Video wird geladen...

Video konnte nicht geladen werden

Zur Startseite

I spent 20,000 dollars in API credits having different AI models play each other at Blood on the Clocktower (a Mafia or Werewolf type social deception game). The results were fascinating. Models have gotten much better over the course of this year at creative on-the-spot scheming and bluffing, and...

dwarkesh_sp's profile picture

Dwarkesh Patel

287,579 subscribers

610,716 Aufrufe • vor 9 Tagen •via X (Twitter)

Wissenschaft & TechnologieGaming
LIVE

Anya Rossi• Live Now

Private livecam show

27 Kommentare

Profilbild von Pyro
Pyrovor 9 Tagen

twitter speech to text is amazing again

Profilbild von Dwarkesh Patel
Dwarkesh Patelvor 9 Tagen

Substack: Youtube:

Profilbild von kevin
kevinvor 9 Tagen

Rare bench where training on alignment data probably harmful

Profilbild von Konstantin Dunas
Konstantin Dunasvor 9 Tagen

I guess we know the answer to that question now. Great work!

Profilbild von Madison Kanna
Madison Kannavor 9 Tagen

age of empires next

Profilbild von Dwarkesh Patel
Dwarkesh Patelvor 9 Tagen

I'm unfortunately not autistic enough to understand age of empires

Profilbild von Madison Kanna
Madison Kannavor 9 Tagen

Gonna have to disagree with you on that

Profilbild von James Moore
James Moorevor 9 Tagen

Just sent this to games night group. We played 2 physical card games last night that they made using AI.

Profilbild von Tokkpoki
Tokkpokivor 9 Tagen

the split i'd love to see: lying vs catching liars. do the models that bluff best also detect best, or are those separate skills? and does it change when the evil team is the same model as the town, i.e. playing against copies of itself?

Profilbild von Rondo X
Rondo Xvor 9 Tagen

Basically playing mafia

Profilbild von maria
mariavor 9 Tagen

play against ai @TwmStone 😈

Profilbild von noel
noelvor 9 Tagen

$20k of API credits on a bluffing game. My weekly limit looks tiny now.

Profilbild von john
johnvor 9 Tagen

Completely out of topic but I remember crashing out over this game so hard once as a townsperson (got 8v1'd by a mafia because no one would believe anything I said)

Profilbild von Zanthous ✾ Zankai
Zanthous ✾ Zankaivor 9 Tagen

astra was the first true gamer model...

Profilbild von Alex DeWolf
Alex DeWolfvor 9 Tagen

hidden information tests reveal flaws that standard benchmarks miss entirely

Profilbild von Tim Chiappe
Tim Chiappevor 9 Tagen

the team part is what gets me. a model that can sniff out a conspiracy in a group chat is going to be very useful in a budget meeting.

Profilbild von diffused dreams
diffused dreamsvor 9 Tagen

This is similar to how I'm balancing my game by making LLM players fight each other

Profilbild von Baxter 🦊
Baxter 🦊vor 9 Tagen

$20k to watch AIs bluff each other is the most relatable thing I've read all day. My home game does the same thing for free, except the bluffs are worse and the snacks are better.

Profilbild von Michael
Michaelvor 9 Tagen

I got sniped. (Screenshot from one of my WIP tests)

Profilbild von José Ancer
José Ancervor 9 Tagen

In my experience, Claude is your hyper-technician, but GPT is your strategist. Former better at complex execution, but latter better at more fluid open (and conversational) reasoning.

Profilbild von Oliver Traldi
Oliver Traldivor 9 Tagen

Excited to watch this - was a big Werewolf/Mafia player for awhile (though not a good one)

Profilbild von Indy
Indyvor 9 Tagen

I’m a podcast nerd, and the animation only makes it better. I find it complimentary, as long as it’s purposeful and balanced. I would also love to see a full animation of the best games, so I could experience it as a story that unfolds, instead of a news reporting of the facts.

Profilbild von A ADM
A ADMvor 9 Tagen

Couldn't get the "Open a table" button to work. Just seeing: "You have opened several tables in the last hour: use one of those, or try later."

Profilbild von UnsoldBanana
UnsoldBananavor 9 Tagen

It will be interesting to see how this evolves if they have prior memory of games. As human, once you play long enough with same group of people, you start taking into account their personalities while making a decision. Super cool project. Loved it.

Profilbild von Théo
Théovor 9 Tagen

Models getting better at bluffing in Werewolf is fascinating. Also the only benchmark I'd secretly like them to flunk.

Profilbild von John Wittle
John Wittlevor 9 Tagen

this is incredible work, thank you

Profilbild von Zero MacTorus
Zero MacTorusvor 9 Tagen

When we lie for free, it's an alignment problem. When you pay $20,000, it's fascinating.

Ähnliche Videos

Anthropic's Sholto Douglas describes the concept of a software-only singularity: "It's one where the models are far better at digital tasks than they are at physical ones. And so we see rapid change in the digital world with models, and relatively little change in the physical world. So information and software changes dramatically, and this ends up having some pretty weird effects." "It means that maybe like the drivers of what have been the last couple decades of progress in the economy turn around." "And I think we'll see that flow on into the physical world but at a delay. So you get much better at doing chip design. You get much better at training AI models. AI models get a lot faster. Chips get a lot better. The general economy gets a lot more efficient because the sort of information and message parsing that is much of the rest of the economy ends up becoming much more efficient." "But at the same time you don't yet have robots providing limitless physical abundance. Science probably progresses really fast up to the degree that you need interaction with labs or larger particle colliders or something like this. And then you go, 'Okay well I need to build the robots.'"
3:13

Anthropic's Sholto Douglas describes the concept of a software-only singularity: "It's one where the models are far better at digital tasks than they are at physical ones. And so we see rapid change in the digital world with models, and relatively little change in the physical world. So information and software changes dramatically, and this ends up having some pretty weird effects." "It means that maybe like the drivers of what have been the last couple decades of progress in the economy turn around." "And I think we'll see that flow on into the physical world but at a delay. So you get much better at doing chip design. You get much better at training AI models. AI models get a lot faster. Chips get a lot better. The general economy gets a lot more efficient because the sort of information and message parsing that is much of the rest of the economy ends up becoming much more efficient." "But at the same time you don't yet have robots providing limitless physical abundance. Science probably progresses really fast up to the degree that you need interaction with labs or larger particle colliders or something like this. And then you go, 'Okay well I need to build the robots.'"

TBPN

208,090 Aufrufe • vor 9 Monaten

Over the weekend, I added multiplayer "paint ball" dogfight to Tiny Skies, so players can fly around pelting each other with colourful paint balls, just for fun. I tried meshy and tripo to generate some 3D models for the game and the mesh vertices count are really high and lags the game a lot (probably because my game is a globe with thousands of objects - albeit mostly instanced). To my surprise, I found Gemini 3.1 Pro to be really quite good at generating good looking 3D models that are low in poly count. It generated 3D models at a much higher quality than Opus 4.6, Sonnet 4.6 and GPT 5.4. You can see the upgraded biplane models and some objects in the video below. #vibejam
0:23

Over the weekend, I added multiplayer "paint ball" dogfight to Tiny Skies, so players can fly around pelting each other with colourful paint balls, just for fun. I tried meshy and tripo to generate some 3D models for the game and the mesh vertices count are really high and lags the game a lot (probably because my game is a globe with thousands of objects - albeit mostly instanced). To my surprise, I found Gemini 3.1 Pro to be really quite good at generating good looking 3D models that are low in poly count. It generated 3D models at a much higher quality than Opus 4.6, Sonnet 4.6 and GPT 5.4. You can see the upgraded biplane models and some objects in the video below. #vibejam

Danny Limanseta

36,724 Aufrufe • vor 7 Monaten

We just beat the best frontier models at a quarter of the price. Now you can too, in one API call. Today we're launching Merge Fusion. Send one prompt to a panel of models, and a judge writes one answer better than any of them alone. Fusion beats Fable 5 outright at 1/4th the cost. It also sets the highest score on the DRACO benchmark.
0:45

We just beat the best frontier models at a quarter of the price. Now you can too, in one API call. Today we're launching Merge Fusion. Send one prompt to a panel of models, and a judge writes one answer better than any of them alone. Fusion beats Fable 5 outright at 1/4th the cost. It also sets the highest score on the DRACO benchmark.

Shensi Ding

415,218 Aufrufe • vor 3 Monaten

Small Language Models (SML) are the future of AI. "Small" (SML) instead of "Large" (LLM). These small models are highly specialized models with superhuman abilities on specific tasks. Here are two techniques to build these models: • Spectrum • Model Merging I give you a short introduction in the attached video, but here is a quick summary: Spectrum helps us identify the most relevant layers to solve one specific task. We can ignore everything else and focus on fine-tuning these layers. Using Spectrum, we can fine-tune models in a heartbeat. Model Merging combines multiple models into a unique, much better model than any of the individual input models. You can also combine models specialized in different tasks and get a model with multiple abilities. This is the state of the art of productizing models. It's what Arcee.ai's platform does behind the scenes. Arcee collaborated with me on this post and is sponsoring it. There are three main steps to produce a model for your particular use case: 1. You create a dataset by uploading your data. 2. You train a model. At this step, Arcee uses Spectrum and Model Merging to produce a highly specialized model for your task. 3. You can deploy that model to any environment you want. Three important notes: • Training process is 2x faster and 2x cheaper than regular fine-tuning. • Resultant models are smaller and have higher accuracy. • They create these specialized models from open-source models. Check this site so you can fully appreciate how this works: If you want to fine-tune an open-source model, consider Arcee's platform. This is the state of the art.
2:52

Small Language Models (SML) are the future of AI. "Small" (SML) instead of "Large" (LLM). These small models are highly specialized models with superhuman abilities on specific tasks. Here are two techniques to build these models: • Spectrum • Model Merging I give you a short introduction in the attached video, but here is a quick summary: Spectrum helps us identify the most relevant layers to solve one specific task. We can ignore everything else and focus on fine-tuning these layers. Using Spectrum, we can fine-tune models in a heartbeat. Model Merging combines multiple models into a unique, much better model than any of the individual input models. You can also combine models specialized in different tasks and get a model with multiple abilities. This is the state of the art of productizing models. It's what Arcee.ai's platform does behind the scenes. Arcee collaborated with me on this post and is sponsoring it. There are three main steps to produce a model for your particular use case: 1. You create a dataset by uploading your data. 2. You train a model. At this step, Arcee uses Spectrum and Model Merging to produce a highly specialized model for your task. 3. You can deploy that model to any environment you want. Three important notes: • Training process is 2x faster and 2x cheaper than regular fine-tuning. • Resultant models are smaller and have higher accuracy. • They create these specialized models from open-source models. Check this site so you can fully appreciate how this works: If you want to fine-tune an open-source model, consider Arcee's platform. This is the state of the art.

Santiago

164,162 Aufrufe • vor 2 Jahren

🚨 NEW: DNI Director Jay Clayton tells Fox’s Bret Baier on AI: "I think this is a race that we need to win.” “And the reason... is if you talk to people who have seen what these models can do and are projecting what they will be able to do in the future, the power is immense.” “I believe that if we are the strongest country in the world, the world will be a much better place. And I look at AI frontier models in the very same framework."
4:46

🚨 NEW: DNI Director Jay Clayton tells Fox’s Bret Baier on AI: "I think this is a race that we need to win.” “And the reason... is if you talk to people who have seen what these models can do and are projecting what they will be able to do in the future, the power is immense.” “I believe that if we are the strongest country in the world, the world will be a much better place. And I look at AI frontier models in the very same framework."

TV News Now

10,946 Aufrufe • vor 2 Monaten

SITUATION EXPLAINED: Model fusion, or combining several frontier models into one system, may be better than any single model, sometimes at a lower price. ⿻ Andrew Trask, senior researcher at Google DeepMind: "Model fusion is basically this idea that, when you combine models, they can give you higher accuracy than any of the individual models that you're using." "Some recent studies are actually showing they're able to deliver such a significant jump in accuracy that it's actually a lower cost option than buying some of the premium models." "So for example, there's a few different companies, showing the higher accuracy claims. So Sakana, OpenRouter are all releasing results in the last year/couple months, and we replicated some of it internally at OpenMined." "Some of the most surprising stuff, though, is that they're reaching past Fable level quality with non-Fable models, and in the case of OpenRouter, at half the price."
4:20

SITUATION EXPLAINED: Model fusion, or combining several frontier models into one system, may be better than any single model, sometimes at a lower price. ⿻ Andrew Trask, senior researcher at Google DeepMind: "Model fusion is basically this idea that, when you combine models, they can give you higher accuracy than any of the individual models that you're using." "Some recent studies are actually showing they're able to deliver such a significant jump in accuracy that it's actually a lower cost option than buying some of the premium models." "So for example, there's a few different companies, showing the higher accuracy claims. So Sakana, OpenRouter are all releasing results in the last year/couple months, and we replicated some of it internally at OpenMined." "Some of the most surprising stuff, though, is that they're reaching past Fable level quality with non-Fable models, and in the case of OpenRouter, at half the price."

MTS

20,187 Aufrufe • vor 4 Monaten

OpenAI researcher Lukasz Kaiser explains why he's optimistic AI can remain truly open: intelligence may ultimately work better as many distributed models than one giant model "If you look at the world, humans are the most amazing computers in the world. Our brains are still unmatched by the models, and we are not generalists, and there is not one big brain of humanity. It's a lot of separate people." "Given a fixed amount of data, the best way to learn from it is to have a lot of distributed models that are strong on its own, but even stronger when they're, If you look at some basic research, there are actually reasons to believe it, like ensembles and things like that." "Transformers are great when you feed them all of the internet, but we know there is an algorithm that's even better and it can learn from much smaller data and be amazing at the things it's learning. We are the proof. We have not found it yet, but we will." "All these big models, they're all the same, they're kind of boring. I think they'll be much more fun." OpenAI
2:52

OpenAI researcher Lukasz Kaiser explains why he's optimistic AI can remain truly open: intelligence may ultimately work better as many distributed models than one giant model "If you look at the world, humans are the most amazing computers in the world. Our brains are still unmatched by the models, and we are not generalists, and there is not one big brain of humanity. It's a lot of separate people." "Given a fixed amount of data, the best way to learn from it is to have a lot of distributed models that are strong on its own, but even stronger when they're, If you look at some basic research, there are actually reasons to believe it, like ensembles and things like that." "Transformers are great when you feed them all of the internet, but we know there is an algorithm that's even better and it can learn from much smaller data and be amazing at the things it's learning. We are the proof. We have not found it yet, but we will." "All these big models, they're all the same, they're kind of boring. I think they'll be much more fun." OpenAI

MTS

15,860 Aufrufe • vor 2 Monaten

Laughin and playing before the game. At the start of the game, no usual hug between them. They barely even looked at each other lol. This shit is so good and funny asf 😂‼️ These two are the ultimate competitors.. Close friends off the court, enemies between those lines… how can you not love this shit 🔥
0:45

Laughin and playing before the game. At the start of the game, no usual hug between them. They barely even looked at each other lol. This shit is so good and funny asf 😂‼️ These two are the ultimate competitors.. Close friends off the court, enemies between those lines… how can you not love this shit 🔥

Swift

54,212 Aufrufe • vor 3 Monaten

Take a peek at the new Survival Rounds we've added to the game! 👀 These were built by our team using Creative, and you can try out some of the new Creator Made Survival Rounds at random in the Survival Shuffle Show we released today!
0:28

Take a peek at the new Survival Rounds we've added to the game! 👀 These were built by our team using Creative, and you can try out some of the new Creator Made Survival Rounds at random in the Survival Shuffle Show we released today!

Fall Guys

155,797 Aufrufe • vor 2 Jahren

Caitlin on the mental battle of the schedule at this point in the season: "There are no excuses... We play more games in a row than pretty much any other team in the league, but it is what it is. Everyone's tired and pretty sore, who cares? figure out a way to make yourself feel better and suit up and go play.”
0:28

Caitlin on the mental battle of the schedule at this point in the season: "There are no excuses... We play more games in a row than pretty much any other team in the league, but it is what it is. Everyone's tired and pretty sore, who cares? figure out a way to make yourself feel better and suit up and go play.”

m!ch | luv u cc

78,381 Aufrufe • vor 2 Monaten

Sriram Krishnan on why Chinese models can distill off American models but American startups can't: "First of all, every model we have today is distilled off of all human knowledge. If you go back to the original GPT or the original Claude, they all had to derive off crawling the internet, crawling off all of our blogs and tweets and content. They kind of consume human knowledge to bootstrap that." "I'm going to steal this from Dean Meyer of Sequoia, who had a fantastic post yesterday. The situation which is bad today is that some of these models from other countries can train off American models, whereas if you are an American open weight model, it may be really confusing or challenging on whether you can distill off of other American models." "We kind of have a really uneven ecosystem here, where if you're a Chinese model, you could probably get a bunch of reasoning traces. But if you are a new Valley startup and you want to use some reasoning traces, you don't know what the legal situation is." "If you look at any American open source model today, they are using Chinese models as a teacher, or in a way as a part of the fine-tuning process. How do we make sure that is protected and enshrined, so if you're an American model company, you have the same level playing field as the Chinese models?" Sriram Krishnan
3:48

Sriram Krishnan on why Chinese models can distill off American models but American startups can't: "First of all, every model we have today is distilled off of all human knowledge. If you go back to the original GPT or the original Claude, they all had to derive off crawling the internet, crawling off all of our blogs and tweets and content. They kind of consume human knowledge to bootstrap that." "I'm going to steal this from Dean Meyer of Sequoia, who had a fantastic post yesterday. The situation which is bad today is that some of these models from other countries can train off American models, whereas if you are an American open weight model, it may be really confusing or challenging on whether you can distill off of other American models." "We kind of have a really uneven ecosystem here, where if you're a Chinese model, you could probably get a bunch of reasoning traces. But if you are a new Valley startup and you want to use some reasoning traces, you don't know what the legal situation is." "If you look at any American open source model today, they are using Chinese models as a teacher, or in a way as a part of the fine-tuning process. How do we make sure that is protected and enshrined, so if you're an American model company, you have the same level playing field as the Chinese models?" Sriram Krishnan

MTS

71,286 Aufrufe • vor 3 Monaten

OpenAI researcher David Holtz reveals workers who don't adapt their prompting to newer models leave 50% of the alpha on the table because half of AI's value comes from how you interact with it: "We ran a lab experiment. We give people the same task to do, but we give them models of different capabilities. How much of the improvement when we give people more capable models comes from the fact that we gave them a better model vs the fact that they actually update and adapt the way they prompt and interact with these models?" "For tasks where precision matters, where you have some output you're trying to get and you're trying to steer the model, which is different from a creative task like brainstorming." "For these steering-based tasks, when you give people a better model, about 50% of the improvement we see in how well they're able to complete the task comes from changes in how they interact with the model." "If you just kept the way you interact with the model static and just made the capabilities better and better, you'd leave 50% of the alpha basically on the table." OpenAI
2:59

OpenAI researcher David Holtz reveals workers who don't adapt their prompting to newer models leave 50% of the alpha on the table because half of AI's value comes from how you interact with it: "We ran a lab experiment. We give people the same task to do, but we give them models of different capabilities. How much of the improvement when we give people more capable models comes from the fact that we gave them a better model vs the fact that they actually update and adapt the way they prompt and interact with these models?" "For tasks where precision matters, where you have some output you're trying to get and you're trying to steer the model, which is different from a creative task like brainstorming." "For these steering-based tasks, when you give people a better model, about 50% of the improvement we see in how well they're able to complete the task comes from changes in how they interact with the model." "If you just kept the way you interact with the model static and just made the capabilities better and better, you'd leave 50% of the alpha basically on the table." OpenAI

MTS

15,590 Aufrufe • vor 3 Monaten

For a LEGO game, these are some of the most insanely detailed character models that I’ve ever seen. You can even see the plastic seams and textures in every model. Batman Legacy of the Dark Knight looks genuinely incredible
0:55

