Links for 2024-07-26
AI achieves silver-medal standard solving International Mathematical Olympiad problems
The fact that the program can come up with a non-obvious construction like this is very impressive, and well beyond what I thought was state of the art.
— Prof Sir Timothy Gowers, IMO gold medalist and Fields Medal winner
I imagine that when people saw Sputnik overhead in 1957, they might have had the same feeling that I do now.
We (human civilization) need to move to high alert, and to act!!
…this AI breakthrough is totally different from #GPT being able to do standardized tests through pattern-matching. It strikes at the heart of discovery.
…the sheer fact that it was able to solve the problems at all is a major advance. The hard part of solving these problems isn’t calculation. It’s inventing a solution pathway. Most people would get 0 points even if they had a year to think.
— Po-Shen Loh, mathematician and former IMO coach https://www.facebook.com/poshenloh/posts/pfbid02t4JRLNxBw87YwyKx8nd4qWrF3Y6BfSxNSpPUrCu1P7J9GV8eBVue415wEDUJAAxXl
The system couples a pre-trained language model with the AlphaZero reinforcement learning algorithm, which previously taught itself to master games like chess, shogi and Go. Each success is used to reinforce its neural network, making it better at tackling subsequent, harder problems.
Read more: https://deepmind.google/discover/blog/ai-solves-imo-problems-at-silver-medal-level/
AI:
OpenAI: “We’re testing SearchGPT, a temporary prototype of new AI search features that give you fast and timely answers with clear and relevant sources.” https://openai.com/index/searchgpt-prototype/
Sam Altman: “AI progress will be immense from here, and AI will be a critical national security issue. Who will control the future of AI? A democratic vision for artificial intelligence must prevail over an authoritarian one.” https://www.washingtonpost.com/opinions/2024/07/25/sam-altman-ai-democracy-authoritarianism-future/ [no paywall: https://archive.is/Uaj2V]
OpenDevin: An Open Platform for AI Software Developers as Generalist Agents https://arxiv.org/abs/2407.16741
Probabilistic Inference in Language Models via Twisted Sequential Monte Carlo https://arxiv.org/abs/2404.17546
Large Language Models can produce responses to debate questions posed in a UK political TV-debate show that were judged as more authentic and relevant than the original responses from the people who were impersonated. https://arxiv.org/abs/2407.12855
IBM proposes a new training-free approach to mitigate hallucination in LLMs. https://arxiv.org/abs/2407.16908
A Playbook for AI Policy https://manhattan.institute/article/a-playbook-for-ai-policy
Miscellaneous:
NASA: "Look closely 👀 These “leopard spots” on a Martian rock are clues pointing to possibly the best signs of ancient microbial life we’ve found yet on Mars. To know for sure, we need to study the rock in labs on Earth." https://www.jpl.nasa.gov/news/nasas-perseverance-rover-scientists-find-intriguing-mars-rock
"Not a single athlete from South Asia appears in the top 3,949 men’s 100m times, despite this region comprising almost 25% of the world’s population." https://www.aporiamagazine.com/p/are-africans-better-runners
Another demonstration of how difficult it is to successfully defend against drones. You have seconds to aim and shoot, and if it crashes within 10 meters, the shrapnel will kill or maim you anyway. https://x.com/GloOouD/status/1816353141031309384