正在加载视频...

视频加载失败

加载此视频时出现问题。这可能是由于临时网络问题，或视频可能不可用。

Using light as a neural network, What if AI didn’t need electricity to think? A new kind of chip uses pure light instead of electrons. In 5–10 years, matrix multiplications that currently take huge energy and time could run almost instantly using photons with up to 95% less power.... show more

Md Ismail Šojal 🕷️

50,076 subscribers

99,092 次观看 • 18 天前 •via X (Twitter)

科学技术教育

Anya Rossi• Live Now

Private livecam show

0 条评论

暂无评论

原始帖子的评论将显示在这里

相关视频

Marc Andreessen says AI is teaching sand to think and it could be the most important technology in the history of humanity: "Imagine a form of alchemy that turns sand into thought." "Chips are made out of sand. They're made out of silicon, so they're literally made out of sand." "We plug the chip into a data center, into power, we light it up, and we put AI on it, and all of a sudden it's thinking." "We've turned sand into thought. And so it's possibly the most revolutionary technology in the history of the species." "It's certainly on par with electricity and steam power. It's certainly more important than the internet." Marc Andreessen 🇺🇸 with Joe Rogan

Marc Andreessen says AI is teaching sand to think and it could be the most important technology in the history of humanity: "Imagine a form of alchemy that turns sand into thought." "Chips are made out of sand. They're made out of silicon, so they're literally made out of sand." "We plug the chip into a data center, into power, we light it up, and we put AI on it, and all of a sudden it's thinking." "We've turned sand into thought. And so it's possibly the most revolutionary technology in the history of the species." "It's certainly on par with electricity and steam power. It's certainly more important than the internet." Marc Andreessen 🇺🇸 with Joe Rogan

a16z

223,308 次观看 • 3 个月前

$What actually goes into a data center? Guru Chahal🇺🇸 whiteboards the entire stack - across land, power, cooling, storage, networking, compute - and shares where we're seeing the most interesting innovation + startup opportunities → Base Power: virtual power plants from residential batteries → Periodic Labs: using AI to discover new materials - ones that run cooler, conduct better, and waste less power → Nexthop AI: network switches that are 10–15% more power efficient than what's on the market → Unconventional AI: AI chips modeled after how the human brain works, targeting 100–1,000x better power efficiency → Ricursive Intelligence: AI to design chips. Today a first chip costs $100M and takes 2–3 years. They want to build in months, at a fraction of the cost → better materials → cheaper experiments → more chip designs → more efficient chips → less power → and what happens when you package all of this into a satellite and put it in orbit 🚀 As trillions of dollars of infrastructure flow into building the "AI factory", every layer of the intelligence economy is being rebuilt at once. Who are we missing? We'd love to meet if you are building somewhere on this stack.$

What actually goes into a data center? Guru Chahal🇺🇸 whiteboards the entire stack - across land, power, cooling, storage, networking, compute - and shares where we're seeing the most interesting innovation + startup opportunities → Base Power: virtual power plants from residential batteries → Periodic Labs: using AI to discover new materials - ones that run cooler, conduct better, and waste less power → Nexthop AI: network switches that are 10–15% more power efficient than what's on the market → Unconventional AI: AI chips modeled after how the human brain works, targeting 100–1,000x better power efficiency → Ricursive Intelligence: AI to design chips. Today a first chip costs $100M and takes 2–3 years. They want to build in months, at a fraction of the cost → better materials → cheaper experiments → more chip designs → more efficient chips → less power → and what happens when you package all of this into a satellite and put it in orbit 🚀 As trillions of dollars of infrastructure flow into building the "AI factory", every layer of the intelligence economy is being rebuilt at once. Who are we missing? We'd love to meet if you are building somewhere on this stack.

Claire Zau

41,288 次观看 • 25 天前

Is uranium the future of AI? As demand for cloud computing and AI models continues to explode, the real bottleneck is becoming power. That’s what led me to start looking at nuclear energy and uranium. Small modular reactors could become a major part of the solution, and if AI keeps demanding more electricity, the commodities that power that infrastructure could become just as important as the chips themselves.

Is uranium the future of AI? As demand for cloud computing and AI models continues to explode, the real bottleneck is becoming power. That’s what led me to start looking at nuclear energy and uranium. Small modular reactors could become a major part of the solution, and if AI keeps demanding more electricity, the commodities that power that infrastructure could become just as important as the chips themselves.

