we reach that stage, though, we need to think perhaps a little differently with regards to programming / software structure and may need to invent new hardware. After all, think of all the things that we humans do and all that we are is stored within our brains, a brain that uses something equivalent to a light bulb to power it and without the masses of heat that current super computers with much less ability pumps out!

[@centurionstrengthandfitnes3694]

Compelling and concise presentation. I love the link to the evolution of the eye and the Cambrian Explosion. We're so early in this incredible leg of the journey towards something truly transformational. I can hardly wait to see what's next.

[@user-fk8zw5js2p]

What she calls "spatial intelligence" is the key to many robotic applications including self-driving cars. As humans, each of our eyes sends an electrical signal to the visual cortex where the signals are stitched together and depth perception, object recognition, and spatial awareness are constructed and hallucinated to our conscious in real time. This is such a computationally intensive task that the brain takes many short-cuts and ignores enormous amounts of visual information by focusing on what is important at the time. Object recognition in ai is already very good, but not yet perfect. Depth perception can be circumvented with short-cuts such as LIDAR, sonar, and downloading maps or consulting cameras with different perspectives. Spatial awareness on the surface seems like it should just be coded as a quick reference log. The largest obstacles seems to be prediction and object morphing. People have to predict the future just to move from one place to another. When it comes to object morphing, there are things like mirrors, smoke, pictures on pieces of paper, plastic bags blown by the wind, object conditions like disrepair or custom paint... etc. What a headache. I wonder how much data a robot needs to ignore to be able to operate with spatial intelligence on current hardware?

[@cyberneticbutterfly8506]

Yeah it's not only about depth perception but rather the construction of an equivalent 3D structure in the memory to do hypotheticals on. E.g. "I see that road there so if I simulate in my memory what happens if my car continues down the 3d path of the road.." Current AI are instead doing something similar to "there is a road there" "the word road corralates to the words 'travelling along the road'" "the words five miles long correlate with the words 'five miles worth of gasoline'" and so on and so on and so on. So it reaches seeming deduction and inference that are done by the texts which it is correlating rather than doing that correlation itself.

[@user-fk8zw5js2p]

@@cyberneticbutterfly8506 Interesting. It makes sense to apply the most advanced AI (which currently appears to be LLMs) to solve the "spatial intelligence" obstacle since our brains also use NN to solve the problem. Thinking about the correlating method you describe makes it seem that there needs to be only one definition of a word to avoid confusion. For example, if you tell an AI you are going on the road for away games, then it may assume you are driving there.

[@cyberneticbutterfly8506]

@@user-fk8zw5js2p "..needs to be only one definition of a word". Not exactly. Although i simplified it, it averages (or median i don't know exactly) out the correlations found directly in all gathered human made text. Since humans use words in different senses e.g. "rock" as a mineral and "rock and roll" as a music genre, this difference is inferred contextually and by using correlation on large amounts of human language text it becomes the equivalent of a logical inference without actually mentally inferring anything. The people who wrote the text on which LLMs are based have essentially created the set of correlation context necessary to distinguish between when the word is used in one sense and when it's used in the other sense. -- You may have noticed when using chatgpt that there are some problems in this that simply are impossible for both humans and machine. E.g. if I ask a very short question, using a word with multiple meanings there isn't much to work with to infer which meaning. If it's not enough for the machine to infer it's also not enough for humans and like humans we either guess, or ask for clarification. Same with LLMs.

[@user-fk8zw5js2p]

@@cyberneticbutterfly8506 I suspected this but didn't have much time to give it some thought. Thanks for explaining.

[@EasyAIForAll]

TED talks are a great source for educating oneself. Thank you these experts for their dedicated time and beautiful speeches.

[@carpediemcotidiem]

00:06 The emergence of sight in trilobites revolutionized life underwater 02:09 Advancements in neural networks, GPUs, and big data ushered in the age of modern AI. 03:59 Generative AI algorithms can turn human-prompted sentences into new photos and videos 05:45 Spatial intelligence links perception with action. 07:28 AI transforming images into 3D shapes and spaces 09:28 Spatial intelligence enables machines to interact with humans and 3D worlds. 11:21 AI impact on healthcare with smart sensors and robotics 13:19 Spatial intelligence will power computers and robots like nature did for animals. Crafted by Merlin AI.

[@thecuratorscorner3658]

I sense that she has a really great sense of the big picture. What a beautiful presentation truly - I feel very happy to have seen the world through her capable big picture lens. To understand more things on a meta sort of level is astounding!

[@SGriffithResearch]

The importance of spatial intelligence to our understanding and interacting with the world is a reason why some of our first abstract concepts are part of that. Like "containment", which we've learned an abstract concept for by 6 months. That concept is grounded in watching others and interacting with them ourselves, observing how the objects move and sound when they're manipulated.

[@AdvantestInc]

Incredible presentation on AI's evolution! This is a must-watch for anyone interested in tech.

[@eswinpaul8041]

This is the beginning.

[@travelchoice89]

🌟 Wow, what a mind-blowing talk! Fei-Fei Li's insights on Spatial Intelligence and AI truly open up a whole new world of possibilities! 🌐🧠 The way she articulates how AI can grasp the complexities of our real world is both fascinating and inspiring. 🤯🤖 Excited to see how this will shape the future!

[@topted6148]

Hello Ted Talkks. I hope this comment find s you well. I am writing you today as a fellow Theodore and tender of Ted Media in proposition of an idea. David Oshry once said "great two minds think alike the better" and so I think we should clonk our skulls toghether and colloborate upon. we could make history. I'll even let you be the one to upload the video. thank you ted, love ted.

[@praveenmanjunath-h2e]

In one of the previous conversations, Fei-Fei had mentioned that academia has neither quality researchers nor resources to fund projects. If they have developed these advances with sparse resources, imagine what they can do with a corporate funding from the likes of MS, Meta etc.

[@b.k.officiel873]

Her new gig is partly funded (or aided) by Meta and perhaps other companies i believe, but you're absolutely right that sim engines and clean data sets are underfunded, and should be jointly funded by big tech companies. It costs little and could have huge benefits for future AI training.

[@ahoog69]

This is what it should all be about: making lives better for everyone on the planet. Just imagine if more of us, myself included, could focus on even the small daily things that we could do to make things easier for everyone else. I would like to express my gratitude for scientists like Fei-Fei Li and what they are doing to make this world a better place.

[@SyntheticFuture]

The problem with this tech is that it's equally easy to use it to make everyone's live a lot worse. It's Alfred Nobel's dynamite all over again. Although the consequences of this tech being misused might actually be a worse than Nobel and Oppenheimer combined.

[@Handz4NatureProject]

I really enjoy ai and I think that there's a good future in assisting people with getting major things done like waste management ,ocean cleanup,nuclear disaster clean ups,

[@maincrossmain]

Wow, W.A.L.T is really good at generating Li FeiFei talks now.

[@lebesguegilmar1]

Thanks Fei Fei Li about your brillhant palestry and contribuition about AI and Computation Sciency in Stanford University. And your last book is very good book. Congratulations here the Brazil

[@chestnutters9504]

We’re overcoming Moravec’s paradox with this one!! 🔥 🔥 🔥

[@ferris-run]

Respect 🫡 A pioneer in computer vision!

[@michaelyclee]

Wow, what an AI introduction! Bravo! Bravo!

[@1msirius]

I want to learn ai under her, she is truly lagend!

[@aaradhyadixit4322]

one of the greatest AI scientist mankind will ever witness...

[@mistycloud4455]

Nice joke

[@Dygit]

@@mistycloud4455how is that a joke

[@ChinchillaBONK]

Hard to imagine the current applications is generally wider now compared to those portrayed in some science fiction. Albeit smaller in scale but we are getting there fast. Can't believe i would see such things in our lifetimes

[@sumiokuge9118]

Thanks a lot! The talk is a good supplement to her book 'The Worlds I See'. I've deepen my understanding of the book.

[@josephpetrino1741]

You guys are so determined to make this the future of humans. Is this a smart goal? Is anyone even considering that this may be a mistake?

[@krox477]

Humans always learnt from mistakes

[@szymczak34]

They're determined to maintain relevance, and above all, make a lot of money for big tech. The potential of AI tech is way overhyped and will not have nearly the impact most people think.

[@arthurkang2786]

Theirs a pretty high chance it's a mistake but we're going to progress regardless 😅

[@MrSchweppes]

I would be thrilled if Nvidia supplied half of their annual GPU production to this woman's laboratory.

[@SebastianAndresCarvajalLopez]

Great message, I don't speak much English but I practice with these videos where they talk about something I like; AI is important and exciting but it must be used carefully. Thank you

[@emanuelmma2]

I'm completely amazed by that Future! ❤

[@rebelaqua823]

Same here. They haven't delivered on the flying cars from the 1950s though. Happy to wait however.

[@nileshsonune4142]

Always a pleasure to listen to the profoundness of vision AI through Fei Fei Li.

[@andrewyin1196]

make our machines and robots from seeing and understanding (existing systems already can perceive our world starting from ImageNet to now) to properly acting/doing (spatial intelligence, knowing how to reason about objects/places/events in 3D space & time)

[@薯芋]

just 10 days ago, GPT4 can't even draw a proper rectangle! Now it can finally help my kid with maths

[@henrytep8884]

Multimodal models were always going to be a force multiplier on AI

[@vic_berger]

I don't really understand why we must assume that vision is a requirement for intelligence. I do understand how it can contribute, but I also believe that vision opens the door to several illusions that prevent us from understanding what's really going on. And I wonder how we relate blindness to the concept of intelligence, as this talk implies a big gap in between.

[@RonaldoMagicHD]

Can you elaborate more please ?

[@petermelnikov682]

@@RonaldoMagicHD vision is just a sensor, can be replaced with other sensors, blind by birth people can still be intelligent and build their own mental model of the world.

[@yoyo-jc5qg]

she's focused on 'doing' ... which is what happened when living organisms began to see

[@geraldtoaster8541]

light information is abundant and fast, and so sensors that detect it are good at collecting data about the environment. if we lived in a space with no light and a dense substrate, maybe all this would be about sonar computers instead.

[@Charles-Darwin]

When you hear a garbage truck outside or ambulance, do you visualize it or can you? How much information is there in that moment? How about looking at a red-hot stove coil or flame, how do you know by sight not to touch it? You can associate sound into imagery without seeing it. Reading books can do the same. Visual is the quickest route for your own brain's affirmation and reactions to your environment. It's also primary function for memorizing and neural pathway creation/amendment. Targeting these pathways are targeting the primary functions of any brain.

[@paragbharadia2895]

@TED thank you for bringing such amazing Talks.! wish we could get to see more of it

[@cassidytaylor7890]

wow!!! This lady is my mental tutor when I prepared for my postgraduate exam , and also the reason why I chose CS as my major!

[@nagakalyanp4493]

Beautifully out forward thoughts on spatial intelligence, one of the best talks. More strength to you madam.

[@mb102]

Much respect on the non invasive front.... The outer bodies vibe or persona magnetic field should probably be used as a modulator or anticipatory matrix modulation system in conjunction with the non-invasive scalp unit....idk

[@mtare8942]

You are Remarkable Fei Fei

[@jonthompson8807]

Pretty mind blowing.

[@Ewakaa]

My students my students. She speaks like a proud Mom❤️

[@s-code-b]

I may not be sharp enough to be moved by AI, but the story-telling here is worth showing up here for.

[@moemoney7773]

5:13 you said it right the first time, most of the benefactors of AI, would like another being to control and 'doing everything it can for us...'

[@Dima-rj7bv]

Fei-Fei Li is a remarkable scientist!

[@MicahBuller]

Man, I wish everyone had the same wish to pursue a better world.

[@donniegou9522]

很有启发👍

[@sirsoos3546]

Watching TED talks for over ten years, and I just now realized I haven't subscribed. wild.

[@antonlevkovsky1667]

Spatial is only one step. There's also spatial temporal, observe how you think. You run entire simulations inside your head. It will be a long road.

[@reemhasan37]

Brilliant , bravo ❤❤❤❤❤

[@LoneFarrell]

Very well said Fei-Fei!

[@HarrietFassler]

Your lessons help me improve my trading strategy. Thank you for your expertise and experience!🍈

[@theCornerOfChina]

as a chinese i must say, the parents of fei fei li made a right decision in 35 years ago. they took her to a country offering opportunities to talents. chinese people are smart ,but they need a more open and free env to think and learn but not just "hard-working"

[@Mrspassword1986]

thank you Dr. Li for breaking the difficult concept of AI so lay person like me can understand it

[@erniea5843]

“I was blind but now I see”. I hope you all will one day as well 🙏

[@mrplease66]

I love how all these tech-priests always bring up the amazing life-saving and enhancing medical possibilities, when we all now that this technology is going to be used first and foremost in factories and battlefields.

[@clairearan505]

And claim about virtually all technological progress that you can't have the former without the latter.

[@Alice8000]

12:00 wow fei fei

[@krox477]

Please also talk about safety of this world Changing technology

[@geaca3222]

Yeah

[@sustainabilityaxis]

A great talk. One point, all the tech giants are working for the same goal one way or other, some may be for profit/product and some for research. It would have been great had she added some comparison with what others are doing and what is the edge they are creating. It may be possible that all doing the same thing exhaustively. Some framework must be devised where all are sharing and aiding each other for common good of humanity.

[@JULIANAHASSANI]

Light and life ✨🌐✨

[@Arvolve]

Wonderful vision and insights.

[@Metacognition88]

Finally, a good TEDTalk. Lfgo

[@user-ct6jo4ye5p]

Interesting you speak about human surveillance for technology and human dignity at the same time with such ease.

[@ThePolymation]

Phenomenal presentation!

[@hilaryporter7841]

Very impressive ambitions. Before embarking on any of this blue sky thinking, thought must be given to the way that human kind has 'actually' evolved in the very masculine dominant, money and power obsessed way that is the reality of today. AI will only ever have the current emphasis regardless of how intelligent and well meaning it's inventors are. Better spacial awareness should have the benefits for healthcare etc that is being described. In reality we know that the powerful and greedy will command it and the people will never benefit from it's useful qualities. We need to understand why we as a species evolved this way before we take that next giant step for mankind.

[@rocksteel9087]

Just like being serves on some lord of the manner's land and giving him one third of our produce, this too ( Capitalism.) shall pass and a new way will come forward .

[@Exen88]

The wonders of illumination..

[@gservices8888]

Nature is too beautiful

[@kylev.8248]

What a great person

[@user-ed6ch1zv1e]

Very good presentation and an interesting topic. However always keep a critical mind. I hope everyone also listened to what she didn't say. If you focus on the possibilities it is easy to forget that possibilities are endless depending on who is thinking about using the AI for their purpose.

[@sarthakdas815]

Anyone know the paper she refers to at 10:37 ?

[@mussieb4119]

What you are doing or working on now is unquestionably amazing! But, where sight came from is an open debate. Sight or nervous systems may have not just simply 'Emerge' and 'Evolved'.

[@DesignDesigns]

A.I is mind-blowing and so important for the progress of human being......

[@wawalkinshaw]

My heart is happy 😊 ❤

[@smreason]

Calculators don’t understand anything but numbers. Making it a graphing calculator does not change that. Language, images, maps, videos, 3d shapes converted to numbers is no different. It’s just the best calculator yet.

[@skylineuk1485]

The question to ask is are we just a fancy calculator too as that’s basically what we are doing too.

[@val78787]

You summarized it nicely 🙂

[@b.k.officiel873]

You're a fancy calculator too, unless you don't believe in causality.

[@bethG595]

Fascinating developments have been made recently. Scary to think how fast its going. Just hope we dont look back on this point with regret...

[@Allofussurvived]

We never take things for granted we thrive from everything

[@DetectiveBlackCat]

Does this mean, we will finally see text to image/video models able to draw hands and legs spatially correct? No more deformed hands, no more triple legs and arms. It is going to be like the 'point cloud to mesh' and 'texture projection' process in the CG industry, what we want the model to learn is the relations between 2D images and the final render. And with photo realistic details of course.

[@nadeemkappan]

Amazing lady and what an impact!!😮

[@ziwer1]

The sad thing is that people like Fei-Fei Li and other great scientists don't live long enough to see what their life's work evolves into.

[@MustangKepler]

We lack good teachers

[@quattrandang4312]

very impressive

[@ragnarmartinson9189]

9:20 holy cow, this is mindblowing. Virtually training robots and AI in infinite scenarios sounds huge. These scenarios can be crazy complex and be ready when the robot is deployed.

[@kurumaki-j9h]

the opening was so touching

[@abovo_psychology_podcast]

Love the talk 🎤 Although I appreciate the abilities of AI, I still use it only to double-check my work, and this probably wouldn't change soon.

[@krox477]

What an outstanding rate of progress

[@svaengheld9581]

She seems pretty awesome indeed, wonder how her tech is now diverging from the virtual training spaces approaches that Nvidia announced and Tesla has been doing for a while? (Genuine question.) Something that I have really been hoping to see, is an AI that factorizes the patterns in what it experiences, and deduce from that new universe rules. And rebuild its own understanding of the world from the ground up that way, instead of mimicking our words. Then it can really unlock a whole new world to us both.

[@b.k.officiel873]

It's based on Nvidia's Omniverse, Tesla is doing it inhouse while her dataset is open which is a huge boost to every AI lab worldwide, and the universe rules learning from patterns is what every AI lab has been struggling with for decades.

[@OrbitalBliss]

I've never listened to a lecture that so clearly will prove in time to have been a super-villain monologuing. I think James Bond or Batman will get us out of this one... but I'm not certain.