• How AI improves power grids:
-it accelerates future energy use calculations to optimize how energy is dispatched, from 10 minutes to 60 seconds
-it predicts individual home energy needs
-it recommends optimal charging times to owners of electric vehicles
-it accelerates the process of identifying potential infrastructure issues and suggests where to trim trees or fix faulty equipment. - MIT Tech Review
• EU researchers admit it’s too soon to know how exactly AI will impact employment and wages, but current data shows that AI-exposed jobs have seen an increase instead of a drop, and that it boosts high-skill and junior-level jobs with minimal effect on medium-skilled jobs – however, the data suggests AI might have a negative effect on wages. - ECB
• Pretty amazing acceleration of text-to-image generation from Stability – SDXL Turbo generates images instantly as you type the prompt, thanks to a new technique that reduces generation steps from 50 to 1, as usual the model is openly available but only for non-commercial use. - Stability, Clipdrop (test the model here)
• Amazon announced Q – a chatbot for workplace use, for up to $25/month, it can help you with AWS, integrates with apps like Slack, connects with over 40 enterprise systems like Salesforce or Microsoft 365, and is expected to be adopted by developers and cloud administrators. - CNBC
• GPT-4 is good at processing radiology reports, including tasks like disease classification and summarization – comparable to experienced radiologists, it could help beyond radiology by making medical reports more understandable, but it requires more research and clinical trials. - Microsoft
• Algorithmic trading company XTX Markets launched a new $10mn challenge fund, the Artificial Intelligence Mathematical Olympiad Prize, offering a grand prize of $5mn to the first publicly-shared AI model to compete at the level of a math olympiad gold medalist. - AIMO
• The current success of Nvidia in AI was 15+ years in the making:
-Jensen Huang’s contribution to AI started with the CUDA platform back in 2006, which enabled the biggest breakthroughs in AI since then, including AI’s “Big Bang moment”, the AlexNet image recognition model
-with CUDA, Huang wanted to democratize supercomputing, but it took years for the market to catch up with his vision
-now, Nvidia is overwhelmed with requests for their hardware, with their $500,000 DGX H100 platform on backorder for months, and their gross profit margin is nearly 70%
-Nvidia is literally powering the AI revolution, because the more compute you give to neural nets, the greater their capabilities
-interesting quote from Ilya Sutskever: “If you allow yourself to believe that an artificial neuron is like a biological neuron, then it’s like you’re training brains. They should do everything we can do”
-Huang predicts that AI will gain reasoning capabilities and start figuring things out by itself in two to three years
-his next big bet is Omniverse, an “industrial metaverse”, or an extremely detailed simulation / digital twin of the world. - Jensen Huang’s profile by the New Yorker