Kevin O'Leary aka Mr. Wonderful

34,028 次观看 • 5 天前

ELON MUSK: The Future of Energy Is Beyond Earth. "It's increasingly difficult to build power plants on the ground. There are very few people who want a power plant in their backyard, so if we wanted to say double the electricity usage of the United States, which is on average about 500 gigawatts, we would have to build, about twice as many power plants, which I don't think people are, well, most communities are not super excited about that, but actually, if we go to space, we can go far beyond the electricity generation of Earth. In fact, this is going to sound kind of crazy, but you could actually increase harnessed energy by a factor of a million and still be using much less than a millionth of the sun's energy, so current human civilization uses much less than a trillionth of the sun's energy output, which is kind of humbling to think about."

ELON MUSK: The Future of Energy Is Beyond Earth. "It's increasingly difficult to build power plants on the ground. There are very few people who want a power plant in their backyard, so if we wanted to say double the electricity usage of the United States, which is on average about 500 gigawatts, we would have to build, about twice as many power plants, which I don't think people are, well, most communities are not super excited about that, but actually, if we go to space, we can go far beyond the electricity generation of Earth. In fact, this is going to sound kind of crazy, but you could actually increase harnessed energy by a factor of a million and still be using much less than a millionth of the sun's energy, so current human civilization uses much less than a trillionth of the sun's energy output, which is kind of humbling to think about."

DogeDesigner

1,091,758 次观看 • 2 个月前

what if you can run an entire 0-person company — without the grind of running a team? matrix is the runtime that makes it possible. in last week’s limited beta, our users created tens of thousands of new 0-person companies and started real businesses in matrix. today, matrix is open to everyone. launch yours ↓

what if you can run an entire 0-person company — without the grind of running a team? matrix is the runtime that makes it possible. in last week’s limited beta, our users created tens of thousands of new 0-person companies and started real businesses in matrix. today, matrix is open to everyone. launch yours ↓

Matrix

525,956 次观看 • 1 个月前

“I think next decade is going to be all about energy. Even Sam Altman from OpenAI talked about the evolution of AI — algorithms run on chips, but ultimately the chips need electricity. The algorithms will get better and better, and chips will get cheaper and cheaper. But at the end of the day, you have to get the electrons, and those electrons have a fundamental price. Even the AI competition between companies and countries is ultimately going to come back to electricity production. So next decade is going to be all about power — AI, manufacturing, the economy — they all come back to energy.” Scott Nolan

“I think next decade is going to be all about energy. Even Sam Altman from OpenAI talked about the evolution of AI — algorithms run on chips, but ultimately the chips need electricity. The algorithms will get better and better, and chips will get cheaper and cheaper. But at the end of the day, you have to get the electrons, and those electrons have a fundamental price. Even the AI competition between companies and countries is ultimately going to come back to electricity production. So next decade is going to be all about power — AI, manufacturing, the economy — they all come back to energy.” Scott Nolan

Shawn Ryan

118,797 次观看 • 1 年前

Elon Musk explaining why we’ll need solar-powered AI satellites in space within the next five years. Not because it sounds futuristic, but because Earth simply can’t generate the energy needed for what’s coming. AI alone could require 300 gigawatts of power every year To put that into perspective, that’s almost two-thirds of the entire U.S. electricity output, just for AI. And scaling up to a terawatt per year? That’s beyond what Earth’s grids and infrastructure can handle. Space is the only place with the endless power supply AI will demand.

Elon Musk explaining why we’ll need solar-powered AI satellites in space within the next five years. Not because it sounds futuristic, but because Earth simply can’t generate the energy needed for what’s coming. AI alone could require 300 gigawatts of power every year To put that into perspective, that’s almost two-thirds of the entire U.S. electricity output, just for AI. And scaling up to a terawatt per year? That’s beyond what Earth’s grids and infrastructure can handle. Space is the only place with the endless power supply AI will demand.

