![DeepMind makes big jump toward interpreting LLMs with sparse autoencoders](https://cdn.statically.io/img/venturebeat.com/wp-content/uploads/2024/07/AI-black-box.jpg?w=350&h=175&crop=1)
![DeepMind makes big jump toward interpreting LLMs with sparse autoencoders](https://cdn.statically.io/img/venturebeat.com/wp-content/uploads/2024/07/AI-black-box.jpg?w=350&h=175&crop=1)
![Gen AI boosts individual creativity at the cost of collective diversity, study finds](https://cdn.statically.io/img/venturebeat.com/wp-content/uploads/2024/07/human-and-robots-writing-together.jpg?w=350&h=175&crop=1)
Gen AI boosts individual creativity at the cost of collective diversity, study finds
![Researchers develop technique to give robots “embodied reasoning” abilities](https://cdn.statically.io/img/venturebeat.com/wp-content/uploads/2024/07/robot-thinking.jpg?w=350&h=175&crop=1)
Researchers develop technique to give robots “embodied reasoning” abilities
![FlashAttention-3 unleashes the power of H100 GPUs for LLMs](https://cdn.statically.io/img/venturebeat.com/wp-content/uploads/2024/07/lightning-fast-GPU.jpg?w=350&h=175&crop=1)
FlashAttention-3 unleashes the power of H100 GPUs for LLMs
![Meta researchers distill System 2 thinking into LLMs, improving performance on complex reasoning](https://cdn.statically.io/img/venturebeat.com/wp-content/uploads/2024/07/Thinking-fast-and-slow.jpg?w=350&h=175&crop=1)
Meta researchers distill System 2 thinking into LLMs, improving performance on complex reasoning
![DeepMind’s PEER scales language models with millions of tiny experts](https://cdn.statically.io/img/venturebeat.com/wp-content/uploads/2024/07/mixture-of-millions-of-experts.jpg?w=350&h=175&crop=1)
DeepMind’s PEER scales language models with millions of tiny experts
![Enterprises embrace generative AI, but challenges remain](https://cdn.statically.io/img/venturebeat.com/wp-content/uploads/2024/07/DSC9647.jpg?w=350&h=175&crop=1)
Enterprises embrace generative AI, but challenges remain
![AI agent benchmarks are misleading, study warns](https://cdn.statically.io/img/venturebeat.com/wp-content/uploads/2024/07/AI-agents.jpg?w=350&h=175&crop=1)
AI agent benchmarks are misleading, study warns
![How AI Agents are changing software development](https://cdn.statically.io/img/venturebeat.com/wp-content/uploads/2024/07/ai-coding-assistant.jpg?w=350&h=175&crop=1)
How AI Agents are changing software development
![Alter3 is the latest GPT-4-powered humanoid robot](https://cdn.statically.io/img/venturebeat.com/wp-content/uploads/2024/06/Alter3-robot.jpg?w=350&h=175&crop=1)
Alter3 is the latest GPT-4-powered humanoid robot
![How Gradient created an open LLM with a million-token context window](https://cdn.statically.io/img/venturebeat.com/wp-content/uploads/2024/06/inifinite-tokens.jpg?w=350&h=175&crop=1)
How Gradient created an open LLM with a million-token context window
![OpenVLA is an open-source generalist robotics model](https://cdn.statically.io/img/venturebeat.com/wp-content/uploads/2024/06/robot-manipulating-object.jpg?w=350&h=175&crop=1)
OpenVLA is an open-source generalist robotics model
![New Transformer architecture could enable powerful LLMs without GPUs](https://cdn.statically.io/img/venturebeat.com/wp-content/uploads/2024/06/glowing-gpu.jpg?w=350&h=175&crop=1)
New Transformer architecture could enable powerful LLMs without GPUs
![What we know about Apple’s on-device AI](https://cdn.statically.io/img/venturebeat.com/wp-content/uploads/2024/02/nuneybits_Vector_art_of_an_apple_sits_on_a_motherboard_amidst_a_43cb7d0d-f63b-4fe1-8496-b2154f4b19fb-transformed.webp?w=350&h=175&crop=1)
What we know about Apple’s on-device AI
![Stanford study finds AI legal research tools prone to hallucinations](https://cdn.statically.io/img/venturebeat.com/wp-content/uploads/2024/06/robot-lawyer.jpg?w=350&h=175&crop=1)
