Links for 2022-10-30
A “Safety Emergency” Happened at the Wuhan Institute of Virology in November of 2019 — “Pro-Publica (with Vanity Fair) is putting their reputation on the line with a very damning report suggesting COVID emerged from a lab leak at the Wuhan Institute of Virology. The new information isn’t about the virus but about political reports that indicate that there was some kind of emergency at the lab in November of 2019, an emergency that was so serious Xi himself got involved.” https://marginalrevolution.com/marginalrevolution/2022/10/a-safety-emergency-happened-at-the-wuhan-institute-of-virology-in-november-of-2019.html
“Finetuning language models on instructions increasingly seems a compute-efficient way to gain performance…Increasing model size continues to yield major gains. Increasing the number of tasks helps, but brings diminishing returns.” https://threadreaderapp.com/thread/1586052079495577600.html
Deep learning with light: A new method uses optics to accelerate machine-learning computations on smart speakers and other low-power connected devices. https://news.mit.edu/2022/optics-deep-learning-computations-1020
“In this work, we present a new learning algorithm, PI-ARS, that combines gradient-based representation learning with gradient-free evolutionary strategy algorithms to leverage the advantages of both.” https://ai.googleblog.com/2022/10/pi-ars-accelerating-evolution-learned.html
Adversarial ML Attack that Secretly Gives a Language Model a Point of View https://www.schneier.com/blog/archives/2022/10/adversarial-ml-attack-that-secretly-gives-a-language-model-a-point-of-view.html
Draft contracts 3x faster with AI: “Spellbook uses GPT-3 to review and suggest language for your contracts, right in Microsoft Word.” https://www.spellbook.legal/
“As a first demo: a Dust app showcasing the now famous capability of models to generate code to answer math questions.” https://twitter.com/dust4ai/status/1582769638282100747
“…we’ve launched our first product, a Person Sensor that is available on SparkFun for $10. This is a small hardware module that detects nearby faces, and returns information about how many there are, where they are relative to the device, and performs facial recognition.” https://petewarden.com/2022/10/20/launching-useful-sensors/
Much more than you ever wanted to know about NEPA [National Environmental Policy Act] https://www.thecgo.org/benchmark/much-more-than-you-ever-wanted-to-know-about-nepa/
"Ukraine is bleeding in the 21st century’s first imperial war of colonial conquest — and yet the places where it finds the least sympathy are former colonies" [Wall Street Journal] https://archive.ph/Uvxk4
“12,000 Russian troops were supposed to defend Kaliningrad... it became cannon fodder for a Ukrainian army that, on paper, was weaker than the Russian army was. Now Kaliningrad is all but defenseless, and the threat the oblast’s troops once posed to NATO … has evaporated.” https://www.forbes.com/sites/davidaxe/2022/10/27/12000-russian-troops-once-posed-a-threat-from-inside-nato-then-they-went-to-ukraine-to-die/?sh=7af9b7843375
Ukraine Unleashes Mass Kamikaze Drone Boat Attack On Russia’s Black Sea Fleet Headquarters https://www.thedrive.com/the-war-zone/ukraine-unleashes-mass-kamikaze-drone-boat-attack-on-russias-black-sea-fleet-headquarters
The other day I watched Animal Kingdom and was curious about an industrial plant that could be seen in the background of one scene. I used Google Lens and it told me it was the San Onofre Nuclear Generating Station.
I'm frequently amazed by how good these AI tools are now. Just 10 years ago this was science fiction.
Previously I used Google Lens to translate an upsidedown children's drawing that contains handwritten Mandarin and have it augment the live-view with a translation in the language of my choice.
Another day I wanted to know when the next garden waste campaign would take place. So I just asked, "Hey Google, when is the next garden waste campaign?", and it told me the date, time, and location for my district. This was much faster than searching for my city's website and manually finding the relevant data.