Links for 2021-12-03
For the first time, mathematicians have partnered with artificial intelligence to suggest and prove new mathematical theorems. — "A new paper in Nature details how machine learning was used to make significant new discoveries in pure mathematics by guiding the intuition of some of the world’s top mathematicians. This work is an example of how AI can accelerate scientific discovery by augmenting and assisting researchers in their fields." https://www.nature.com/articles/s41586-021-04086-x
Show Your Work: Scratchpads for Intermediate Computation with Language Models — “We show that huge language models (137B params!) can be trained to solve algorithmic tasks by “showing their work”---writing intermediate text to a scratchpad. This “scratchpad” technique even allows us to predict the execution of Python code.” https://arxiv.org/abs/2112.00114
A 2006 article about the hardest math course ever offered: 45 percent Jewish, 18 percent Asian, 100 percent male — “Each week, their heads huddled together, these students dedicate 30 to 50 hours to problem sets—proving significant theorems with only definitions to guide them...Writing one’s own textbook, which is basically what the students do, is not for everyone...It’s a game of “Survivor”: Outwit, Outplay, Outmath. Before the fifth Monday of the term, students who can’t seem to stay in the game start dropping like flies...Seventy started it, 20 finished it, and only 10 understood it...The final course drop forms are dutifully submitted, finalizing the class roster: 45 percent Jewish, 18 percent Asian, 100 percent male. The tribe has spoken.” https://www.thecrimson.com/article/2006/12/6/burden-of-proof-at-1002-am/
“If you think A is true and B is false, and find an argument that A implies B, you have three choices: you can decide A is false after all; you can decide B is true after all; or you can decide that the argument actually isn’t valid. Or you can adopt some probabilistic combination: it’s perfectly consistent to believe A is 60% likely to be true, B 60% likely to be false, and the argument 60% likely to be correct. But fundamentally you have to make a choice about which of the three pieces to adjust, and by how much.” https://jaydaigle.net/blog/pascalian-medicine/
"Results from neural networks support the idea that brains use predictions to create perceptions—and that they work that way to conserve power." https://www.wired.com/story/your-brain-is-an-energy-efficient-prediction-machine
Is Technological Progress Slowing? The Case of American Agriculture https://mattsclancy.substack.com/p/is-technological-progress-slowing
Researchers have developed a jelly-like material that can withstand the equivalent of an elephant standing on it, and completely recover to its original shape, even though it’s 80% water. https://www.cam.ac.uk/research/news/super-jelly-can-survive-being-run-over-by-a-car
Memes, merchandise and Mars cocktails: Russia’s mania for Elon Musk has no bounds https://www.washingtonpost.com/world/europe/russia-musk-tesla-mars/2021/11/25/f63eab4e-4179-11ec-9404-50a28a88b9cd_story.html
The Rejuvenome Project at Astera Institute is now accepting proposals for interventions to try to rejuvenate mice, form is at the bottom of https://astera.org/longevity/
Elon Musk says Raptor engine production is a ‘disaster’ that puts SpaceX at risk of bankruptcy https://www.theverge.com/2021/11/30/22809720/elon-musk-spacex-raptor-engine-crisis-bankruptcy-starship
Trigger of rare blood clots with AstraZeneca jab found by scientists https://www.google.com/amp/s/www.bbc.com/news/health-59418123.amp
Here are some ways in which the near future will be radically different from today:
1. Noninvasive brain-computer interfaces in combination with deep learning will revolutionize how we interact with technology. No more keyboards, mice, or touchscreens. A lot will happen before you become consciously aware that you want it to happen, e.g. turning on the lights.
2. Ultra-low-power sensors will be so cheap that we'll be able to gather huge amounts of high-resolution data. In combination with advances in synthetic data, this will allow us to automate many tasks even in the absence of further machine learning breakthroughs.
3. Virtual reality will be as popular and widespread as smartphones are today because the hardware will be as lightweight and cheap as normal glasses. This adoption will eventually pave the way for the augmented reality revolution merging the real and virtual world forever.
4. Virtual assistants like Alexa will be good enough to serve as tutors for children. They will be able to hold conversations that may continue for days, weeks, or even months. This will alleviate loneliness and make computer games much more interesting.
5. Next-generation nuclear power will lift millions out of poverty and enable climate-neutral growth.
6. Universal real-time voice and text translation will finally break the language barrier and allow people to live and work in countries without speaking the local language.
7. Advances in medical monitoring, wearable devices and sensors (e.g. smart toilets) will allow us to detect many medical conditions such as cancer or heart problems early enough to successfully treat them.