Stanford study finds AI legal research tools prone to hallucinations
![How foundation agents can revolutionize AI decision-making in the real world](https://cdn.statically.io/img/venturebeat.com/wp-content/uploads/2024/06/robots-real-world.jpg?w=350&h=175&crop=1)
How foundation agents can revolutionize AI decision-making in the real world
![Meta and Google researchers’ new data curation method could transform self-supervised learning](https://cdn.statically.io/img/venturebeat.com/wp-content/uploads/2024/05/sea-of-data.jpg?w=350&h=175&crop=1)
Meta and Google researchers’ new data curation method could transform self-supervised learning
![Microsoft, Beihang release MoRA, an efficient LLM fine-tuning technique](https://cdn.statically.io/img/venturebeat.com/wp-content/uploads/2024/05/power-cube.jpg?w=350&h=175&crop=1)
Microsoft, Beihang release MoRA, an efficient LLM fine-tuning technique
![Microsoft’s vs. Apple’s AI computer strategies: Why Satya is winning (for now)](https://cdn.statically.io/img/venturebeat.com/wp-content/uploads/2023/11/nuneybits_A_neural_network_art_piece_depicting_the_contrast_bet_040309b7-fbda-4b61-b34e-702688bbe854-transformed.png?w=350&h=175&crop=1)
Microsoft’s vs. Apple’s AI computer strategies: Why Satya is winning (for now)
![Meta introduces Chameleon, a state-of-the-art multimodal model](https://cdn.statically.io/img/venturebeat.com/wp-content/uploads/2024/05/robot-chameleon.jpg?w=350&h=175&crop=1)
Meta introduces Chameleon, a state-of-the-art multimodal model
![How attention offloading reduces the costs of LLM inference at scale](https://cdn.statically.io/img/venturebeat.com/wp-content/uploads/2024/05/spaceship-light-speed.jpg?w=350&h=175&crop=1)
How attention offloading reduces the costs of LLM inference at scale
![Nvidia’s DrEureka outperforms humans in training robotics systems](https://cdn.statically.io/img/venturebeat.com/wp-content/uploads/2024/05/walking_globe.png?w=350&h=175&crop=1)
Nvidia’s DrEureka outperforms humans in training robotics systems
![Meta’s new multi-token prediction makes AI models up to 3X faster](https://cdn.statically.io/img/venturebeat.com/wp-content/uploads/2024/05/Comic_Book_Style_Robot_Typing.png?w=350&h=175&crop=1)
Meta’s new multi-token prediction makes AI models up to 3X faster
![DeepMind researchers discover impressive learning capabilities in long-context LLMs](https://cdn.statically.io/img/venturebeat.com/wp-content/uploads/2024/04/nuneybits_A_vibrant_purple_and_blue_data_visualization_with_an__ca7ae6be-04ea-491e-b2fc-f8be54f76028.webp?w=350&h=175&crop=1)
DeepMind researchers discover impressive learning capabilities in long-context LLMs
![Meta challenges transformer architecture with Megalodon LLM](https://cdn.statically.io/img/venturebeat.com/wp-content/uploads/2024/04/cfr0z3n_cybernetic_megalodon_shark_swimming_through_a_sea_of_gl_a8c3f098-8021-4f46-afdf-415f52e23e3d.png?w=350&h=175&crop=1)
Meta challenges transformer architecture with Megalodon LLM
![How LLMs are ushering in a new era of robotics](https://cdn.statically.io/img/venturebeat.com/wp-content/uploads/2024/04/robots-hanging-from-cables-in-an-office.png?w=350&h=175&crop=1)
How LLMs are ushering in a new era of robotics
![Google’s new technique gives LLMs infinite context](https://cdn.statically.io/img/venturebeat.com/wp-content/uploads/2024/01/The_universe_of_open_source_large_language_models_is_small_in_number-e1706200036796.png?w=350&h=175&crop=1)
Google’s new technique gives LLMs infinite context
![Sakana AI’s evolutionary algorithm discovers new architectures for generative models](https://cdn.statically.io/img/venturebeat.com/wp-content/uploads/2024/03/cfr0z3n_two_neon_colored_tubes_full_of_red_and_blue_fish_merge__da8f1986-9595-4d7e-bd1f-a8be824bca00.png?w=350&h=175&crop=1)
