As a former materials physicist, I am genuinely unsure of what to make of this superconductivity paper.
🧵 👇
Ted Sanders
861 posts
Research at OpenAI. Be kind to others, and yourself.
- Last year, Sam Bankman-Fried scammed me into writing an essay, by dangling a $1.5M essay contest that disappeared along with FTX. My loss is your gain. We just published 114 pages on why we think transformative* AGI in the next 20 years is <1% likely.
- Just dropped: updated guide for Q&A with GPT and embeddings-based search. Discusses: - Why search is better than fine-tuning - What types of questions this fails on - How to troubleshoot & improve your system If you have ideas to improve, I'll do them.
- Replying to @sanderstedIn favor: the scientists look legit, the graphs look legit, the materials system looks legit (related to cuprates), the video shows one sample halfway levitating, and, most speculatively, they just published a three-author hype paper (Nobel prize maxes out at 3 winners per).
- Replying to @sanderstedAgainst: They didn’t even measure a Tc (because their PPMS maxed out at 400 K and apparently they didn’t bother to use an oven??), they relatedly didn’t measure a Meissner effect, and they published this three months ago with little fanfare. Can fear of scooping explain this?
- Replying to @sanderstedThis paper examines the way of thinking and limitations of physicists regarding the phenomenon of superconductivity and outlines how room-temperature and ambient-pressure superconductors can be developed through the statistical thermodynamic background of the liquid state theory.
- Replying to @sanderstedLastly, if you had discovered room-temperature ambient-pressure superconductivity, would this really be the first sentence of your paper:
- Replying to @sandersted*in hindsight, not actually sure how related to cuprates this material is. i was just comforted by the stalwart copper and oxygen in the chemical formula. i’ve been out of the complex oxides game for a while now (although i did write the wikipedia article on the subject).
- Replying to @sanderstedupdate: my best guess at this point is that the measurements are real but misinterpreted. maybe a metal-insulator transition/breakdown in a normally diamagnetic insulator. would really love to see meissner effect measurements etc. that don't rely on bulk conductivity measurements
- Replying to @sanderstedLike, really? In any case, I’m curious to see where this all goes. April paper: koreascience.or.kr/article/JAKO20…
- Replying to @mattparlmerpaper probably isn’t correct, but you can probably make the same arguments about high Tc superconductors or blue LEDs. a lot of geniuses toiling for years by baking powders in ovens. the combinatorial recipe space is huge, and it can take weeks or months per material.
- Jotted down a list of some interesting prompting libraries and prompting papers in a new section of the OpenAI Cookbook. Any ones you like that we're still missing?
- AGI is hard to define. my preferred definition of AGI is a computer system that can can accomplish a task impossible for 100 human geniuses working together, such as publishing a blog post with a single canonical spelling of GPT-4o / gpt-4o / gpt4o