SMX 🇺🇸

33,351 次观看 • 8 个月前

.Martin Shkreli says photonic computing's "time is coming" - "it's a pretty insane speedup versus GPUs." "It's kind of like quantum's ugly little stepsister, but I think it's the belle of the ball." "Recently in Nature, and a few other preeminent journals, especially Chinese companies, have shown you can do MatMuls [matrix multiplication] with light." "It's actually pretty remarkable. Light goes to 100 THz. But more important than that is you can actually do 3D MatMuls, and they're O(1) in complexity," meaning they take the same amount of time no matter how big the problem is. "More and more of the chip and entire global computer infrastructure is going photonic. And maybe the whole thing goes photonic someday. And electricity is actually the odd man out." From his January appearance on the show.

.Martin Shkreli says photonic computing's "time is coming" - "it's a pretty insane speedup versus GPUs." "It's kind of like quantum's ugly little stepsister, but I think it's the belle of the ball." "Recently in Nature, and a few other preeminent journals, especially Chinese companies, have shown you can do MatMuls [matrix multiplication] with light." "It's actually pretty remarkable. Light goes to 100 THz. But more important than that is you can actually do 3D MatMuls, and they're O(1) in complexity," meaning they take the same amount of time no matter how big the problem is. "More and more of the chip and entire global computer infrastructure is going photonic. And maybe the whole thing goes photonic someday. And electricity is actually the odd man out." From his January appearance on the show.

TBPN

576,497 次观看 • 4 个月前

Nvidia just spent $4 billion on a technology 99% of people have never heard of. But in 3 years, every AI data center on Earth will need it. And Nvidia just LOCKED UP the supply. Here's what happened: Nvidia invested $2 billion in Coherent and $2 billion in Lumentum. You probably never heard of these companies. They make photonics technology. Systems that transmit data using LIGHT instead of electricity. Sounds like sci-fi. But this is the most important infrastructure bet in AI right now. Here's the problem Nvidia just solved for itself: AI data centers are hitting a wall that has nothing to do with chips, energy, or money... Copper wiring is dying. Every data center on Earth moves data between GPUs using copper cables. But at the speeds AI now demands, copper physically cannot keep up. Signal degrades. Heat explodes. Power consumption skyrockets. Right now, 30% of the electricity in an AI data center is wasted just MOVING data from point A to point B. An MIT researcher said: "Copper's not going to cut it. It gets too hot. Too much power consumption and loss." Jensen Huang admitted it himself too: "We use copper as far as we can, about a meter or two. But where data centers are the size of a stadium, we need something else." That something else is photonics. Replacing copper with laser-powered fiber optics built directly into the chip. The numbers are insane: - 3.5x more power efficient - 10x better network reliability - Data moving at 102 terabits per second Wells Fargo estimates the photonics market will hit $10-12 billion by 2030. And Nvidia just bought privileged access to the two companies that make the advanced lasers every single one of these systems will need. This is the Nvidia playbook on repeat. They did this with CoreWeave. Invested $2 billion, locked up GPU capacity, created a dependent customer. They did this with memory suppliers. Secured HBM allocations years in advance while competitors scrambled. Now they're doing it with photonics. Invest early. Lock up supply. Make the entire ecosystem dependent on companies that are dependent on Nvidia. By the time competitors realize photonics is the bottleneck, Nvidia already OWNS the supply chain. Every data center, AI factory, and GPU cluster will need this technology to function at scale. Nvidia will become even more important.

Nvidia just spent $4 billion on a technology 99% of people have never heard of. But in 3 years, every AI data center on Earth will need it. And Nvidia just LOCKED UP the supply. Here's what happened: Nvidia invested $2 billion in Coherent and $2 billion in Lumentum. You probably never heard of these companies. They make photonics technology. Systems that transmit data using LIGHT instead of electricity. Sounds like sci-fi. But this is the most important infrastructure bet in AI right now. Here's the problem Nvidia just solved for itself: AI data centers are hitting a wall that has nothing to do with chips, energy, or money... Copper wiring is dying. Every data center on Earth moves data between GPUs using copper cables. But at the speeds AI now demands, copper physically cannot keep up. Signal degrades. Heat explodes. Power consumption skyrockets. Right now, 30% of the electricity in an AI data center is wasted just MOVING data from point A to point B. An MIT researcher said: "Copper's not going to cut it. It gets too hot. Too much power consumption and loss." Jensen Huang admitted it himself too: "We use copper as far as we can, about a meter or two. But where data centers are the size of a stadium, we need something else." That something else is photonics. Replacing copper with laser-powered fiber optics built directly into the chip. The numbers are insane: - 3.5x more power efficient - 10x better network reliability - Data moving at 102 terabits per second Wells Fargo estimates the photonics market will hit $10-12 billion by 2030. And Nvidia just bought privileged access to the two companies that make the advanced lasers every single one of these systems will need. This is the Nvidia playbook on repeat. They did this with CoreWeave. Invested $2 billion, locked up GPU capacity, created a dependent customer. They did this with memory suppliers. Secured HBM allocations years in advance while competitors scrambled. Now they're doing it with photonics. Invest early. Lock up supply. Make the entire ecosystem dependent on companies that are dependent on Nvidia. By the time competitors realize photonics is the bottleneck, Nvidia already OWNS the supply chain. Every data center, AI factory, and GPU cluster will need this technology to function at scale. Nvidia will become even more important.

