Video yükleniyor...

Video Yüklenemedi

Bu video yüklenirken bir sorun oluştu. Bu geçici bir ağ sorunundan kaynaklanıyor olabilir veya video kullanılamıyor olabilir.

Ana Sayfaya Dön

Excited to introduce Intelligent Go-Explore: Foundation model (FM) agents have the potential to be invaluable, but struggle to learn hard-exploration tasks! Our new algorithm drastically improves their exploration abilities via Go-Explore + FM intelligence. Led by Cong Lu🧵1/

Jeff Clune

32,741 subscribers

67,116 görüntüleme • 2 yıl önce •via X (Twitter)

Eğitim Sağlık & İyilik Bilim & Teknoloji

Anya Rossi• Live Now

Private livecam show

10 Yorum

Jeff Clune profil fotoğrafı

Jeff Clune2 yıl önce

We revisit Go-Explore, a powerful exploration algorithm based on archiving discovered states & returning to and exploring from promising states. With the help of domain-specific heuristics, Go-Explore (Nature, 2021) achieved superhuman performance on Atari games and robotics! 2/

Jeff Clune profil fotoğrafı

Jeff Clune2 yıl önce

Key insight: leverage the ability of Foundation Models to judge interestingness in place of heuristics, extending Go-Explore to automatically tackle virtually any problem. IGE captures serendipity on the fly, including discoveries that are impossible to predict ahead of time! 3/

Jeff Clune profil fotoğrafı

Jeff Clune2 yıl önce

We evaluate IGE on a variety of text-based tasks that require search and exploration, with impressive results. For example, in Game of 24, a mathematical reasoning problem, IGE achieves 100% success rate 70.8% faster 📈 than classic graph search. 4/

Jeff Clune profil fotoğrafı

Jeff Clune2 yıl önce

On the hard TextWorld Coin Collector problem, a classic text-agent benchmark, IGE succeeds where prior SOTA FM agents completely fail. It also outperforms on medium and easier problems. 5/

Jeff Clune profil fotoğrafı

Jeff Clune2 yıl önce

IGE combines the tremendous strengths of foundation models with the powerful Go-Explore algorithm, opening up a new frontier of research into creating more generally capable agents with strong exploration capabilities. We're excited to see what new problems this could solve! 6/

Jeff Clune profil fotoğrafı

Jeff Clune2 yıl önce

IGE was driven by @cong_ml, in collaboration with @shengranhu and myself. 👩‍💻All code is open-sourced at: 🌐Website: 📄Paper: 7/7

Abel TM profil fotoğrafı

Abel TM2 yıl önce

@cong_ml Bigger environments? NetHack?

Jeff Clune profil fotoğrafı

Jeff Clune2 yıl önce

@cong_ml Coming soon to a theater near you!

Ω.KendrickPlumard profil fotoğrafı

Ω.KendrickPlumard2 yıl önce

@cong_ml Really cool work!

Sumon Ahmed profil fotoğrafı

Sumon Ahmed2 yıl önce

@cong_ml Highly informative

Benzer Videolar

[ 💌 143OFC ] We are so excited to introduce Yes My Love, the rising star in the world of P-Pop! ❤️‍🔥 Are you ready to learn their stories and get to know them better? 🤔 — Go dive into YMLFlix and explore what awaits! 💫 🔗 #YMLFlixRelease #143YesMyLove

[ 💌 143OFC ] We are so excited to introduce Yes My Love, the rising star in the world of P-Pop! ❤️‍🔥 Are you ready to learn their stories and get to know them better? 🤔 — Go dive into YMLFlix and explore what awaits! 💫 🔗 #YMLFlixRelease #143YesMyLove

143 Official

15,541 görüntüleme • 3 yıl önce

🎥🐼 Say hello to giant pandas Bao Li and Qing Bao in our new series: Prepping for Pandas! Go behind the scenes as we introduce you to our new animal ambassadors, look back at 52 years of conservation and explore their connection to D.C. #DCPandas

🎥🐼 Say hello to giant pandas Bao Li and Qing Bao in our new series: Prepping for Pandas! Go behind the scenes as we introduce you to our new animal ambassadors, look back at 52 years of conservation and explore their connection to D.C. #DCPandas

National Zoo

51,248 görüntüleme • 1 yıl önce

Real-time world model interaction meets space exploration. Star Junkers is a space scavenger RPG — a drifter girl, a loyal weasel, and a ship barely holding together. Explore uncharted planets, uncover hidden treasure, take on risky jobs, and trade your way to a better ship. The exploration is enhanced by a real-time world model, adding a layer of unpredictability to what you find. With an ever-expanding universe of UGC planets, there's always somewhere new to go. Built on Yoroll

Real-time world model interaction meets space exploration. Star Junkers is a space scavenger RPG — a drifter girl, a loyal weasel, and a ship barely holding together. Explore uncharted planets, uncover hidden treasure, take on risky jobs, and trade your way to a better ship. The exploration is enhanced by a real-time world model, adding a layer of unpredictability to what you find. With an ever-expanding universe of UGC planets, there's always somewhere new to go. Built on Yoroll

Yoroll

47,443 görüntüleme • 5 ay önce

Today, I'm incredibly excited to announce Default's Series A led by 8VC and our new product. We spent the last three years building the tools and orchestration layer companies like Owner and Airbyte have used to build infrastructure across their go-to-market functions. Today, we're launching the first real-time data layer for go-to-market, a powerful new revenue agent, and a suite of tools for revenue teams and their agents. It's Day 1.

Today, I'm incredibly excited to announce Default's Series A led by 8VC and our new product. We spent the last three years building the tools and orchestration layer companies like Owner and Airbyte have used to build infrastructure across their go-to-market functions. Today, we're launching the first real-time data layer for go-to-market, a powerful new revenue agent, and a suite of tools for revenue teams and their agents. It's Day 1.

nico

141,669 görüntüleme • 3 ay önce

⚡️BREAKING Iran's FM after holding talks with the Americans: 'We agreed on guiding principles to begin drafting text of a potential deal Good progress was made, the path ahead is clearer. Drafting will be hard work, but the path has started Delegations will now go back to their countries for consultation'

⚡️BREAKING Iran's FM after holding talks with the Americans: 'We agreed on guiding principles to begin drafting text of a potential deal Good progress was made, the path ahead is clearer. Drafting will be hard work, but the path has started Delegations will now go back to their countries for consultation'

Iran Observer

164,000 görüntüleme • 6 ay önce

$1/ I am excited to announce the new Playground: Design graphics like a Pro. You can make T-shirts, Logos, Social Media Posts, and even invent new kinds of memes. Powered by our next image foundation model Playground v3 (beta). Go to playground.\com. No wait list.$

1/ I am excited to announce the new Playground: Design graphics like a Pro. You can make T-shirts, Logos, Social Media Posts, and even invent new kinds of memes. Powered by our next image foundation model Playground v3 (beta). Go to playground.\com. No wait list.

Suhail

1,688,590 görüntüleme • 2 yıl önce

Super excited to share 🧠MLGym 🦾 – the first Gym environment for AI Research Agents 🤖🔬 We introduce MLGym and MLGym-Bench, a new framework and benchmark for evaluating and developing LLM agents on AI research tasks. The key contributions of our work are: 🕹️ Enables the exploration of different training algorithms for AI Research Agents such as RL 🛠️ Provides a flexible evaluation framework that can accommodate different artifacts such as models, algorithms, or predictions 🤖 Allows researchers to evaluate any model without the need to develop a custom agentic harness 🎯 Introduces 13 diverse open-ended AI Research tasks for evaluating AI Research Agents on a wide range of domains such as computer vision, natural language processing, reinforcement learning, game theory, and logical reasoning. 📈 Proposes a new evaluation metric for AI Research Agents MLGym makes it easy to: 1) Add new tasks 2) Evaluate new models 3) Integrate new agents Check out a video of the MLGym Agent to see how it performs the full pipeline of idea generation💡, implementation 👩‍💻, experimentation 👩‍🔬, and iteration 🔄 to improve on ML tasks. Huge thanks to the exceptionally talented Deepak Nathani who led this work and to all the other amazing collaborators who made this possible 🙏🫶🚀

Super excited to share 🧠MLGym 🦾 – the first Gym environment for AI Research Agents 🤖🔬 We introduce MLGym and MLGym-Bench, a new framework and benchmark for evaluating and developing LLM agents on AI research tasks. The key contributions of our work are: 🕹️ Enables the exploration of different training algorithms for AI Research Agents such as RL 🛠️ Provides a flexible evaluation framework that can accommodate different artifacts such as models, algorithms, or predictions 🤖 Allows researchers to evaluate any model without the need to develop a custom agentic harness 🎯 Introduces 13 diverse open-ended AI Research tasks for evaluating AI Research Agents on a wide range of domains such as computer vision, natural language processing, reinforcement learning, game theory, and logical reasoning. 📈 Proposes a new evaluation metric for AI Research Agents MLGym makes it easy to: 1) Add new tasks 2) Evaluate new models 3) Integrate new agents Check out a video of the MLGym Agent to see how it performs the full pipeline of idea generation💡, implementation 👩‍💻, experimentation 👩‍🔬, and iteration 🔄 to improve on ML tasks. Huge thanks to the exceptionally talented Deepak Nathani who led this work and to all the other amazing collaborators who made this possible 🙏🫶🚀

Roberta Raileanu

105,115 görüntüleme • 1 yıl önce

What happens when millions of AI agents start negotiating, transacting, and delegating to one another? Nenad Tomasev joined our podcast with Hannah Fry to explore the rise of agentic economies – and how we can diversify agent decision-making to avoid AI groupthink. Timecodes: 00:00 Intro 1:07 Defining AI agents 4:44 Agentic exploration in science and research 15:46 Delegation between agents 22:46 Agentic security and traps 29:31 Building an agentic economy 33:22 Cognitive monoculture 36:29 Distributed intelligence

What happens when millions of AI agents start negotiating, transacting, and delegating to one another? Nenad Tomasev joined our podcast with Hannah Fry to explore the rise of agentic economies – and how we can diversify agent decision-making to avoid AI groupthink. Timecodes: 00:00 Intro 1:07 Defining AI agents 4:44 Agentic exploration in science and research 15:46 Delegation between agents 22:46 Agentic security and traps 29:31 Building an agentic economy 33:22 Cognitive monoculture 36:29 Distributed intelligence

Google DeepMind

116,168 görüntüleme • 2 ay önce

.CapIX Dev's vision is to bring quantum computing to retail The three layers underneath: - Layer 1: GPU mesh aggregating Surplus Intelligence, Pod the Squire, OpenRouter - Layer 2: Smart router (heavy tasks go to the GPU, light tasks go to the CPU, quantum computing in the future?) - Layer 3: Community cloud with end-to-end flows Ruqaiyah built it solo with an 86-agent dev The whole EasyA Kickstart stack is forming under the $CPX umbrella - ARTHUR LABS builds the 3D model - Bit Agents handles the commerce - Covenant Foundation governs the agents - Factory in China prints the physical product - All running on CapIX compute

.CapIX Dev's vision is to bring quantum computing to retail The three layers underneath: - Layer 1: GPU mesh aggregating Surplus Intelligence, Pod the Squire, OpenRouter - Layer 2: Smart router (heavy tasks go to the GPU, light tasks go to the CPU, quantum computing in the future?) - Layer 3: Community cloud with end-to-end flows Ruqaiyah built it solo with an 86-agent dev The whole EasyA Kickstart stack is forming under the $CPX umbrella - ARTHUR LABS builds the 3D model - Bit Agents handles the commerce - Covenant Foundation governs the agents - Factory in China prints the physical product - All running on CapIX compute

