Links for 2022-11-19
Talk to books: Get quotes from more than 100 000 books that respond to your question. A creativity tool by Google to explore new ideas and get relevant quotes. http://books.google.com/talktobooks/
Extreme Generative Image Compression by Learning Text Embedding from Diffusion Models: “The optimization is applied together with a learning compression model to achieve extreme compression of low bitrates <0.1 bpp.” https://arxiv.org/abs/2211.07793
Neural nets decompiling mixed Beatles tracks powered the new remixed Beatles album 'Revolver' https://www.vulture.com/2022/11/the-beatles-revolver-super-deluxe-peter-jackson-get-back.html
The Next Generation of Large Language Models Will Blow Your Mind and Disrupt Your Business https://blog.swift.vc/the-next-generation-of-large-language-models-will-blow-your-mind-and-disrupt-your-business-3b913d6dfa8a
“Quite remarkable that Japan in the 1500s bought a few examples of the latest tech, reverse engineered it, and mass produced their own.” https://twitter.com/whyvert/status/1593242423672864772
Adherence to emotion norms is greater in individualist cultures than in collectivist cultures – the opposite of what most people expect. (N = 200,000) https://psycnet.apa.org/doi/10.1037/pspi0000409
"Citizens do in fact express higher trust in less politicized scientific fields (mathematics, astronomy etc.) than politicized ones (economics and climate science)." https://onlinelibrary.wiley.com/doi/10.1111/pops.12866
"liberals perceive the stock market as a more dangerous & riskier place than conservatives. …explains liberals’ cautious investment behaviors, … support for regulation of the stock market, & their greater opposition to privatization of Social Security." https://onlinelibrary.wiley.com/doi/full/10.1111/pops.12805?campaign=woletoc
“being reflexively anti-establishment is a sloppy heuristic that inevitably devolves into conspiracism and paranoia.” [Quillette] https://archive.ph/wmxuZ
Park et al. chop dendrites off V1 Layer 2/3 neurons to see how orientation selectivity changes. https://www.nature.com/articles/s41467-019-13029-0
"We are contemplating the escape of a virus that could infect individuals one after the other, leading to a..pandemic involving hundreds to millions of deaths, not to mention the..economic impact–put bluntly, a small risk with catastrophic consequences." https://www.the-microbiologist.com/opinion/engineering-viruses-to-make-the-world-a-more-dangerous-place/341.article
Until recently, Twitter roughly had 7,100 employees and 5,500 contractors.
For comparison:
1. WhatsApp only had 55 employees in 2014 and 500 million users when it was acquired by Facebook.
2. OpenAI only had ~120 employees when it released GPT-3.
3. Just a handful of people maintain software libraries underlying our scientific infrastructure such as matplotlib, NumPy, and pandas.
4. DARPA has 220 employees.
And Twitter has seen almost no innovation. C'mon, man!
How Tesla caught an employee who leaked data in 2008: