Links for 2023-04-11
...generative agents--computational software agents that simulate believable human behavior. Generative agents wake up, cook breakfast, and head to work; artists paint, while authors write; they form opinions, notice each other, and initiate conversations; they remember and reflect on days past as they plan the next day.
To enable generative agents, we describe an architecture that extends a large language model to store a complete record of the agent's experiences using natural language, synthesize those memories over time into higher-level reflections, and retrieve them dynamically to plan behavior.
We demonstrate that, with generative agents, it is sufficient to simply tell one agent that she wants to throw a party. Despite many potential points of failure—the party planner must remember to tell other agents about the party, attendees must remember the invitation, those who remember must decide to actually show up, and other possible points of failure—agents in our environment succeed. They spread the word about the party and then show up, with one agent even asking another agent on a date to the party, all from this single user-generated seed suggestion.
Paper: https://arxiv.org/abs/2304.03442
Project page: https://reverie.herokuapp.com/arXiv_Demo/
Previously: Generally capable agents emerge from open-ended play https://www.deepmind.com/blog/generally-capable-agents-emerge-from-open-ended-play
See also: Using Large Language Models to Simulate Multiple Humans and Replicate Human Subject Studies https://arxiv.org/abs/2208.10264
Neurons can use their spiking to identify their causal effects, allowing an approximation to gradient descent learning. https://journals.plos.org/ploscompbiol/article?id=10.1371/journal.pcbi.1011005
The brain applies rhythms to physical patches of the cortex to selectively control just the right neurons at the right times to do the right things. https://news.mit.edu/2023/spatial-computing-enables-flexible-working-memory-0330
Our new brains: neurotechnology advances that could change everything https://www.freethink.com/hard-tech/visionary-neurotechnology
Comprehensive market analysis on the rapidly growing field of neurotechnology, with a focus on Brain-Computer Interfaces (BCIs) https://www.canva.com/design/DAFKWDyTHH0/h5RgsTiQ35zWCh2IiiebSA/view?utm_content=DAFKWDyTHH0&utm_campaign=designshare&utm_medium=link&utm_source=viewer
Brain Computer Interface : an investor perspective https://blog.newfundcap.com/brain-computer-interface-an-investor-perspective/
Brain morphology seems to have a large general dimension that's related to genetic and phenotypic general intelligence. https://onlinelibrary.wiley.com/doi/10.1002/hbm.26283
Maze-solving agents: Add a top-right vector, make the agent go to the top-right https://www.lesswrong.com/posts/gRp6FAWcQiCWkouN5/maze-solving-agents-add-a-top-right-vector-make-the-agent-go
How well do Large Language Models perform in Arithmetic tasks? https://arxiv.org/abs/2304.02015
“Bringing back analog computers in much more advanced forms than their historic ancestors will change the world of computing drastically and forever.” [Wired] https://archive.is/ZMTku
We cannot afford not to invest in preventing aging: "a slowdown in aging that increases life expectancy by 1 year is worth US$38 trillion, and by 10 years, US$367 trillion." https://www.nature.com/articles/s43587-021-00080-0
DNA Confirms Oral History of Swahili People [The New York Times] https://archive.is/0BnMK
Programmable protein delivery with a bacterial contractile injection system https://www.nature.com/articles/s41586-023-05870-7
“Eternal September or the September that never ended is Usenet slang for a period beginning around 1993[2] when Internet service providers began offering Usenet access to many new users. The flood of new users overwhelmed the existing culture for online forums and the ability to enforce existing norms.” https://en.wikipedia.org/wiki/Eternal_September
Long ago, when shepherds wanted to see if two herds of sheep were isomorphic, they would look for an explicit isomorphism.
In other words, they would line up both herds and try to match each sheep in one herd with a sheep in the other. But one day along came a shepherd who invented decategorification.
She realized one could take each herd and ‘count’ it, setting up an isomorphism between it and some set of ‘numbers’, which were nonsense words like ‘one, two, three, . . . ’ specially designed for this purpose.
By comparing the resulting numbers, she could show that two herds were isomorphic without explicitly establishing an isomorphism!
In short, by decategorifying the category of finite sets, the set of natural numbers was invented.
— Categorification https://arxiv.org/abs/math/9802029