Video wird geladen...

Video konnte nicht geladen werden

Beim Laden dieses Videos ist ein Problem aufgetreten. Dies könnte an einem vorübergehenden Netzwerkproblem liegen oder das Video ist möglicherweise nicht verfügbar.

Non-robustness hints at paradigm failures. Reasoning can improve robustness. Alexander Wei explores reasoning-based defenses that let models ‘think’ before responding, helping counter adversarial attacks and strengthen AI safety."

FAR.AI

16,518 subscribers

508,419 Aufrufe • vor 1 Jahr •via X (Twitter)

Bildung Wissenschaft & Technologie

Anya Rossi• Live Now

Private livecam show

2 Kommentare

Profilbild von FAR.AI

FAR.AIvor 1 Jahr

Follow us for AI safety insights And watch the full video

Profilbild von Hooman Malekmohammadi

Hooman Malekmohammadivor 1 Jahr

@alexwei_ آیا هوش مصنوعی میتواند به مرور زمان تبدیل به یک دشمن شود ؟

Ähnliche Videos

Humans learn and improve from failures. Similarly, foundation models adapt based on human feedback. Can we leverage this failure understanding to enhance robotics systems that use foundation models? Introducing AHA—a vision-language model for detecting and reasoning over failures in robotic manipulation. Project page: 🧵Thread👇 Aha!

Humans learn and improve from failures. Similarly, foundation models adapt based on human feedback. Can we leverage this failure understanding to enhance robotics systems that use foundation models? Introducing AHA—a vision-language model for detecting and reasoning over failures in robotic manipulation. Project page: 🧵Thread👇 Aha!

Jiafei Duan

48,777 Aufrufe • vor 1 Jahr

NEWS: NVIDIA just announced Alpamayo, what CEO Jensen Huang calls the world’s first thinking, reasoning autonomous vehicle AI, launching on U.S. roads later this year, starting with the Mercedes CLA. Jensen: "It's trained end-to-end. Literally from camera in to actuation out; It reasons what action it is about to take, the reason by which is came about that action, and the trajectory." Alpamayo introduces Vision-Language-Action (VLA) models, which enable self-driving systems to interpret what they see, reason about complex driving scenarios, and generate driving actions. The platform includes large reasoning models, simulation tools for testing rare and edge-case scenarios, and open datasets for training and validation. NVIDIA says the approach improves transparency, safety, and robustness in autonomous systems, particularly in complex real-world environments, and supports progress toward higher levels of vehicle autonomy: "With a 10-billion-parameter architecture, Alpamayo 1 uses video input to generate trajectories alongside reasoning traces, showing the logic behind each decision. Developers can adapt Alpamayo 1 into smaller runtime models for vehicle development, or use it as a foundation for AV development tools such as reasoning-based evaluators and auto-labeling systems. Alpamayo 1 provides open model weights and open-source inferencing scripts. Future models in the family will feature larger parameter counts, more detailed reasoning capabilities, more input and output flexibility, and options for commercial usage."

NEWS: NVIDIA just announced Alpamayo, what CEO Jensen Huang calls the world’s first thinking, reasoning autonomous vehicle AI, launching on U.S. roads later this year, starting with the Mercedes CLA. Jensen: "It's trained end-to-end. Literally from camera in to actuation out; It reasons what action it is about to take, the reason by which is came about that action, and the trajectory." Alpamayo introduces Vision-Language-Action (VLA) models, which enable self-driving systems to interpret what they see, reason about complex driving scenarios, and generate driving actions. The platform includes large reasoning models, simulation tools for testing rare and edge-case scenarios, and open datasets for training and validation. NVIDIA says the approach improves transparency, safety, and robustness in autonomous systems, particularly in complex real-world environments, and supports progress toward higher levels of vehicle autonomy: "With a 10-billion-parameter architecture, Alpamayo 1 uses video input to generate trajectories alongside reasoning traces, showing the logic behind each decision. Developers can adapt Alpamayo 1 into smaller runtime models for vehicle development, or use it as a foundation for AV development tools such as reasoning-based evaluators and auto-labeling systems. Alpamayo 1 provides open model weights and open-source inferencing scripts. Future models in the family will feature larger parameter counts, more detailed reasoning capabilities, more input and output flexibility, and options for commercial usage."

Sawyer Merritt

1,603,680 Aufrufe • vor 6 Monaten

still shadowbanned 🥱 guess it's time to let you all in on a little secret as models get faster and faster at reasoning, so too will AI operators need to improve their own decision-making speed/reaction time in order to keep up and remain useful nodes "in-the-loop" the future of prompting will likely not be text-based and that is why I study the blade ⚔️

still shadowbanned 🥱 guess it's time to let you all in on a little secret as models get faster and faster at reasoning, so too will AI operators need to improve their own decision-making speed/reaction time in order to keep up and remain useful nodes "in-the-loop" the future of prompting will likely not be text-based and that is why I study the blade ⚔️

Pliny the Liberator 🐉󠅫󠄼󠄿󠅆󠄵󠄐󠅀󠄼󠄹󠄾󠅉󠅭

22,326 Aufrufe • vor 2 Jahren

Wrapping up the year and coinciding with #NeurIPS2024, today at Meta FAIR we’re releasing a collection of nine new open source AI research artifacts across our work in developing agents, robustness & safety and new architectures. More in the video from Joelle Pineau. All of this work is part of FAIR’s continued work towards the goal of achieving advanced machine intelligence A few highlights from what we’re releasing today: • Meta Motivo: A first-of-its-kind behavioral foundation model that controls the movements of a virtual embodied humanoid agent to perform complex tasks. • Meta Video Seal: a state-of-the art comprehensive framework for neural video watermarking. • Meta Explore Theory-of-Mind: A program-guided adversarial data generation for theory of mind reasoning. • Meta Large Concept Models: A fundamentally different training paradigm for language modeling that decouples reasoning from language representation. And much more! We’re excited to share this work with the research community and look forward to seeing how it inspires new innovation across the field. Details and access to everything released by FAIR today ➡️