MCG

13,184 görüntüleme • 2 ay önce

We're extremely excited to announce our $15.6M Series A led by a16z! We're building the AI communications platform for logistics. Go check out our new website and book a demo: You might be surprised who calls you back ;)

We're extremely excited to announce our $15.6M Series A led by a16z! We're building the AI communications platform for logistics. Go check out our new website and book a demo: You might be surprised who calls you back ;)

HappyRobot

97,158 görüntüleme • 1 yıl önce

People with high conscientiousness tend to get better grades in school and earn more money—even when you control for IQ. So if you have two people with the same score on an intelligence test, the person with higher conscientiousness will likely go on to get better grades and have higher earnings. There's also a very small but significant inverse correlation between intelligence and conscientiousness. In other words, the higher you go in intelligence, the lower your conscientiousness score tends to be. It's a very small effect, but it seems to be driven largely by the orderliness aspect of conscientiousness. Some psychologists have suggested there's a compensatory mechanism at work here. Very intelligent people can afford to be less organized because they're able to use their intelligence to compensate for their lack of diligence and focus. Meanwhile, people who are relatively less intelligent are forced to be more orderly to compensate for their lack of intellectual aptitude. Relatedly, men with spouses who score higher on conscientiousness tend to earn more money, even when their own level of conscientiousness is controlled for.

People with high conscientiousness tend to get better grades in school and earn more money—even when you control for IQ. So if you have two people with the same score on an intelligence test, the person with higher conscientiousness will likely go on to get better grades and have higher earnings. There's also a very small but significant inverse correlation between intelligence and conscientiousness. In other words, the higher you go in intelligence, the lower your conscientiousness score tends to be. It's a very small effect, but it seems to be driven largely by the orderliness aspect of conscientiousness. Some psychologists have suggested there's a compensatory mechanism at work here. Very intelligent people can afford to be less organized because they're able to use their intelligence to compensate for their lack of diligence and focus. Meanwhile, people who are relatively less intelligent are forced to be more orderly to compensate for their lack of intellectual aptitude. Relatedly, men with spouses who score higher on conscientiousness tend to earn more money, even when their own level of conscientiousness is controlled for.

Rob Henderson

46,865 görüntüleme • 9 ay önce

🚨NEW: Elon Musk explains his vision for future space exploration, mass drivers, and harnessing the power of the sun. "In order to understand the universe, you must explore the universe. We're only using roughly 1% of the potential energy of Earth. The only way to access the energy of the sun is to expand beyond Earth. We then need to extend to the galaxy and other galaxies. The next step beyond Earth datacenters are Earth Orbital datacenters. It will be 100-200 GW/year. We see a path to 1 TW/year. By being on the moon, building AI satellites, and having a real mass driver on the moon, we can go to 1000GW+ year. You could even work your way up to a few % of the sun's energy. It's difficult to imagine what an intelligence of that scale would think about but it will be exciting to see what happens. I want to see a mass driver on the moon go SHOOM SHOOM! One after the other, shooting satellites into space. I can't think of anything more exciting. Maybe we'll meet aliens. Maybe we'll see remnants of ancient civilizations. But the only way we're going to do that is to get out there and explore."

🚨NEW: Elon Musk explains his vision for future space exploration, mass drivers, and harnessing the power of the sun. "In order to understand the universe, you must explore the universe. We're only using roughly 1% of the potential energy of Earth. The only way to access the energy of the sun is to expand beyond Earth. We then need to extend to the galaxy and other galaxies. The next step beyond Earth datacenters are Earth Orbital datacenters. It will be 100-200 GW/year. We see a path to 1 TW/year. By being on the moon, building AI satellites, and having a real mass driver on the moon, we can go to 1000GW+ year. You could even work your way up to a few % of the sun's energy. It's difficult to imagine what an intelligence of that scale would think about but it will be exciting to see what happens. I want to see a mass driver on the moon go SHOOM SHOOM! One after the other, shooting satellites into space. I can't think of anything more exciting. Maybe we'll meet aliens. Maybe we'll see remnants of ancient civilizations. But the only way we're going to do that is to get out there and explore."