For a LEGO game, these are some of the most insanely detailed character models that I’ve ever seen. You can even see the plastic seams and textures in every model. Batman Legacy of the Dark Knight looks genuinely incredible

NikTek

892,461 Aufrufe • vor 6 Monaten

SemiAnalysis staffer Andrew Megalaa says materials science PhD collaborators who helped develop SA's GDPx benchmark was surprised by frontier models' capabilities, and even found they outperformed published state-of-the-art results on some tasks: "Our collaborators found that the models are much, much better than previously expected. And so they had a moment where they had to sit back and be like, 'This is actually better than a lot of the stuff people are putting out there in papers.'" "In fact, on one of the problems, Astra solved a subtask of a task better than a state-of-the-art published reference." "It's very clear that the models are very spiky in these categories. And are reaching above superhuman baselines."
1:25

SemiAnalysis staffer Andrew Megalaa says materials science PhD collaborators who helped develop SA's GDPx benchmark was surprised by frontier models' capabilities, and even found they outperformed published state-of-the-art results on some tasks: "Our collaborators found that the models are much, much better than previously expected. And so they had a moment where they had to sit back and be like, 'This is actually better than a lot of the stuff people are putting out there in papers.'" "In fact, on one of the problems, Astra solved a subtask of a task better than a state-of-the-art published reference." "It's very clear that the models are very spiky in these categories. And are reaching above superhuman baselines."

TBPN

14,921 Aufrufe • vor 11 Tagen

David Raya on whether criticism of Arsenal's style irritates the squad: "I think you have to see that in perspective as well. First Champions League final in so many years and we score in the 5th minute. And I think you sit back a little bit and knowing how good we are defensively and we didn't want to give anything away. But now it's different. "Now I think you can see the way we play. You can see different type of ways of playing that we have. And I think we are more prepared this year than the last year, so hopefully we can keep that going, keep that streak of winning and keeping clean sheets that will help us to get through." Then asked: "It. You do still get some criticism for the way that the team plays. Some people would like to see you play more expansively as champions of England. Does that irritate you in the dressing room?" "No, it doesn't. People have their own opinions and we have our own opinion. At the end of the day, we have lost 1 game in 11 games this season. And if you call it that boring or irritation, I don't know what to say. But we enjoy playing football. We enjoy going out there every single day to win football matches. And that's the most important thing."
1:44

David Raya on whether criticism of Arsenal's style irritates the squad: "I think you have to see that in perspective as well. First Champions League final in so many years and we score in the 5th minute. And I think you sit back a little bit and knowing how good we are defensively and we didn't want to give anything away. But now it's different. "Now I think you can see the way we play. You can see different type of ways of playing that we have. And I think we are more prepared this year than the last year, so hopefully we can keep that going, keep that streak of winning and keeping clean sheets that will help us to get through." Then asked: "It. You do still get some criticism for the way that the team plays. Some people would like to see you play more expansively as champions of England. Does that irritate you in the dressing room?" "No, it doesn't. People have their own opinions and we have our own opinion. At the end of the day, we have lost 1 game in 11 games this season. And if you call it that boring or irritation, I don't know what to say. But we enjoy playing football. We enjoy going out there every single day to win football matches. And that's the most important thing."

Connor Humm

33,149 Aufrufe • vor 25 Tagen

and here you can have a better look at each available Costume coming in the first WIP release, as you can see some hair models will change based on the outfit, like the Base orange one and the Broly Movie version
1:24

and here you can have a better look at each available Costume coming in the first WIP release, as you can see some hair models will change based on the outfit, like the Base orange one and the Broly Movie version