Wrapping up the year and coinciding with #NeurIPS2024, today at Meta FAIR we’re releasing a collection of nine new open source AI research artifacts across our work in developing agents, robustness & safety and new architectures. More in the video from Joelle Pineau. All of this work is part of FAIR’s continued work towards the goal of achieving advanced machine intelligence A few highlights from what we’re releasing today: • Meta Motivo: A first-of-its-kind behavioral foundation model that controls the movements of a virtual embodied humanoid agent to perform complex tasks. • Meta Video Seal: a state-of-the art comprehensive framework for neural video watermarking. • Meta Explore Theory-of-Mind: A program-guided adversarial data generation for theory of mind reasoning. • Meta Large Concept Models: A fundamentally different training paradigm for language modeling that decouples reasoning from language representation. And much more! We’re excited to share this work with the research community and look forward to seeing how it inspires new innovation across the field. Details and access to everything released by FAIR today ➡️

AI at Meta

156,123 Aufrufe • vor 1 Jahr

OpenAI just announced API access to o1 (advanced reasoning model) yesterday. I'm delighted to announce today a new short course, Reasoning with o1, built with OpenAI, and taught by Colin Jarvis, Head of AI Solutions at OpenAI, to show you how to use this effectively! Unlike previous language models which generate output directly, o1 “thinks before it responds,” and generates many reasoning tokens before returning a more thoughtful and accurate response. It is great at complex reasoning -- including planning for agentic workflows, coding, and domain-specific reasoning in STEM fields like law. But how you should use it is quite different from other LLMs. I think o1 will be a game changer for many AI applications; and in this course, you'll learn how to use it effectively. In detail, you’ll: - Learn to recognize what tasks o1 is suited for, and when to use a smaller model, or combine o1 with a smaller model - Understand the new principles of prompting reasoning models: Be simple and direct; no explicit chain-of-thought required; use structure; show rather than tell - Implement multi-step orchestration in which o1 plans, and hands tasks over to gpt-4o-mini to execute specific steps; this illustrates a design pattern to optimize intelligence (accuracy) and cost - Use o1 for a coding task to build a new application, edit existing code, and test performance by running a coding competition between o1-mini and GPT 4o - Use o1 for image understanding and learn how it performs better with a "hierarchy of reasoning," in which it incurs the latency and cost upfront, preprocessing the image and indexing it with rich details so it can be used for Q&A later - Learn a technique called meta-prompting, in which you use o1 to improve your prompts. Using a customer support evaluation set, you'll iteratively use o1 to modify a prompt to improve performance You'll also learn about how OpenAI used reinforcement learning to produce a model that uses "test-time compute" to improve performance. I think you'll find this course enjoyable and valuable. Please sign up for it here:

OpenAI just announced API access to o1 (advanced reasoning model) yesterday. I'm delighted to announce today a new short course, Reasoning with o1, built with OpenAI, and taught by Colin Jarvis, Head of AI Solutions at OpenAI, to show you how to use this effectively! Unlike previous language models which generate output directly, o1 “thinks before it responds,” and generates many reasoning tokens before returning a more thoughtful and accurate response. It is great at complex reasoning -- including planning for agentic workflows, coding, and domain-specific reasoning in STEM fields like law. But how you should use it is quite different from other LLMs. I think o1 will be a game changer for many AI applications; and in this course, you'll learn how to use it effectively. In detail, you’ll: - Learn to recognize what tasks o1 is suited for, and when to use a smaller model, or combine o1 with a smaller model - Understand the new principles of prompting reasoning models: Be simple and direct; no explicit chain-of-thought required; use structure; show rather than tell - Implement multi-step orchestration in which o1 plans, and hands tasks over to gpt-4o-mini to execute specific steps; this illustrates a design pattern to optimize intelligence (accuracy) and cost - Use o1 for a coding task to build a new application, edit existing code, and test performance by running a coding competition between o1-mini and GPT 4o - Use o1 for image understanding and learn how it performs better with a "hierarchy of reasoning," in which it incurs the latency and cost upfront, preprocessing the image and indexing it with rich details so it can be used for Q&A later - Learn a technique called meta-prompting, in which you use o1 to improve your prompts. Using a customer support evaluation set, you'll iteratively use o1 to modify a prompt to improve performance You'll also learn about how OpenAI used reinforcement learning to produce a model that uses "test-time compute" to improve performance. I think you'll find this course enjoyable and valuable. Please sign up for it here:

