Loading video...

Video Failed to Load

Go Home

We currently train AIs by having them predict human text outputs. Could we train it by having them predict human neural patterns? w. Adam Marblestone

dwarkesh_sp's profile picture

Dwarkesh Patel

242,535 subscribers

37,610 views • 3 months ago •via X (Twitter)

LIVE

Anya Rossi• Live Now

Private livecam show

0 Comments

No comments available

Comments from the original post will appear here

Related Videos

Cognitive scientist Joscha Bach: "at the moment, we are not building systems that are minds — we are building systems that predict text and visual patterns made for human consumption" True machine perception has never been tried, but once it is, it may surpass human reasoning
0:47

Cognitive scientist Joscha Bach: "at the moment, we are not building systems that are minds — we are building systems that predict text and visual patterns made for human consumption" True machine perception has never been tried, but once it is, it may surpass human reasoning

Haider.

61,279 views • 28 days ago

Wouldn't it be great if we could train robots without any teleoperation! In our latest paper, we train robots to mimic a human video of the task by simply matching the object features using RL. We only need one video and under an hour of robot training.
0:22

Wouldn't it be great if we could train robots without any teleoperation! In our latest paper, we train robots to mimic a human video of the task by simply matching the object features using RL. We only need one video and under an hour of robot training.

Lerrel Pinto

46,221 views • 1 year ago

"By defending human rights, we can prevent harm before it happens. "By advancing human rights, we create a fairer world. "By protecting human rights, we protect our future." - United Nations Human Rights Council President Omar Zniber #HumanRightsDay #OurRightsOutFuture
1:04

"By defending human rights, we can prevent harm before it happens. "By advancing human rights, we create a fairer world. "By protecting human rights, we protect our future." - United Nations Human Rights Council President Omar Zniber #HumanRightsDay #OurRightsOutFuture

UN Human Rights Council

29,762 views • 1 year ago

1/10 image-based screening advanced drug discovery but scaling to massive perturbation space is hard! Given a cell image, we asked if we could predict the morphological effect induced by a perturbation! Led by Alessandro Palma & w Fabian Theis we propose IMPA
1:03

1/10 image-based screening advanced drug discovery but scaling to massive perturbation space is hard! Given a cell image, we asked if we could predict the morphological effect induced by a perturbation! Led by Alessandro Palma & w Fabian Theis we propose IMPA

Mo Lotfollahi

40,663 views • 3 years ago

Another cinematic ruined by not having human eyes so we just don’t know what emotion they were trying to convey. (Skill issssssue)
0:55

Another cinematic ruined by not having human eyes so we just don’t know what emotion they were trying to convey. (Skill issssssue)

Grox

104,828 views • 11 months ago

Meta's Joe Spisak explains how AI models can train themselves by generating images, asking itself questions about them, and choosing the best answers, in order to move beyond human data and human fine-tuning, and teach itself from synthetic data
1:13

Meta's Joe Spisak explains how AI models can train themselves by generating images, asking itself questions about them, and choosing the best answers, in order to move beyond human data and human fine-tuning, and teach itself from synthetic data

Tsarathustra

42,390 views • 1 year ago

What if AI learned physics the way Newton did – by experiencing it? We built Sim2Reason: train LLMs inside virtual worlds governed by real physics laws, zero human annotation. Result: +5–10% improvement on International Physics Olympiad, zero-shot. 🧵
1:00

What if AI learned physics the way Newton did – by experiencing it? We built Sim2Reason: train LLMs inside virtual worlds governed by real physics laws, zero human annotation. Result: +5–10% improvement on International Physics Olympiad, zero-shot. 🧵

Mihir Prabhudesai

247,253 views • 3 months ago

There are currently foreign nationals from more than 40 countries being held hostage by Hamas. We ask you as parents. As children. As human beings. Share their stories and help us bring them home.
0:53

There are currently foreign nationals from more than 40 countries being held hostage by Hamas. We ask you as parents. As children. As human beings. Share their stories and help us bring them home.

Israel ישראל

710,905 views • 2 years ago

“Helping animals by having better vaccines” When Bill Gates isn’t trying to push experimental mRNA vaccines into every Human on Earth, he’s trying to get them into animals instead.
0:34

“Helping animals by having better vaccines” When Bill Gates isn’t trying to push experimental mRNA vaccines into every Human on Earth, he’s trying to get them into animals instead.

Concerned Citizen

114,941 views • 1 year ago

Does GPT understand the world? Here is what Ilya Sutskever, co-founder of OpenAI, says during a discussion with Jensen Huang, CEO of Nvidia: (1) When we train a large neural network to accurately predict the next word in lots of different texts from the internet, the AI is learning a world model. (2) On the surface, it may look like learning correlations in text, but it turns out that to 'just learn' statistical correlations in text, to compress information really well, what the neural network learns is some representation of the process that produced the text. (3) This text is a projection of the world...what the neural network is learning is aspects of the world, of people, of the human conditions, their hopes, dreams, motivations, their interactions...the situations we are in. The neural network learns a compressed, abstract, usable representation." Do you think learning representations = understanding? Are large language models simply stochastic parrots, or are they much more?
3:10

Does GPT understand the world? Here is what Ilya Sutskever, co-founder of OpenAI, says during a discussion with Jensen Huang, CEO of Nvidia: (1) When we train a large neural network to accurately predict the next word in lots of different texts from the internet, the AI is learning a world model. (2) On the surface, it may look like learning correlations in text, but it turns out that to 'just learn' statistical correlations in text, to compress information really well, what the neural network learns is some representation of the process that produced the text. (3) This text is a projection of the world...what the neural network is learning is aspects of the world, of people, of the human conditions, their hopes, dreams, motivations, their interactions...the situations we are in. The neural network learns a compressed, abstract, usable representation." Do you think learning representations = understanding? Are large language models simply stochastic parrots, or are they much more?

Alex Ker 🔭

1,367,097 views • 2 years ago

The protection of human rights – and the commitment to uphold them by governments – should not diminish as we age. Demand a world where human rights last a lifetime, #ProtectTheProtest and #AgeLoud 📢
1:00

The protection of human rights – and the commitment to uphold them by governments – should not diminish as we age. Demand a world where human rights last a lifetime, #ProtectTheProtest and #AgeLoud 📢

Amnesty International

15,011 views • 1 year ago

A must-watch clip. Konstantin Kisin asks, “How will AI destroy humanity?” AI safety expert Roman Yampolskiy explains why that’s the wrong question. We assume AI will think like us. It won’t. A superintelligence could act far outside human understanding, like humans compared to squirrels. It wouldn’t need to hate us. It could wipe us out simply by pursuing something else. And the uncomfortable truth? We’re not in full control. Dr. Roman Yampolskiy explains that we don’t code these systems, we train them, then study what they become.
2:36

A must-watch clip. Konstantin Kisin asks, “How will AI destroy humanity?” AI safety expert Roman Yampolskiy explains why that’s the wrong question. We assume AI will think like us. It won’t. A superintelligence could act far outside human understanding, like humans compared to squirrels. It wouldn’t need to hate us. It could wipe us out simply by pursuing something else. And the uncomfortable truth? We’re not in full control. Dr. Roman Yampolskiy explains that we don’t code these systems, we train them, then study what they become.

TRIGGERnometry

135,052 views • 3 months ago

Super excited to share the last paper of my PhD: "Hallucination in World Models is Predictable and Preventable"✨ We train a 350M-param generative world model on a large dataset w/ 210 tasks and show that we can predict *when* hallucination happens and use that to fix it! 🧵1/n
0:40

Super excited to share the last paper of my PhD: "Hallucination in World Models is Predictable and Preventable"✨ We train a 350M-param generative world model on a large dataset w/ 210 tasks and show that we can predict *when* hallucination happens and use that to fix it! 🧵1/n

Nicklas Hansen

54,573 views • 1 month ago

“We told our daughter we were having our son, we did it by giving her papa bear and mama bear and the baby bear, little arrow pointing to the mama bear’ stomach” aww🥺🥺 “But she was so disappointed it was a human” ROSE 😂😂
0:27

“We told our daughter we were having our son, we did it by giving her papa bear and mama bear and the baby bear, little arrow pointing to the mama bear’ stomach” aww🥺🥺 “But she was so disappointed it was a human” ROSE 😂😂

Keeping Up with Scarlett

29,080 views • 2 years ago

“Food is a human right, water is a human right, shelter is a human right, and if we taxed people appropriately, we could provide all that sh*t” -economic liberal explaining how we need to tax more again ✨🌈 Why is barista math and economics so difficult for them?
1:01

“Food is a human right, water is a human right, shelter is a human right, and if we taxed people appropriately, we could provide all that sh*t” -economic liberal explaining how we need to tax more again ✨🌈 Why is barista math and economics so difficult for them?

Meme’nOnLibs

61,642 views • 1 year ago

Number one economy on earth, the American economy. How was it built? Immigration. We curate the best, we bring them here to create jobs and companies. We curate these amazing students from all over the world and we bring them into a cohort to train them how to become entrepreneurs in America, they want to stay. If we don't do it the right way, the Chinese will. Do we want the Chinese to train them or do we want them? We want them to stay here and I want them get their passports, their families get passports, they'll create jobs and I will invest in them.
0:58

Number one economy on earth, the American economy. How was it built? Immigration. We curate the best, we bring them here to create jobs and companies. We curate these amazing students from all over the world and we bring them into a cohort to train them how to become entrepreneurs in America, they want to stay. If we don't do it the right way, the Chinese will. Do we want the Chinese to train them or do we want them? We want them to stay here and I want them get their passports, their families get passports, they'll create jobs and I will invest in them.

Kevin O'Leary aka Mr. Wonderful

33,081 views • 8 months ago

.Sen. Cory Booker: "It is very frustrating to me that we still have a repressive regime in power in Venezuela suppressing human rights, yet we are cooperating with them, we are funding them..."
6:51

.Sen. Cory Booker: "It is very frustrating to me that we still have a repressive regime in power in Venezuela suppressing human rights, yet we are cooperating with them, we are funding them..."

CSPAN

14,180 views • 6 months ago

SHOCKING NEWS! When we asked if it would be easier to train this lab puppy NOT to steal the leather sandals or if it would be easier to train an adult human male to put them away, most of you said the puppy! The results are in! Video description for inclusivity: a light yellow lab puppy sitting by a door. Beside her is a cabinet. It is opened to show a pair of Birkenstock sandals.
0:33

SHOCKING NEWS! When we asked if it would be easier to train this lab puppy NOT to steal the leather sandals or if it would be easier to train an adult human male to put them away, most of you said the puppy! The results are in! Video description for inclusivity: a light yellow lab puppy sitting by a door. Beside her is a cabinet. It is opened to show a pair of Birkenstock sandals.

Team Servicerottie🇨🇦🐕‍🦺🦽

16,223 views • 9 months ago

"There were what we might call hominids- human like creatures- we might even call them proto-humans" Apologist N.T. Wright explains that if Adam and Eve literally existed, they were likely were 'human-like' primates that lived 200,000-5 million years ago.
1:31

"There were what we might call hominids- human like creatures- we might even call them proto-humans" Apologist N.T. Wright explains that if Adam and Eve literally existed, they were likely were 'human-like' primates that lived 200,000-5 million years ago.

Protestia

141,768 views • 3 months ago