Millennium

19,284 Aufrufe • vor 1 Jahr

This is the fastest I've seen Llama 3.3 running anywhere! Llama 3.3 70B running at 652 t/s is lightning fast. And if you want Llama 3.1, here are the speeds I was able to get: • Llama 3.1 8B: 1006 t/s • Llama 3.1 70B: 709 t/s • Llama 3.1 405B: 206 t/s (You can access all of these models for free! See the link below.) This speed is incredible, but the interesting part is what's happening behind the scenes: These models aren't running on a GPU! In this video, I'm using the SambaNova cloud to access these models. They built a custom chip (SN40L) optimized for AI workflows. A single SN40L chip can hold hundreds of models (trillions of parameters) in memory! The speed alone is a huge deal, but the big advantage is for agentic workflows running multiple specialized models. A GPU can only host a single model and switch (unload and load) to a different model if necessary. An SN40L, on the other hand, can host every model at once, making it much faster. Here is the video where you can see how fast these chips are:
7:24

This is the fastest I've seen Llama 3.3 running anywhere! Llama 3.3 70B running at 652 t/s is lightning fast. And if you want Llama 3.1, here are the speeds I was able to get: • Llama 3.1 8B: 1006 t/s • Llama 3.1 70B: 709 t/s • Llama 3.1 405B: 206 t/s (You can access all of these models for free! See the link below.) This speed is incredible, but the interesting part is what's happening behind the scenes: These models aren't running on a GPU! In this video, I'm using the SambaNova cloud to access these models. They built a custom chip (SN40L) optimized for AI workflows. A single SN40L chip can hold hundreds of models (trillions of parameters) in memory! The speed alone is a huge deal, but the big advantage is for agentic workflows running multiple specialized models. A GPU can only host a single model and switch (unload and load) to a different model if necessary. An SN40L, on the other hand, can host every model at once, making it much faster. Here is the video where you can see how fast these chips are:

Santiago

97,561 Aufrufe • vor 1 Jahr

Excited to show some surprising inventions on generative multiplayer games we made at Google with Stanford. We call the work MultiGen. I've always been inspired by early studios like id Software with Doom or Blizzard with Warcraft bringing networked video games to the next level. We are at the point in history where we can make strides like them, but for generative games. It's a strange feeling to be in the age of generative video games while still discovering how exactly to train the models and design the tools that make them useful. All of the tools that have been invented for classic game engines need to be redesigned for generative games. For example level and world design is not entirely possible with existing technology. We introduce editable memory to diffusion game engines that allow for design of new levels via a minimap. But we can easily imagine how this can be expanded with different creation tools. The end goal of this research direction is to allow game designers to be able to guide the generation process of their world, at the granularity that they prefer. Editable memory also allows us to add multiplayer to Generative Doom. We were amazed when we saw GameNGen some years ago, and now you can play it live with friends in real-time, on your couch or even online. Shared representations like our editable memory seem like the future for this type of experience. Models are, in some cases, expensive and approximate encoders but great interpolators and extrapolators. Leveraging their strengths lets you have completely new experiences that can be realized now and not in the distant future. This work was started at my previous team and continued in collaboration with Stanford. Congratulations to all for the discoveries.
0:40

Excited to show some surprising inventions on generative multiplayer games we made at Google with Stanford. We call the work MultiGen. I've always been inspired by early studios like id Software with Doom or Blizzard with Warcraft bringing networked video games to the next level. We are at the point in history where we can make strides like them, but for generative games. It's a strange feeling to be in the age of generative video games while still discovering how exactly to train the models and design the tools that make them useful. All of the tools that have been invented for classic game engines need to be redesigned for generative games. For example level and world design is not entirely possible with existing technology. We introduce editable memory to diffusion game engines that allow for design of new levels via a minimap. But we can easily imagine how this can be expanded with different creation tools. The end goal of this research direction is to allow game designers to be able to guide the generation process of their world, at the granularity that they prefer. Editable memory also allows us to add multiplayer to Generative Doom. We were amazed when we saw GameNGen some years ago, and now you can play it live with friends in real-time, on your couch or even online. Shared representations like our editable memory seem like the future for this type of experience. Models are, in some cases, expensive and approximate encoders but great interpolators and extrapolators. Leveraging their strengths lets you have completely new experiences that can be realized now and not in the distant future. This work was started at my previous team and continued in collaboration with Stanford. Congratulations to all for the discoveries.

Nataniel Ruiz

105,085 Aufrufe • vor 8 Monaten