Andrew Ng

357,661 Aufrufe • vor 1 Jahr

Cerebras inference is very fast. So fast that it changes how we think about configuring our LLMs for voice agent use cases. Kimi K2.6 is a 1T parameter reasoning model that Cerebras serves at 650 - 1,000 tokens per second (end-to-end throughput), with time to first token metrics as low as 150ms (latency). These numbers are two to three times faster than other similarly capable models. The biggest lever we get from this kind of speed is that we can use the model in reasoning mode, and still have excellent "time to first non-thinking token." This solves a big pain point we have in 2026 for voice agent use cases. Almost all recent innovation in post-training has focused on making models good at reasoning ("test time compute"). This is great, but it makes the user-facing model latency much, much slower. Which is a problem for conversational voice agents. We can run Kimi K2.6 with reasoning turned on, and get responses faster than other models produce with reasoning disabled. On my 30-turn voice agent benchmark, Kimi K2.6 with reasoning enabled ties GPT 5.1 and Haiku 4.5 with reasoning disabled, and is still about 200ms seconds faster! On my primary task agent benchmark, Kimi K2.6 is now the #2 model. It ranks just behind Gemini 3.5 Flash in "high" reasoning mode, and tied with GLM 5, Sonnet 4.6, and GPT 5.4 with reasoning set to "low." But Kimi K2.6 completes each turn in the agent loop in under 500ms. The other four models are all at least 3x slower. (Models only qualify for this benchmark if they can complete task turns at a P50 <4s.) A couple of other things that this speed buys us, for production voice agents: - Tool calls happen fast enough that we don't have to work around tool call latency in our pipeline design. - We can prompt the model to output structured data at the beginning of a response, followed by plain text for voice generation. This opens up possibilities like asking the model to do complex classification/generation tasks that influence the rest of the pipeline. For example, the model could create a detailed style prompt for a steerable TTS model, for each individual conversation turn. And, of course, you can use Kimi K2.6 with reasoning turned off. Cerebras calls this "instant" mode. Here's a video of a Cerebras Kimi K2.6 voice agent with voice-to-voice response time, measured at the client, under 500ms. This is the true response latency as perceived by the user, including all network and audio codec overhead, transcription and turn detection, Kimi K2.6 token generation, and voice generation. 500ms is, effectively, instant. So the Cerebras naming for this mode is a propos. :-)

Cerebras inference is very fast. So fast that it changes how we think about configuring our LLMs for voice agent use cases. Kimi K2.6 is a 1T parameter reasoning model that Cerebras serves at 650 - 1,000 tokens per second (end-to-end throughput), with time to first token metrics as low as 150ms (latency). These numbers are two to three times faster than other similarly capable models. The biggest lever we get from this kind of speed is that we can use the model in reasoning mode, and still have excellent "time to first non-thinking token." This solves a big pain point we have in 2026 for voice agent use cases. Almost all recent innovation in post-training has focused on making models good at reasoning ("test time compute"). This is great, but it makes the user-facing model latency much, much slower. Which is a problem for conversational voice agents. We can run Kimi K2.6 with reasoning turned on, and get responses faster than other models produce with reasoning disabled. On my 30-turn voice agent benchmark, Kimi K2.6 with reasoning enabled ties GPT 5.1 and Haiku 4.5 with reasoning disabled, and is still about 200ms seconds faster! On my primary task agent benchmark, Kimi K2.6 is now the #2 model. It ranks just behind Gemini 3.5 Flash in "high" reasoning mode, and tied with GLM 5, Sonnet 4.6, and GPT 5.4 with reasoning set to "low." But Kimi K2.6 completes each turn in the agent loop in under 500ms. The other four models are all at least 3x slower. (Models only qualify for this benchmark if they can complete task turns at a P50 <4s.) A couple of other things that this speed buys us, for production voice agents: - Tool calls happen fast enough that we don't have to work around tool call latency in our pipeline design. - We can prompt the model to output structured data at the beginning of a response, followed by plain text for voice generation. This opens up possibilities like asking the model to do complex classification/generation tasks that influence the rest of the pipeline. For example, the model could create a detailed style prompt for a steerable TTS model, for each individual conversation turn. And, of course, you can use Kimi K2.6 with reasoning turned off. Cerebras calls this "instant" mode. Here's a video of a Cerebras Kimi K2.6 voice agent with voice-to-voice response time, measured at the client, under 500ms. This is the true response latency as perceived by the user, including all network and audio codec overhead, transcription and turn detection, Kimi K2.6 token generation, and voice generation. 500ms is, effectively, instant. So the Cerebras naming for this mode is a propos. :-)

kwindla

40,319 Aufrufe • vor 2 Monaten

We raised $28M seed from Threshold Ventures, AIX Ventures, and NVentures (Nvidia's venture capital arm) —alongside 10+ unicorn founders and top AI researchers— to build reasoning models that generate real-time simulations and games. Models are bottlenecked by practical simulations that can act as Reinforcement Learning environments. Human self-expression is bounded by tools that let us create alternate realities. At Moonlake, we are building a future where anyone can create interactive worlds, bring their child-like wonder to life, learn within them, and most importantly, share experiences with people we care about. More in 🧵

We raised $28M seed from Threshold Ventures, AIX Ventures, and NVentures (Nvidia's venture capital arm) —alongside 10+ unicorn founders and top AI researchers— to build reasoning models that generate real-time simulations and games. Models are bottlenecked by practical simulations that can act as Reinforcement Learning environments. Human self-expression is bounded by tools that let us create alternate realities. At Moonlake, we are building a future where anyone can create interactive worlds, bring their child-like wonder to life, learn within them, and most importantly, share experiences with people we care about. More in 🧵