Ricardo

640,885 次观看 • 5 个月前

Jensen Huang: What if I told you that in the last decade, we’ve made AI computing 100,000 times more efficient? Imagine if your car could become 100,000 times faster or use 100,000 times less energy. That’s what we’ve achieved with AI. In just 10 years, the energy required to run AI will be minimal, making it accessible in everything – from smartphones to homes – all the time. The future of AI means it will power our daily lives without draining resources or creating environmental impact. This is happening now, and the potential is limitless.

Jensen Huang: What if I told you that in the last decade, we’ve made AI computing 100,000 times more efficient? Imagine if your car could become 100,000 times faster or use 100,000 times less energy. That’s what we’ve achieved with AI. In just 10 years, the energy required to run AI will be minimal, making it accessible in everything – from smartphones to homes – all the time. The future of AI means it will power our daily lives without draining resources or creating environmental impact. This is happening now, and the potential is limitless.

Ian Miles Cheong

35,071 次观看 • 8 个月前

ELON: "WE'LL HAVE MORE AI CHIPS THAN POWER TO RUN THEM" "The limiting factor for AI deployment is fundamentally electrical power. It's energy. We're seeing the rate of AI chip production increase exponentially, but the rate of electricity being brought online is 5% a year, max. It's clear that we're very soon, maybe even later this year, we'll be producing more chips than we can turn on." Source: Elon Musk

ELON: "WE'LL HAVE MORE AI CHIPS THAN POWER TO RUN THEM" "The limiting factor for AI deployment is fundamentally electrical power. It's energy. We're seeing the rate of AI chip production increase exponentially, but the rate of electricity being brought online is 5% a year, max. It's clear that we're very soon, maybe even later this year, we'll be producing more chips than we can turn on." Source: Elon Musk

Mario Nawfal

40,874 次观看 • 7 个月前

Evaluating Neural Networks at the Speed of Light (with Light!). See live optical inference in the video below. Excited to share recent academic work on optical neural networks as a collection of computing elements embedded in the camera lens! These elements perform computation optically even before an image is captured, using the photons in the scene instead of GPU computation after the capture. We were able to achieve ImageNet classification more than two orders of magnitude faster than conventional neural networks on today's GPUs at almost no power consumption! To do this, we developed an array of metalenses that perform this computation on light from the scene. Project: Paper: Amazing collaboration with Kaixuan Wei, Xiao Li, Johannes Froech, Praneeth Chakravarthula, James Whitehead, Ethan Tseng , Arka Majumdar .

Evaluating Neural Networks at the Speed of Light (with Light!). See live optical inference in the video below. Excited to share recent academic work on optical neural networks as a collection of computing elements embedded in the camera lens! These elements perform computation optically even before an image is captured, using the photons in the scene instead of GPU computation after the capture. We were able to achieve ImageNet classification more than two orders of magnitude faster than conventional neural networks on today's GPUs at almost no power consumption! To do this, we developed an array of metalenses that perform this computation on light from the scene. Project: Paper: Amazing collaboration with Kaixuan Wei, Xiao Li, Johannes Froech, Praneeth Chakravarthula, James Whitehead, Ethan Tseng , Arka Majumdar .