Autism Capital 🧩

13,432,803 görüntüleme • 6 ay önce

The SIX Digital Exchange (SDX) in Switzerland is using a stablecoin for their wholesale Central Bank Digital Currency (CBDC) pilot to explore the use of private money as a settlement asset for tokenized transactions. The Swiss National Bank (SNB) is considering a private Swiss franc token money that would be backed by sight deposits at the central bank, aiming to have a risk profile comparable to central bank money. This approach allows for the exploration of using stablecoins, which are privately-issued digital currencies, as a potential solution for settling financial transactions. By utilizing stablecoins (Private money) in their CBDC pilot, SIX aims to enhance transaction settlement efficiency and explore innovative payment solutions within the financial landscape.

Mr. Man

24,860 görüntüleme • 2 yıl önce

SilkAI is live. Try it now: 🔥 The first version of our intelligent web experience is ready for you to explore, directly in your browser, no downloads needed. You can: • Chat with intelligent AI agents • Create and train your own characters • Explore a new layer of interactive AI Built from scratch with love, obsession, and way too many late nights, this is your moment to dive in. And behind the scenes? We’re already working on something wilder: real-world robotics. Talk. Create. Go deep. Try it free. Let SilkAI surprise you.

SilkAI is live. Try it now: 🔥 The first version of our intelligent web experience is ready for you to explore, directly in your browser, no downloads needed. You can: • Chat with intelligent AI agents • Create and train your own characters • Explore a new layer of interactive AI Built from scratch with love, obsession, and way too many late nights, this is your moment to dive in. And behind the scenes? We’re already working on something wilder: real-world robotics. Talk. Create. Go deep. Try it free. Let SilkAI surprise you.

NetVRk

28,908 görüntüleme • 1 yıl önce

LLM agents have demonstrated promise in their ability to automate computer tasks, but face challenges with multi-step reasoning and planning. Towards addressing this, we propose an inference-time tree search algorithm for LLM agents to explicitly perform exploration and multi-step planning in interactive web environments. It is the first tree search algorithm for LLM agents that shows effectiveness on realistic and complex web environments: on the challenging VisualWebArena benchmark, applying our search algorithm on top of a GPT-4o agent yields a 39.7% relative increase in success rate compared to the same baseline without search, setting a state-of-the-art success rate of 26.4%. On WebArena, search also yields a 28.0% relative improvement over a baseline agent, setting a competitive success rate of 19.2%.

LLM agents have demonstrated promise in their ability to automate computer tasks, but face challenges with multi-step reasoning and planning. Towards addressing this, we propose an inference-time tree search algorithm for LLM agents to explicitly perform exploration and multi-step planning in interactive web environments. It is the first tree search algorithm for LLM agents that shows effectiveness on realistic and complex web environments: on the challenging VisualWebArena benchmark, applying our search algorithm on top of a GPT-4o agent yields a 39.7% relative increase in success rate compared to the same baseline without search, setting a state-of-the-art success rate of 26.4%. On WebArena, search also yields a 28.0% relative improvement over a baseline agent, setting a competitive success rate of 19.2%.

Jing Yu Koh

124,894 görüntüleme • 2 yıl önce

🎮 Computer Use Agent Arena is LIVE! 🚀 🔥 Easiest way to test computer-use agents in the wild without any setup 🌟 Compare top VLMs: OpenAI Operator, Claude 3.7, Gemini 2.5 Pro, Qwen 2.5 vl and more 🕹️ Test agents on 100+ real apps & webs with one-click config 🔒 Safe & free access on cloud-hosted machines Page: Leaderboard (tentative): Blog: Data & Code (coming soon): ⭐️Why Computer Agent Arena? 1️⃣Beyond Static Benchmarks: We use computers to perform enormous tasks and workflows every day, and AI agents have the potential to automate these tasks. However, existing benchmarks are very limited (e.g., only 369 tasks in OSWorld and 812 tasks in WebArena). To better measure their capabilities, we introduce Computer Agent Arena for users to easily compare & test AI agents on all kinds of crowdsourced real-world computer use tasks. 2️⃣Cloud Testing, Simplified: As agents like OpenAI’s Operator and Claude 3.7 sonnet release, users face configuration challenges and privacy hurdles to deploy on their own computers. Our platform integrates these agents with cloud-hosted machines, providing users with quick and secure access. 3️⃣Unified Embodied Digital Environment: Unlike Chatbot Arena, we provide users with a real embodied environment—computers—where all agents are grounded in real computer tasks and environments. Led by @XLANG_Lab [1/🧵]