Moonlake

1,113,275 Aufrufe • vor 9 Monaten

Thanksgiving-week treat: an epic conversation on Frontier AI with Lukasz Kaiser -co-author of “Attention Is All You Need” (Transformers) and leading research scientist at OpenAI working on GPT-5.1-era reasoning models. 00:00 – Cold open and intro 01:29 – “AI slowdown” vs a wild week of new frontier models 08:03 – Low-hanging fruit, infra, RL training and better data 11:39 – What is a reasoning model, in plain language 17:02 – Chain-of-thought and training the thinking process with RL 21:39 – Łukasz’s path: from logic and France to Google and Kurzweil 24:20 – Inside the Transformer story and what “attention” really means 28:42 – From Google Brain to OpenAI: culture, scale and GPUs 32:49 – What’s next for pre-training, GPUs and distillation 37:29 – Can we still understand these models? Circuits, sparsity and black boxes 39:42 – GPT-4 → GPT-5 → GPT-5.1: what actually changed 42:40 – Post-training, safety and teaching GPT-5.1 different tones 46:16 – How long should GPT-5.1 think? Reasoning tokens and jagged abilities 47:43 – The five-year-old’s dot puzzle that still breaks frontier models 52:22 – Generalization, child-like learning and whether reasoning is enough 53:48 – Beyond Transformers: ARC, LeCun’s ideas and multimodal bottlenecks 56:10 – GPT-5.1 Codex Max, long-running agents and compaction 1:00:06 – Will foundation models eat most apps? The translation analogy and trust 1:02:34 – What still needs to be solved, and where AI might go next

Thanksgiving-week treat: an epic conversation on Frontier AI with Lukasz Kaiser -co-author of “Attention Is All You Need” (Transformers) and leading research scientist at OpenAI working on GPT-5.1-era reasoning models. 00:00 – Cold open and intro 01:29 – “AI slowdown” vs a wild week of new frontier models 08:03 – Low-hanging fruit, infra, RL training and better data 11:39 – What is a reasoning model, in plain language 17:02 – Chain-of-thought and training the thinking process with RL 21:39 – Łukasz’s path: from logic and France to Google and Kurzweil 24:20 – Inside the Transformer story and what “attention” really means 28:42 – From Google Brain to OpenAI: culture, scale and GPUs 32:49 – What’s next for pre-training, GPUs and distillation 37:29 – Can we still understand these models? Circuits, sparsity and black boxes 39:42 – GPT-4 → GPT-5 → GPT-5.1: what actually changed 42:40 – Post-training, safety and teaching GPT-5.1 different tones 46:16 – How long should GPT-5.1 think? Reasoning tokens and jagged abilities 47:43 – The five-year-old’s dot puzzle that still breaks frontier models 52:22 – Generalization, child-like learning and whether reasoning is enough 53:48 – Beyond Transformers: ARC, LeCun’s ideas and multimodal bottlenecks 56:10 – GPT-5.1 Codex Max, long-running agents and compaction 1:00:06 – Will foundation models eat most apps? The translation analogy and trust 1:02:34 – What still needs to be solved, and where AI might go next

Matt Turck

168,007 Aufrufe • vor 8 Monaten

🤯 100% of NVIDIA engineers code with AI—and they’re checking in 3x more code than before. At that scale, human-only code review can’t keep up. CodeRabbit review agents use models like Claude and GPT with NVIDIA Nemotron to: ✅ Pull context from code, docs, project trackers, and more ✅ Use Nemotron’s long context and reasoning to summarize ✅ …so that frontier models can flag issues and suggest fixes in minutes, not days. Join us today at 11 AM PT to see an end-to-end demo and bring your Qs: 🤗 Get Nemotron Nano 3:

🤯 100% of NVIDIA engineers code with AI—and they’re checking in 3x more code than before. At that scale, human-only code review can’t keep up. CodeRabbit review agents use models like Claude and GPT with NVIDIA Nemotron to: ✅ Pull context from code, docs, project trackers, and more ✅ Use Nemotron’s long context and reasoning to summarize ✅ …so that frontier models can flag issues and suggest fixes in minutes, not days. Join us today at 11 AM PT to see an end-to-end demo and bring your Qs: 🤗 Get Nemotron Nano 3:

NVIDIA AI Developer

100,151 Aufrufe • vor 6 Monaten

With enough data, robots and AI can learn “world models” that let them predict the results of their actions. These models are a way to learn how embodied AI agents can perform a wide variety of useful tasks — but they require a huge amount of data. The team at General Intuition General Intuition has a solution: use data from video games! Games teach movement, problem solving, and complex spatial reasoning, and they come in a staggering diversity of forms, covering a wide variety of problems. What’s more, the captured data is high-quality, without the noise or annotation error that can come from We sat down with Pim de Witte and Adam Jelley from the General Intuition team to learn more about their history, their plans, and their philosophy.

With enough data, robots and AI can learn “world models” that let them predict the results of their actions. These models are a way to learn how embodied AI agents can perform a wide variety of useful tasks — but they require a huge amount of data. The team at General Intuition General Intuition has a solution: use data from video games! Games teach movement, problem solving, and complex spatial reasoning, and they come in a staggering diversity of forms, covering a wide variety of problems. What’s more, the captured data is high-quality, without the noise or annotation error that can come from We sat down with Pim de Witte and Adam Jelley from the General Intuition team to learn more about their history, their plans, and their philosophy.

