Stephen McAleese, at Less Wrong, used scaling rules and formulas to try to predict the performance of GPT-4. He underpredicted the capabilities of GPT-4. However, he was not widely wrong. GPT-4 came out earlier than he expected and had better performance than expected. Multi-task Language Understanding on MMLU was better than expected. GPT-4 can work …
The OpenAI GPT4 can score higher than 90% of law students writing the bar exam while the old version (GPT3) was in the bottom 10% of human bar exam test takers. GPT-4 can output 25000 words. GPT-4 can write a higher quality novel while GPT3.5 could only output a very short story. GPT-4 can score …
Ark Invest forecasts a massive drop in the cost to train generative AI models and a massive increase in the size of AI models. AI models will increase to hundreds of trillions of parameters. The limitation will not be the training cost. The limitation will be the training data and those with better and proprietary …
The three main ingredients leading to predictable improvements in AI performance are training data, computation, and improved algorithms. In the mid-2010s, some AI researchers noticed that larger AI systems were consistently smarter, and so they theorized that the most important ingredient in AI performance might be the total budget for AI training computation. When this …
Dr Alan D. Thompson is a world expert in artificial intelligence (AI), specializing in the augmentation of human intelligence, and advancing the evolution of ‘integrated AI’. He is predicting Artificial General Intelligence will arrive in about 35 months. Alan provides AI consulting and advisory to intergovernmental organizations including member states of the United Nations, Non-Aligned …
The BaGuaLu AI system used the Chinese Sunway exaflop supercomputer to train the largest AI model with over 174 trillion parameters. The miraculous capabilities of neural net AI systems like ChatGPT (AI generate novel text and stories) and Dall-E (AI generate novel pictures) and Alphafold2 (protein folding) comes from the growth of the AI models. …
Demis Hassabis is the CEO and co-founder of DeepMind and he had a great talk with Lex Fridman about AI. DeepMind beat the best human players with Go and they also made the best AI for chess. DeepMind also created AlphaFold 2 which has solved protein folding. As advanced mathematics and Calculus were critical to …
Arxiv – Deepmind introduces GATO, a generalist AI agent, which could be a path to AGI, Artificial General Intelligence. Inspired by progress in large-scale language modeling, Deepmind apply a similar approach towards building a single generalist agent beyond the realm of text outputs. The agent, which Deepmind refer to as Gato, works as a multi-modal, …
Elon Musk believes 2029 is the year we can achieve both AGI (Artificial General Intelligence) and humans on Mars. Clearly, Elon believes Teslabot will be used to achieve realworld AGI and realworld AI. He also believes that SpaceX Super Heavy Starship will be used to place humans on Mars. 2029 feels like a pivotal year. …
The Cofounder and CTO of AI Unicorn has said that the Tesla the Factory-Bot is a done deal. The case is made that FSD AI is sufficient for a Tesla factory bot. Tesla FSD AI can already be trained to tell the difference between different parts in a factory. The arms and leg actuators have …
Elon expects a pretty good prototype Teslabot this year and a low volume production for Teslabot will start next year. Elon also says Full self-driving safer than human driving and the fully rapidly reusable orbital SpaceX Starship should happen this year. SOURCES- Dave Lee Investing Brian WangBrian Wang is a Futurist Thought Leader and a …