🎮 Computer Use Agent Arena is LIVE! 🚀 🔥 Easiest way to test computer-use agents in the wild without any setup 🌟 Compare top VLMs: OpenAI Operator, Claude 3.7, Gemini 2.5 Pro, Qwen 2.5 vl and more 🕹️ Test agents on 100+ real apps & webs with one-click config 🔒 Safe & free access on cloud-hosted machines Page: Leaderboard (tentative): Blog: Data & Code (coming soon): ⭐️Why Computer Agent Arena? 1️⃣Beyond Static Benchmarks: We use computers to perform enormous tasks and workflows every day, and AI agents have the potential to automate these tasks. However, existing benchmarks are very limited (e.g., only 369 tasks in OSWorld and 812 tasks in WebArena). To better measure their capabilities, we introduce Computer Agent Arena for users to easily compare & test AI agents on all kinds of crowdsourced real-world computer use tasks. 2️⃣Cloud Testing, Simplified: As agents like OpenAI’s Operator and Claude 3.7 sonnet release, users face configuration challenges and privacy hurdles to deploy on their own computers. Our platform integrates these agents with cloud-hosted machines, providing users with quick and secure access. 3️⃣Unified Embodied Digital Environment: Unlike Chatbot Arena, we provide users with a real embodied environment—computers—where all agents are grounded in real computer tasks and environments. Led by @XLANG_Lab [1/🧵]

Bowen Wang

93,200 görüntüleme • 1 yıl önce

Did you know that suborbital research can serve orbital, lunar and mars missions? 🌝 In the latest episode of #AustinExplains, we go behind the scenes at Virgin Galactic to explore how their spacecraft enables a new generation o microgravity research. With Redwire technology onboard, researchers can run experiments in suborbital microgravity labs accelerating discoveries that will support future space exploration missions. Check it out! 🚀

Did you know that suborbital research can serve orbital, lunar and mars missions? 🌝 In the latest episode of #AustinExplains, we go behind the scenes at Virgin Galactic to explore how their spacecraft enables a new generation o microgravity research. With Redwire technology onboard, researchers can run experiments in suborbital microgravity labs accelerating discoveries that will support future space exploration missions. Check it out! 🚀

Redwire

16,981 görüntüleme • 5 ay önce

Haven't been to a conference in a while, really excited to be at #NeurIPS2024! I'll be helping present 4 of our group's recent papers: 1. Overcoming the Sim-to-Real Gap: Leveraging Simulation to Learn to Explore for Real-World RL 2. Distributional Successor Features Enable Zero-Shot Policy Optimization 3. Learning to Cooperate with Humans using Generative Agents 4. Personalizing Reinforcement Learning from Human Feedback with Variational Preference Learning Find more details on each paper and where to find us in this thread (1/6)

Haven't been to a conference in a while, really excited to be at #NeurIPS2024! I'll be helping present 4 of our group's recent papers: 1. Overcoming the Sim-to-Real Gap: Leveraging Simulation to Learn to Explore for Real-World RL 2. Distributional Successor Features Enable Zero-Shot Policy Optimization 3. Learning to Cooperate with Humans using Generative Agents 4. Personalizing Reinforcement Learning from Human Feedback with Variational Preference Learning Find more details on each paper and where to find us in this thread (1/6)

Abhishek Gupta

10,803 görüntüleme • 1 yıl önce