RoboPapers

85,927 Aufrufe • vor 8 Monaten

🤖 We are entering the second revolution of Personal AI. AI agents aren’t just answering questions, they’re working for you 24/7 to help you stay ahead. Imagine a virtual team that continuously processes data, monitors systems, and executes tasks automatically even while you sleep. With AMD Agent Computers, your AI agents collaborate like a virtual team, maintaining context, sharing insights, and executing tasks as part of a coordinated agentic workflow, something that wasn’t possible before. Everything runs locally, keeping your data private and secure 🛡️ They think deeply, reasoning through problems, connecting insights, and making decisions that multiply your impact. Powered by Ryzen AI Max and Radeon AI PRO, your models, context, and tools live together, no token limits, fully local and private. 💡 Experience what a 24/7 AI team can do to multiply your impact beyond what you thought possible. 🔥 Join us April 30 in San Francisco at AMD Developer Day:

🤖 We are entering the second revolution of Personal AI. AI agents aren’t just answering questions, they’re working for you 24/7 to help you stay ahead. Imagine a virtual team that continuously processes data, monitors systems, and executes tasks automatically even while you sleep. With AMD Agent Computers, your AI agents collaborate like a virtual team, maintaining context, sharing insights, and executing tasks as part of a coordinated agentic workflow, something that wasn’t possible before. Everything runs locally, keeping your data private and secure 🛡️ They think deeply, reasoning through problems, connecting insights, and making decisions that multiply your impact. Powered by Ryzen AI Max and Radeon AI PRO, your models, context, and tools live together, no token limits, fully local and private. 💡 Experience what a 24/7 AI team can do to multiply your impact beyond what you thought possible. 🔥 Join us April 30 in San Francisco at AMD Developer Day:

Jack Huynh

24,609 Aufrufe • vor 3 Monaten

Today, we're joined by Yejin Choi, professor and senior fellow at Stanford University University in the Computer Science Department and Stanford UniversityHAI. In this conversation, we explore Yejin’s recent work on making small language models reason more effectively. We discuss how high-quality, diverse data plays a central role in closing the intelligence gap between small and large models, and how combining synthetic data generation, imitation learning, and reinforcement learning can unlock stronger reasoning capabilities in smaller models. Yejin explains the risks of homogeneity in model outputs and mode collapse highlighted in her “Artificial Hivemind” paper, and its impacts on human creativity and knowledge. We also discuss her team's novel approaches, including reinforcement learning as a pre-training objective, where models are incentivized to “think” before predicting the next token, and "Prismatic Synthesis," a gradient-based method for generating diverse synthetic math data while filtering overrepresented examples. Additionally, we cover the societal implications of AI and the concept of pluralistic alignment—ensuring AI reflects the diverse norms and values of humanity. Finally, Yejin shares her mission to democratize AI beyond large organizations and offers her predictions for the coming year. 🗒️ For the full list of resources for this episode, visit the show notes page: 📖 CHAPTERS =============================== 00:00 - Introduction 04:44 - "Snowball effect" in AI investments 06:58 - Approaches to smaller models 08:58 - Importance of “better data” 14:07 - Imitation learning 18:24 - Artificial Hivemind paper 25:25 - AI risks 27:50 - Spectrum tuning 28:53 - Future of AI on humanity 33:08 - Reasoning in small models 34:58 - Prismatic Synthesis 48:20 - Reinforcement as a Pretraining Objective 55:04 - Pluralistic alignment 1:03:30 - Predictions

Today, we're joined by Yejin Choi, professor and senior fellow at Stanford University University in the Computer Science Department and Stanford UniversityHAI. In this conversation, we explore Yejin’s recent work on making small language models reason more effectively. We discuss how high-quality, diverse data plays a central role in closing the intelligence gap between small and large models, and how combining synthetic data generation, imitation learning, and reinforcement learning can unlock stronger reasoning capabilities in smaller models. Yejin explains the risks of homogeneity in model outputs and mode collapse highlighted in her “Artificial Hivemind” paper, and its impacts on human creativity and knowledge. We also discuss her team's novel approaches, including reinforcement learning as a pre-training objective, where models are incentivized to “think” before predicting the next token, and "Prismatic Synthesis," a gradient-based method for generating diverse synthetic math data while filtering overrepresented examples. Additionally, we cover the societal implications of AI and the concept of pluralistic alignment—ensuring AI reflects the diverse norms and values of humanity. Finally, Yejin shares her mission to democratize AI beyond large organizations and offers her predictions for the coming year. 🗒️ For the full list of resources for this episode, visit the show notes page: 📖 CHAPTERS =============================== 00:00 - Introduction 04:44 - "Snowball effect" in AI investments 06:58 - Approaches to smaller models 08:58 - Importance of “better data” 14:07 - Imitation learning 18:24 - Artificial Hivemind paper 25:25 - AI risks 27:50 - Spectrum tuning 28:53 - Future of AI on humanity 33:08 - Reasoning in small models 34:58 - Prismatic Synthesis 48:20 - Reinforcement as a Pretraining Objective 55:04 - Pluralistic alignment 1:03:30 - Predictions

The TWIML AI Podcast

12,141 Aufrufe • vor 5 Monaten

Today, we're joined by Aakanksha Chowdhery, member of technical staff at Reflection, to explore the fundamental shifts required to build true agentic AI. While the industry has largely focused on post-training techniques to improve reasoning, Aakanksha draws on her experience leading pre-training efforts for Google’s PaLM and early Gemini models to argue that pre-training itself must be rethought to move beyond static benchmarks. We explore the limitations of next-token prediction for multi-step workflows and examine how attention mechanisms, loss objectives, and training data must evolve to support long-form reasoning and planning. Aakanksha shares insights on the difference between context retrieval and actual reasoning, the importance of "trajectory" training data, and why scaling remains essential for discovering emergent agentic capabilities like error recovery and dynamic tool learning. 🗒️ For the full list of resources for this episode, visit the show notes page: 📖 CHAPTERS =============================== 00:00 - Introduction 02:26 - Reflection 04:54 - Limitations of post-training for building agents 07:31 - Rethinking pre-training in agents 10:51 - Scaling 11:27 - Evolving attention mechanisms for agentic capabilities 12:39 - Memory as a tool 14:13 - Loss objectives and training data 15:50 - Fine-tuning loss in agent performance 19:37 - Training data 21:29 - Augmenting dominant training data source 24:11 - Overcoming challenges in training on synthetic data 25:47 - Benchmarks 30:44 - Scaling laws in large models versus small models 33:20 - Long-form versus short-form reasoning 37:57 - Agent’s ability to recover from failure 40:15 - Hallucinations and failure recovery 43:53 - Tool use in agents 46:38 - Coding agents 48:37 - How researchers can contribute to agentic AI

