This week:
- I just finished reading The MANIAC, a fantastic historical fiction account of science and technology - from physics to John von Neumann to artificial intelligence. It ended with a story of Lee Sedol and the AI program AlphaGo in 2016.
- Now we get a view of the aftermath of how players have evolved in Go-playing - "Were the moves getting better because some players cheated by using the AI? No. They really were getting better."
- On the UI side, I just came across ScreenAI from Google, released last month. Followed by this paper noting that thus far, models "fall short in their ability to comprehend and interact effectively with ... screens" - releasing Ferret-UI, with "enhanced understanding of mobile UI screens, equipped with referring, grounding, and reasoning capabilities". Personally - this is one of the more exciting frontiers for AI integration into our day-to-day - its ability to interact (beyond search) with our existing internet infrastructure.
- Sequoia and Forbes publish the AI 50: Companies of the Future, calling out:
- "Workflow automation platform ServiceNow is achieving case avoidance rates of nearly 20% with their AI-powered Now Assist.
- Palo Alto Networks has reduced the cost of processing expenses with AI.
- Hubspot has scaled customer support with AI.
- And Swedish fintech Klarna recently announced over $40 million in run-rate savings by building AI into their customer support."
- Intel releases the Intel Gaudi 3 AI accelerator promising "significant performance improvements for training and inference tasks" versus NVIDIA.
- Apple's M4 chip will be optimized for AI capabilities.
- Google launches CodeGemma and Gemma 1.1 with performance improvements.
- As for a paper I am spending some time with this week - TimeGPT-1 from last year - "the first foundation model for time series, capable of generating accurate predictions for diverse datasets not seen during training"