Imagine the confusion as my AI submits ideas…

https://www.bespacific.com/this-tiny-website-is-googles-first-line-of-defense-in-the-patent-wars/

This Tiny Website Is Google’s First Line of Defense in the Patent Wars

Wired: “TDCommons is a free space for inventors to lay claim to breakthroughs without having to file a patent. Why is it so off the radar? A trio of Google engineers recently came up with a futuristic way to help anyone who stumbles through presentations on video calls. They propose that when algorithms detect a speaker’s pulse racing or “umms” lengthening, a generative AI bot that mimics their voice could simply take over. That cutting-edge idea wasn’t revealed at a big company event or in an academic journal. Instead, it appeared in a 1,500-word post on a little-known, free website called TDCommons.org that Google has quietly owned and funded for nine years. Until WIRED received a link to an idea on TDCommons last year and got curious, Google had never spoken with the media about its website. Scrolling through TDCommons, you can read Google’s latest ideas for coordinating smart home gadgets for better sleep, preserving privacy in mobile search results, and using AI to summarize a person’s activities from their photo archives. And the submissions aren’t exclusive to Google; about 150 organizations, including HP, Cisco, and Visa, also have posted inventions to the website. The website is a home for ideas that seem potentially valuable but not worth spending tens of thousands of dollars seeking a patent for. By publishing the technical details and establishing “prior art,” Google and other companies can head off future disputes by blocking others from filing patents for similar concepts. Google gives employees a $1,000 bonus for each invention they post to TDCommons—a tenth of what it awards its patent seekers—but they also get an immediately shareable link to gloat about otherwise secretive work. TDCommons adds to Google’s long-standing, and far more vocal, efforts to carve out greater space for freewheeling innovation in an industry where patents can be used to hobble or extract cash from competitors. The site may be dowdy and obscure, but it does the trick. “The beauty of defensive publications is that this website can be pretty simple,” says Laura Sheridan, Google’s head of patent policy. “It needs to establish a date. And it needs to have documents be accessible. There’s not much more we need to do.” In reality, the experiment has struggled to cut through government bureaucracy and overcome competition from more robust archives. Sheridan acknowledges it’s a work in progress. TDCommons needs a bigger flow of uploads to become less peculiar and more vital. It offers a unique hope of expanding public access to the technical creativity happening inside corporate walls—and shifting more resources toward that work.”

Tools & Techniques.

https://www.nature.com/articles/s43588-024-00593-9

Automated discovery of algorithms from data

To automate the discovery of new scientific and engineering principles, artificial intelligence must distill explicit rules from experimental data. This has proven difficult because existing methods typically search through the enormous space of possible functions. Here we introduce deep distilling, a machine learning method that does not perform searches but instead learns from data using symbolic essence neural networks and then losslessly condenses the network parameters into a concise algorithm written in computer code. This distilled code, which can contain loops and nested logic, is equivalent to the neural network but is human-comprehensible and orders-of-magnitude more compact. On arithmetic, vision and optimization tasks, the distilled code is capable of out-of-distribution systematic generalization to solve cases orders-of-magnitude larger and more complex than the training data. The distilled algorithms can sometimes outperform human-designed algorithms, demonstrating that deep distilling is able to discover generalizable principles complementary to human expertise.