Sakana AI’s evolutionary algorithm discovers new architectures for generative models
![DeepMind and Stanford’s new robot control model follow instructions from sketches](https://venturebeat.com/https://venturebeat.com/wp-content/uploads/2024/03/Screenshot-2024-03-11-at-1.24.45 PM.png?w=350&h=175&crop=1)
DeepMind and Stanford’s new robot control model follow instructions from sketches
![State Dept-backed report provides action plan to avoid catastrophic AI risks](https://cdn.statically.io/img/venturebeat.com/wp-content/uploads/2023/04/VB_security-breach-padlock_3_1200x800.jpg?w=350&h=175&crop=1)
State Dept-backed report provides action plan to avoid catastrophic AI risks
![Why Meta’s V-JEPA model can be a big deal for real-world AI](https://cdn.statically.io/img/venturebeat.com/wp-content/uploads/2024/02/nuneybits_Vector_art_of_Microsoft_Windows_and_cloud_computing_i_cd799e9f-c949-42c9-97e6-4b857760e4ee.png?w=350&h=175&crop=1)
Why Meta’s V-JEPA model can be a big deal for real-world AI
![What is Apple’s generative AI strategy?](https://cdn.statically.io/img/venturebeat.com/wp-content/uploads/2024/02/nuneybits_Abstract_art_of_an_apple_sits_on_a_motherboard_amidst_79ee03f2-1f42-480c-b6ad-96894ce1233c-transformed.webp?w=350&h=175&crop=1)
What is Apple’s generative AI strategy?
![DeepMind’s GenEM uses LLMs to generate expressive behaviors for robots](https://cdn.statically.io/img/venturebeat.com/wp-content/uploads/2024/02/nuneybits_Abstract_painting_of_a_robot_waving_hello_to_a_busine_74f1a4ba-4c0f-4978-b79a-f82c0407ebfe-transformed.webp?w=350&h=175&crop=1)
DeepMind’s GenEM uses LLMs to generate expressive behaviors for robots
![Meta’s OK-Robot performs zero-shot pick-and-drop in unseen environments](https://venturebeat.com/https://venturebeat.com/wp-content/uploads/2024/01/Screenshot-2024-01-29-at-2.31.38 PM.png?w=350&h=175&crop=1)
Meta’s OK-Robot performs zero-shot pick-and-drop in unseen environments
![Beyond chatbots: The wide world of embeddings](https://cdn.statically.io/img/venturebeat.com/wp-content/uploads/2024/01/nuneybits_A_visualization_of_the_key_concepts_in_the_story_disp_39a15483-41b1-470c-b4fb-e9756850fe6e-transformed.webp?w=350&h=175&crop=1)
Beyond chatbots: The wide world of embeddings
![Stanford’s mobile ALOHA robot learns from humans to cook, clean, do laundry](https://cdn.statically.io/img/venturebeat.com/wp-content/uploads/2024/01/Screen-Shot-2024-01-05-at-1.40.51-PM.png?w=350&h=175&crop=1)
Stanford’s mobile ALOHA robot learns from humans to cook, clean, do laundry
![2023 was a great year for open-source LLMs](https://cdn.statically.io/img/venturebeat.com/wp-content/uploads/2023/12/nuneybits_A_pixel_art_visualization_showing_an_AI_safety_archit_cfecf0ba-f2a4-4e52-9dff-d7422dc97203.png?w=350&h=175&crop=1)
2023 was a great year for open-source LLMs
![UC Berkeley’s transformer-based robot control system generalizes to unseen environments](https://cdn.statically.io/img/venturebeat.com/wp-content/uploads/2023/12/Screen-grab-from-UC-Berkeley-humanoid-robot-1.png?w=350&h=175&crop=1)
UC Berkeley’s transformer-based robot control system generalizes to unseen environments
![New reinforcement learning method uses human cues to correct its mistakes](https://cdn.statically.io/img/venturebeat.com/wp-content/uploads/2023/12/424f0e10-d1a7-4094-a579-a24019a2394f.jpeg?w=350&h=175&crop=1)
New reinforcement learning method uses human cues to correct its mistakes
![New transformer architecture can make language models faster and resource-efficient](https://cdn.statically.io/img/venturebeat.com/wp-content/uploads/2023/12/nuneybits_Abstract_art_of_the_classic_and_new_transformer_block_9df8da79-f6cb-46ae-bd19-62485f061633-transformed.webp?w=350&h=175&crop=1)