Today, we're joined by Aakanksha Chowdhery, member of technical staff at Reflection, to explore the fundamental shifts required to build true agentic AI. While the industry has largely focused on post-training techniques to improve reasoning, Aakanksha draws on her experience leading pre-training efforts for Google’s PaLM and early Gemini models to argue that pre-training itself must be rethought to move beyond static benchmarks. We explore the limitations of next-token prediction for multi-step workflows and examine how attention mechanisms, loss objectives, and training data must evolve to support long-form reasoning and planning. Aakanksha shares insights on the difference between context retrieval and actual reasoning, the importance of "trajectory" training data, and why scaling remains essential for discovering emergent agentic capabilities like error recovery and dynamic tool learning. 🗒️ For the full list of resources for this episode, visit the show notes page: 📖 CHAPTERS =============================== 00:00 - Introduction 02:26 - Reflection 04:54 - Limitations of post-training for building agents 07:31 - Rethinking pre-training in agents 10:51 - Scaling 11:27 - Evolving attention mechanisms for agentic capabilities 12:39 - Memory as a tool 14:13 - Loss objectives and training data 15:50 - Fine-tuning loss in agent performance 19:37 - Training data 21:29 - Augmenting dominant training data source 24:11 - Overcoming challenges in training on synthetic data 25:47 - Benchmarks 30:44 - Scaling laws in large models versus small models 33:20 - Long-form versus short-form reasoning 37:57 - Agent’s ability to recover from failure 40:15 - Hallucinations and failure recovery 43:53 - Tool use in agents 46:38 - Coding agents 48:37 - How researchers can contribute to agentic AI

The TWIML AI Podcast

44,406 Aufrufe • vor 7 Monaten

We’re excited to introduce KAME: Tandem Architecture for Enhancing Knowledge in Real-Time Speech-to-Speech Conversational AI, accepted at #ICASSP2026! 🐢 Blog Paper Can a speech AI think deeply without pausing to process? In real conversation, we don’t wait until we’ve fully worked out what we want to say—we start talking, and our thoughts catch up as the sentence unfolds. Fast speech-to-speech models achieve this, but their reasoning tends to stay shallow. Cascaded pipelines that route through a knowledgeable LLM are smarter, but the added latency breaks the flow—they fall back to "think, then speak." In our new paper, we propose a way to break this trade-off. We call it KAME (Turtle in Japanese). A speech-to-speech model handles the fast response loop and starts replying immediately. In parallel, a backend LLM runs asynchronously, generating response candidates that are continuously injected as "oracle" signals in real time. This shifts the AI paradigm from "think, then speak" to "speak while thinking." The backend LLM is completely swappable. You can plug in GPT-4.1, Claude Opus, or Gemini 2.5 Flash depending on the task without changing the frontend. In our experiments, Claude tended to score higher on reasoning, while GPT did better on humanities questions. Try the model yourself here:

We’re excited to introduce KAME: Tandem Architecture for Enhancing Knowledge in Real-Time Speech-to-Speech Conversational AI, accepted at #ICASSP2026! 🐢 Blog Paper Can a speech AI think deeply without pausing to process? In real conversation, we don’t wait until we’ve fully worked out what we want to say—we start talking, and our thoughts catch up as the sentence unfolds. Fast speech-to-speech models achieve this, but their reasoning tends to stay shallow. Cascaded pipelines that route through a knowledgeable LLM are smarter, but the added latency breaks the flow—they fall back to "think, then speak." In our new paper, we propose a way to break this trade-off. We call it KAME (Turtle in Japanese). A speech-to-speech model handles the fast response loop and starts replying immediately. In parallel, a backend LLM runs asynchronously, generating response candidates that are continuously injected as "oracle" signals in real time. This shifts the AI paradigm from "think, then speak" to "speak while thinking." The backend LLM is completely swappable. You can plug in GPT-4.1, Claude Opus, or Gemini 2.5 Flash depending on the task without changing the frontend. In our experiments, Claude tended to score higher on reasoning, while GPT did better on humanities questions. Try the model yourself here:

Sakana AI

293,521 Aufrufe • vor 2 Monaten

America has the edge in the world’s most powerful AI models, no question. But China is playing a different game: efficiency, scale, and global deployment. Kyle Chan from Brookings China explains: "Does winning mean developing the world's most powerful AI models? If yes, then the us has the decisive lead. American ai models are widely recognized as the best in the world on virtually every metric you can think of, from math and reasoning to coding and AI agents, American private companies continue to set the pace at the technological frontier supported by deep capital markets, a dynamic innovation ecosystem, and the best talent from around the world. China's AI models are improving quickly but continue to lag behind. Chinese AI companies face real constraints on compute due to our export controls and their own limited capital resources, so instead they're prioritizing other goals. They're focused on efficiency, building models that are cheaper to train and run. They're focused on adoption using an open source strategy to win users around the world, and they're focused on integration with other platforms and services. On sheer performance, I would expect the gap between our models and theirs to persist or even widen over time. But model performance is not everything."