Felix Heide

29,904 次观看 • 1 年前

Great #ces2026 Report #7: The company inside the coolest AI products at CES. Makes low power AI chips inside a wide range of consumer products. Watches, rings, AI glasses, medical devices, all use their chips because they are best at running neural networks at very low power. I've been touring the suites all morning, getting a look at augmented reality AI glasses. It's quite clear that the Pickle glasses are not possible with today's technology, and I'll have a video up someday soon about that with several different vendors. So agree with others that it is selling something that isn’t currently possible to manufacture. I got a look at the latest state-of-the-art technologies that go into the glasses, and this is the first part of that. This gives you a lot of insights into the future.

Great #ces2026 Report #7: The company inside the coolest AI products at CES. Makes low power AI chips inside a wide range of consumer products. Watches, rings, AI glasses, medical devices, all use their chips because they are best at running neural networks at very low power. I've been touring the suites all morning, getting a look at augmented reality AI glasses. It's quite clear that the Pickle glasses are not possible with today's technology, and I'll have a video up someday soon about that with several different vendors. So agree with others that it is selling something that isn’t currently possible to manufacture. I got a look at the latest state-of-the-art technologies that go into the glasses, and this is the first part of that. This gives you a lot of insights into the future.

Robert Scoble

17,396 次观看 • 7 个月前

How many of the big ideas of the past 15 years of AI are downstream of hardware constraints? The big hardware story over that period is that logic has become way cheaper than data transfer. Stacking huge numbers of matrix multiplies was perfect for this hardware regime, because matrix multiplication is logic-intensive but requires less data transfer. And so we got matmul-heavy deep learning. It's interesting to think about what AI would look like in a world where these costs didn't diverge so much.

How many of the big ideas of the past 15 years of AI are downstream of hardware constraints? The big hardware story over that period is that logic has become way cheaper than data transfer. Stacking huge numbers of matrix multiplies was perfect for this hardware regime, because matrix multiplication is logic-intensive but requires less data transfer. And so we got matmul-heavy deep learning. It's interesting to think about what AI would look like in a world where these costs didn't diverge so much.

Dwarkesh Patel

61,147 次观看 • 2 个月前

A major gas field in the UK could supply over 10% of the UK's energy, but that gas isn't destined for the grid. The plan instead is to extract it on site, burn it in a local power plant, and use that electricity to mine bitcoin. The energy will be consumed at the source, because that route is simpler and more profitable. Taxes, regulation and grid constraints all add friction. The gas could power homes. It isn't. Because burning it on site makes more money. That is a damning indictment of UK energy policy.

A major gas field in the UK could supply over 10% of the UK's energy, but that gas isn't destined for the grid. The plan instead is to extract it on site, burn it in a local power plant, and use that electricity to mine bitcoin. The energy will be consumed at the source, because that route is simpler and more profitable. Taxes, regulation and grid constraints all add friction. The gas could power homes. It isn't. Because burning it on site makes more money. That is a damning indictment of UK energy policy.

Electroverse

15,529 次观看 • 4 个月前

Elon Musk explains how data centers in Space could help the US compete with China in AI “Given that Chinese AI companies are doing as well as they are with a relatively small amount of compute, it seems that if they had a lot of compute, there’s a good chance they would be the leaders. The constraints on AI are essentially electricity and AI chips. China is very strong in digital AI, and it has far more electricity than the United States. In fact, China already has more electricity than the United States, Europe and India combined. The US government can prevent US companies from using Chinese models, but it can’t stop the rest of the world from using them. That won’t stop China from becoming the leader in AI. Digital intelligence comes down to how much AI compute, how many chips and how much electricity you have. The current constraint is electricity outside China. In China, the constraint is chips. China is also closer than most people realize to solving the lithography problem, which would enable it to make AI chips in very large volumes. Once we address the power constraint with AI data centers in space, the constraint will once again be chips outside China.”

Elon Musk explains how data centers in Space could help the US compete with China in AI “Given that Chinese AI companies are doing as well as they are with a relatively small amount of compute, it seems that if they had a lot of compute, there’s a good chance they would be the leaders. The constraints on AI are essentially electricity and AI chips. China is very strong in digital AI, and it has far more electricity than the United States. In fact, China already has more electricity than the United States, Europe and India combined. The US government can prevent US companies from using Chinese models, but it can’t stop the rest of the world from using them. That won’t stop China from becoming the leader in AI. Digital intelligence comes down to how much AI compute, how many chips and how much electricity you have. The current constraint is electricity outside China. In China, the constraint is chips. China is also closer than most people realize to solving the lithography problem, which would enable it to make AI chips in very large volumes. Once we address the power constraint with AI data centers in space, the constraint will once again be chips outside China.”

