Grok 4 had lower lmarena scores than I had projected based upon the amount AI training compute used. I had projections for Grok 3.5 which was renamed Grok 4. We do not have elo lmarena score for Grok 4 heavy. My old projection had assumed what will now be called Grok 5 was called Grok …
Figure AI videos of the Helix Humanoid Robot loading washing machines in a residential setting and sorting objects from a conveyor belt for industrial settings. Brian WangBrian Wang is a Futurist Thought Leader and a popular Science blogger with 1 million readers per month. His blog Nextbigfuture.com is ranked #1 Science News Blog. It covers …
XAI Grok 4 Benchmarks are showing it is the leading model. Humanity Last Exam at 35 and 45 for reasoning is a big improvement from about 21 for other top models. If these leaked Grok 4 benchmarks are correct, 95 AIME, 88 GPQA, 75 SWE-bench, then XAI has the most powerful model on the market. …
Reinforcement Pre-Training (RPT) is a new method for training large language models (LLMs) by reframing the standard task of predicting the next token in a sequence as a reasoning problem solved using reinforcement learning (RL). Unlike traditional RL methods for LLMs that need expensive human data or limited annotated data, RPT uses verifiable rewards based …
Cursor AI raised $900 million in Series C funding from Thrive, Accel, Andreessen Horowitz, and DST. Cursor has grown to over $500 million in ARR and is used by more than half of the Fortune 500, including NVIDIA, Uber, and Adobe. This scale will help us push the frontier of AI coding research. Brian WangBrian …
China has the second largest sovereign wealth fund in the world at $1.3 trillion and it has over $16 trillion in debt. Singapore also has both a large national debt and a large sovereign wealth fund. The Federal Reserve holds about $4.4 trillion in T-bills and bonds. The US already holds a lot of financial …
Making AI to 10 to 30 times more efficient for AI inference and getting more value from training will increase AI demand. Increased AI efficiency on training and inference will accelerate the improvement and usefulness of AI. Dr Know It All went over the DeepSeek paper and explains how they automated the Reinforcement Learning. AlphaZero …
OpenaI o3 sets new records in several key areas, particularly in reasoning, coding and mathematical problem-solving. It scores 75.7% on the semi-private eval in low-compute mode (for $20 per task in compute ) and 87.5% in high-compute mode (thousands of $ per task). It’s very expensive. It is not just brute force. These capabilities are …
XAI is releasing Grok 2 in beta. This will be about equal to OpenAI GPT4 artificial intelligence models. Elon Musk has said XAI is assembled 100,000 Nvidia H100 chips for a total of 400 Exaflops of compute to train Grok 3 by the end of this year. Grok 3 will be at about the power …
Llama 3.1 405B is the first openly available model that rivals the top AI models when it comes to state-of-the-art capabilities in general knowledge, steerability, math, tool use, and multilingual translation. With the release of the 405B model, Meta supercharges innovation—with unprecedented opportunities for growth and exploration. They believe the latest generation of Llama will …
If Large Language Models debate their answers they can reach better answers. A complementary approach to improve language responses where multiple language model instances propose and debate their individual responses and reasoning processes over multiple rounds to arrive at a common final answer. The findings indicate that this approach significantly enhances mathematical and strategic reasoning …