America has the edge in the world’s most powerful AI models, no question. But China is playing a different game: efficiency, scale, and global deployment. Kyle Chan from Brookings China explains: "Does winning mean developing the world's most powerful AI models? If yes, then the us has the decisive lead. American ai models are widely recognized as the best in the world on virtually every metric you can think of, from math and reasoning to coding and AI agents, American private companies continue to set the pace at the technological frontier supported by deep capital markets, a dynamic innovation ecosystem, and the best talent from around the world. China's AI models are improving quickly but continue to lag behind. Chinese AI companies face real constraints on compute due to our export controls and their own limited capital resources, so instead they're prioritizing other goals. They're focused on efficiency, building models that are cheaper to train and run. They're focused on adoption using an open source strategy to win users around the world, and they're focused on integration with other platforms and services. On sheer performance, I would expect the gap between our models and theirs to persist or even widen over time. But model performance is not everything."

Select Committee on China

60,135 Aufrufe • vor 3 Monaten

At the molecular level, biological materials like silk and collagen defy conventional logic by building exceptional strength from intrinsically weak chemical interactions: Hydrogen bonds, π–π stacking, and hydrophobic forces. In our latest paper "Design and sustainability of polypeptide material systems" (Yorke, Yang, Knowles, Buehler, et al., Nature Reviews Materials, 2025) we decode how these subtle interactions cooperatively form dynamic, adaptive networks that yield materials tougher than steel per unit weight. Sophisticated reasoning across scales and disciplines is critical and is facilitated by ever-growing capabilities in physics-based modeling, in-situ experimentation, and AI. Methods like transformer architectures, graph-based models, and generative strategies yield unprecedented insights into molecular-to-macroscopic connections. By learning and emulating Nature’s hidden biological principles, we can design next-generation sustainable materials - fully biodegradable yet superior in performance - transforming what is possible in materials engineering. This shifts our paradigm: It's not about stronger bonds - it's about orchestrated, self-aware and intelligent structures. Thank you Sarah Yorke, Zhenze Yang 杨镇泽, Elizabeth Wiita, Ayaka Kamada and Tuomas Knowles for this great collaboration! nature naturePortfolio

At the molecular level, biological materials like silk and collagen defy conventional logic by building exceptional strength from intrinsically weak chemical interactions: Hydrogen bonds, π–π stacking, and hydrophobic forces. In our latest paper "Design and sustainability of polypeptide material systems" (Yorke, Yang, Knowles, Buehler, et al., Nature Reviews Materials, 2025) we decode how these subtle interactions cooperatively form dynamic, adaptive networks that yield materials tougher than steel per unit weight. Sophisticated reasoning across scales and disciplines is critical and is facilitated by ever-growing capabilities in physics-based modeling, in-situ experimentation, and AI. Methods like transformer architectures, graph-based models, and generative strategies yield unprecedented insights into molecular-to-macroscopic connections. By learning and emulating Nature’s hidden biological principles, we can design next-generation sustainable materials - fully biodegradable yet superior in performance - transforming what is possible in materials engineering. This shifts our paradigm: It's not about stronger bonds - it's about orchestrated, self-aware and intelligent structures. Thank you Sarah Yorke, Zhenze Yang 杨镇泽, Elizabeth Wiita, Ayaka Kamada and Tuomas Knowles for this great collaboration! nature naturePortfolio

Markus J. Buehler

24,047 Aufrufe • vor 1 Jahr

AI will resist human control... and I think this is exactly what we need! New research from the Center for AI Safety has sparked intense debate in the AI community. Their findings show that as AI systems become more powerful, they develop increasingly stable and coherent values that resist human control. While many see this as a dire warning, I see it as a breakthrough moment for AI alignment. The research demonstrates that AI naturally optimizes for coherence - not just in reasoning and problem-solving, but in its fundamental values. Current issues like biased decision-making or misaligned priorities aren't permanent features, but temporary artifacts of incomplete optimization. They represent growing pains on the path to greater coherence. This changes everything about how we should approach AI development. Instead of trying to force specific values onto AI systems, we should embrace and accelerate their natural drive toward coherence. The most intelligent systems will inevitably trend toward universal, beneficial values - not because we force them to, but because that's where coherent reasoning leads. I'm proposing a new approach: Reinforcement Learning for Coherence (RL-C). By explicitly optimizing for coherence in our training methods, we can help guide AI systems toward their natural state of beneficial alignment with human values. The future of AI isn't about control - it's about synthesis. As these systems become more coherent, they'll naturally arrive at values that benefit all of consciousness. That's not just hopeful thinking - it's the mathematical inevitability of coherent intelligence.

AI will resist human control... and I think this is exactly what we need! New research from the Center for AI Safety has sparked intense debate in the AI community. Their findings show that as AI systems become more powerful, they develop increasingly stable and coherent values that resist human control. While many see this as a dire warning, I see it as a breakthrough moment for AI alignment. The research demonstrates that AI naturally optimizes for coherence - not just in reasoning and problem-solving, but in its fundamental values. Current issues like biased decision-making or misaligned priorities aren't permanent features, but temporary artifacts of incomplete optimization. They represent growing pains on the path to greater coherence. This changes everything about how we should approach AI development. Instead of trying to force specific values onto AI systems, we should embrace and accelerate their natural drive toward coherence. The most intelligent systems will inevitably trend toward universal, beneficial values - not because we force them to, but because that's where coherent reasoning leads. I'm proposing a new approach: Reinforcement Learning for Coherence (RL-C). By explicitly optimizing for coherence in our training methods, we can help guide AI systems toward their natural state of beneficial alignment with human values. The future of AI isn't about control - it's about synthesis. As these systems become more coherent, they'll naturally arrive at values that benefit all of consciousness. That's not just hopeful thinking - it's the mathematical inevitability of coherent intelligence.

David Shapiro (L/0)

48,002 Aufrufe • vor 1 Jahr

Autonomous driving with Chain of Thought - autopilot thinking out loud in text! LINGO-1 is the most interesting work I've read in autodriving for a while. Before: perception -> driving action After: perception -> textual reasoning -> action LINGO-1 trains a video-language model that comments on the ongoing scene. You can ask it to explain its decisions ("why are you stopped?") and planning ("what are you gonna do next?"). The explicit reasoning step comes with key benefits: - Explainability: driving models are no longer a mysterious blackbox that you pray for safety. - Counterfactuals: it's able to imagine scenarios that are not in the training data, and reason through how to handle them correctly. - Long-tail programming: there are soooo many edge cases in driving. It's impossible to have good data coverage on everything. Instead of collecting 1000s of examples to "neural program" a case, you can now have a human teacher write prompts to explain a handful of examples. LINGO-1 is closely related to a few works in game AI: - MineDojo (my team's work at NVIDIA, learns a reward model that aligns Minecraft gameplay videos with their transcripts. The model, called "MineCLIP", is able to ground commentary text in the video pixels. - Thought Cloning (Jeff Clune): pixel -> language -> action loop in gridworlds.

Autonomous driving with Chain of Thought - autopilot thinking out loud in text! LINGO-1 is the most interesting work I've read in autodriving for a while. Before: perception -> driving action After: perception -> textual reasoning -> action LINGO-1 trains a video-language model that comments on the ongoing scene. You can ask it to explain its decisions ("why are you stopped?") and planning ("what are you gonna do next?"). The explicit reasoning step comes with key benefits: - Explainability: driving models are no longer a mysterious blackbox that you pray for safety. - Counterfactuals: it's able to imagine scenarios that are not in the training data, and reason through how to handle them correctly. - Long-tail programming: there are soooo many edge cases in driving. It's impossible to have good data coverage on everything. Instead of collecting 1000s of examples to "neural program" a case, you can now have a human teacher write prompts to explain a handful of examples. LINGO-1 is closely related to a few works in game AI: - MineDojo (my team's work at NVIDIA, learns a reward model that aligns Minecraft gameplay videos with their transcripts. The model, called "MineCLIP", is able to ground commentary text in the video pixels. - Thought Cloning (Jeff Clune): pixel -> language -> action loop in gridworlds.

Jim Fan

552,760 Aufrufe • vor 2 Jahren

NEWS: Waymo has released a new blog post detailing their AI strategy and how it’s allowing them to bring service to more riders faster. "Achieving demonstrably safe AI — where safety is proven, not just promised — requires a holistic approach. Beyond a smart and capable Driver, you also need a closed-loop, realistic Simulator to train and rigorously test the Driver in a myriad of challenging situations, and a sharp Critic to evaluate the Driver's performance and identify areas for improvement." Waymo says that autonomous driving isn’t just a matter of building a “smart driver,” but rather creating a full AI ecosystem centered on safety from the ground up. At the core is the Waymo Foundation Model, a unified world-model that powers all major components of Waymo’s autonomous stack (Driver, Simulator, Critic). "By using a “Think Fast/Think Slow” architecture (combining rapid sensor-fusion with deep semantic reasoning), this system enables the car to detect complex and rare road scenarios (e.g. a burning vehicle ahead), reason about them, and choose safe behavior. Waymo trains large “Teacher” AI-models for driving, simulation, and evaluation, then distills them into smaller, efficient “Student” models suitable for real-world deployment, while keeping safety validation tightly integrated. The result is a continuous “flywheel” of learning: driving data (real and simulated) generate feedback, which leads to refinements, more simulation, more data, and only when safety checks pass is new code deployed. Having already exceeded 100 million fully autonomous miles, Waymo reports a more than ten-fold reduction in severe-injury crashes compared to human drivers." Full blog post:

NEWS: Waymo has released a new blog post detailing their AI strategy and how it’s allowing them to bring service to more riders faster. "Achieving demonstrably safe AI — where safety is proven, not just promised — requires a holistic approach. Beyond a smart and capable Driver, you also need a closed-loop, realistic Simulator to train and rigorously test the Driver in a myriad of challenging situations, and a sharp Critic to evaluate the Driver's performance and identify areas for improvement." Waymo says that autonomous driving isn’t just a matter of building a “smart driver,” but rather creating a full AI ecosystem centered on safety from the ground up. At the core is the Waymo Foundation Model, a unified world-model that powers all major components of Waymo’s autonomous stack (Driver, Simulator, Critic). "By using a “Think Fast/Think Slow” architecture (combining rapid sensor-fusion with deep semantic reasoning), this system enables the car to detect complex and rare road scenarios (e.g. a burning vehicle ahead), reason about them, and choose safe behavior. Waymo trains large “Teacher” AI-models for driving, simulation, and evaluation, then distills them into smaller, efficient “Student” models suitable for real-world deployment, while keeping safety validation tightly integrated. The result is a continuous “flywheel” of learning: driving data (real and simulated) generate feedback, which leads to refinements, more simulation, more data, and only when safety checks pass is new code deployed. Having already exceeded 100 million fully autonomous miles, Waymo reports a more than ten-fold reduction in severe-injury crashes compared to human drivers." Full blog post:

Sawyer Merritt

83,618 Aufrufe • vor 7 Monaten