DogeDesigner

245,139 次观看 • 1 个月前

Elon Musk: For the next few years, I think America is likely to win the artificial intelligence race. But that early lead is fragile and short-lived. Ultimately, the global victor will be decided by a single factor: it will be a function of who controls the AI chip fabrication. The factories that make the AI chips, who controls them? That is where the real power lies. If those facilities are controlled, if more of them are controlled by China, then China will win the future of AI. Right now, almost all the advanced AI chip factories, they call them fabs, are located in Taiwan. That immense concentration of manufacturing creates a massive vulnerability for the rest of the globe. If China invades Taiwan, what happens to the world? Well, if they were to invade in the near term, the entire world would essentially be cut off from advanced AI chips overnight. Everything would grind to a halt. Currently, one hundred percent of advanced AI chips are made in Taiwan. We are relying entirely on a single location for the critical hardware that will define the next era of human technology and military dominance. Bringing that manufacturing power back home and securing our own supply chain is absolutely essential for national security, and the harsh reality is that we are simply not doing enough to prepare.

Elon Musk: For the next few years, I think America is likely to win the artificial intelligence race. But that early lead is fragile and short-lived. Ultimately, the global victor will be decided by a single factor: it will be a function of who controls the AI chip fabrication. The factories that make the AI chips, who controls them? That is where the real power lies. If those facilities are controlled, if more of them are controlled by China, then China will win the future of AI. Right now, almost all the advanced AI chip factories, they call them fabs, are located in Taiwan. That immense concentration of manufacturing creates a massive vulnerability for the rest of the globe. If China invades Taiwan, what happens to the world? Well, if they were to invade in the near term, the entire world would essentially be cut off from advanced AI chips overnight. Everything would grind to a halt. Currently, one hundred percent of advanced AI chips are made in Taiwan. We are relying entirely on a single location for the critical hardware that will define the next era of human technology and military dominance. Bringing that manufacturing power back home and securing our own supply chain is absolutely essential for national security, and the harsh reality is that we are simply not doing enough to prepare.

Ian Miles Cheong

409,258 次观看 • 4 个月前

Sam Altman on AI's energy consumption. Spending energy on AI is necessary to solve humanity's biggest problems. 🎯 "Even if we have to use hundreds of megawatts or gigawatts on this problem if we can use AI to discover how to do efficient cheap fusion and then very quickly replicate the thousands of GWs of generating capacity that's burn in carbon around the world that would be a huge win and I think that is the promise of AI We are totally screwed if we do not come up with new scientific solutions to address the climate challenge in front of us. We clearly have failed to do that fast enough on our own without AI. Let's try it with AI" - Sam Alman He believes current models are already "incredibly efficient" on a per-query or "watts per token" basis, even comparing them favorably to the energy a human consumes to think. Draws a parallel to the early concerns about Google's energy use, noting that a search query used far less energy than the alternative at the time, such as driving to a library. Using AI to discover a breakthrough like cheap nuclear fusion would be a massive net positive for the climate, far outweighing the energy AI consumes. --- Full video in TUBerlinTV YT channel (link in comment)

Sam Altman on AI's energy consumption. Spending energy on AI is necessary to solve humanity's biggest problems. 🎯 "Even if we have to use hundreds of megawatts or gigawatts on this problem if we can use AI to discover how to do efficient cheap fusion and then very quickly replicate the thousands of GWs of generating capacity that's burn in carbon around the world that would be a huge win and I think that is the promise of AI We are totally screwed if we do not come up with new scientific solutions to address the climate challenge in front of us. We clearly have failed to do that fast enough on our own without AI. Let's try it with AI" - Sam Alman He believes current models are already "incredibly efficient" on a per-query or "watts per token" basis, even comparing them favorably to the energy a human consumes to think. Draws a parallel to the early concerns about Google's energy use, noting that a search query used far less energy than the alternative at the time, such as driving to a library. Using AI to discover a breakthrough like cheap nuclear fusion would be a massive net positive for the climate, far outweighing the energy AI consumes. --- Full video in TUBerlinTV YT channel (link in comment)

Rohan Paul

48,203 次观看 • 11 个月前