This Nation is Doomed ☠️

The future of a nation can be measured by two types of noise: Babies crying and the roar of the nation's fighter jets in training. These are the sounds of hope and freedom. Their absence means a nation's future is in jeopardy.

Link: https://archive.is/TZJcr

This is awesome and I fully support it. But I'd wager that it's a little inconsistent with his (and my own) AI timelines. Whether a self-sustaining Mars colony can be established before the singularity is questionable. But definitely worth trying.

https://x.com/elonmusk/status/1832550322293837833

Links for 2024-09-03

AI:

1. Survey: How Do Elite Chinese Students Feel About the Risks of AI? https://www.lesswrong.com/posts/gxCGKHpX8G8D8aWy5/survey-how-do-elite-chinese-students-feel-about-the-risks-of

2. Can Meta clean up its data centers with geothermal energy? https://www.theverge.com/2024/8/29/24231301/meta-facebook-geothermal-energy-sage-data-center-partnership

3. "ChatGPT’s weekly users have doubled in less than a year" ("API use has doubled following...GPT-4o-mini") https://www.theverge.com/2024/8/29/24231685/openai-chatgpt-200-million-weekly-users

4. Feds to get early access to OpenAI, Anthropic AI to test for doomsday scenarios https://arstechnica.com/tech-policy/2024/08/feds-to-get-early-access-to-openai-anthropic-ai-to-test-for-doomsday-scenarios/

5. Companies like JPMorgan and Walmart are shifting from restricting generative AI tools like ChatGPT to developing their own internal AI assistants. https://www.cnbc.com/2024/08/28/why-jpmorgan-and-walmart-are-opting-for-internal-gen-ai-assistants.html

6. Crypto is the Money for AIs https://marginalrevolution.com/marginalrevolution/2024/09/crypto-is-the-money-for-ais.html

Neuroscience:

1. Nervous System’s Master Matchmaker Sparks Breakthrough in Computer Science https://www.cshl.edu/the-nervous-systems-matchmaker/

2. Researchers give adult zebra finches back their ability to learn new songs https://www.mpg.de/23351337/zebra-finches-learn-songs-in-old-age

3. “The information throughput of a human being is about 10 bits/s. In comparison, our sensory systems gather data at an enormous rate, no less than 1 gigabits/s. The stark contrast between these numbers remains unexplained.” https://arxiv.org/abs/2408.10234

4. UK researchers find Alzheimer’s-like brain changes in long COVID patients https://uknow.uky.edu/research/uk-researchers-find-alzheimer-s-brain-changes-long-covid-patients

Technology:

1. Robot Metalsmiths Are Resurrecting Toroidal Tanks for NASA: Now bots can form sheet metal in complex geometries https://spectrum.ieee.org/metal-forming-robot

2. Researchers discover a surprising way to jump-start battery performance: Charging lithium-ion batteries at high currents just before they leave the factory is 30 times faster and increases battery lifespans by 50%, according to a study at the SLAC-Stanford Battery Center https://www.eurekalert.org/news-releases/1056171

3. Scientists at Peking University have developed a highly efficient tensor processing unit (TPU) based on carbon nanotubes. https://techxplore.com/news/2024-08-tensor-processor-chip-based-carbon.html

4. If you don’t understand why things don’t always add up when you use floating point numbers—and in particular, if you don’t understand why you should never use floating point when working with money—you should read the Floating Point Guide. https://floating-point-gui.de/

5. You really can just make a fake sky with a reflector and an LED https://www.youtube.com/watch?v=6bqBsHSwPgw

6. A reusable ramming drone from Poland. The drone has a reinforced structure, destroys enemy drones with kinetic force, and will theoretically survive such an attack and return to the "base". https://x.com/wolski_jaros/status/1830995429313650912

Miscellaneous:

1. Astronomers puzzled by little red galaxies that seem impossibly dense https://www.newscientist.com/article/2445967-astronomers-puzzled-by-little-red-galaxies-that-seem-impossibly-dense/ [no paywall: https://archive.is/NhWAs]

2. Earthworms have ‘completely scrambled’ genomes. Did that enable their ancestors to leave the sea? https://www.science.org/content/article/earthworms-have-completely-scrambled-genomes-did-help-their-ancestors-leave-sea [no paywall: https://archive.is/2qzx5]

Links for 2024-08-31

AI:

1. “Here we show that symbolic search over the space of metaprograms—programs that revise programs—dramatically improves learning efficiency. On a behavioral benchmark of 100 algorithmically rich rules, this approach fits human learning more accurately than alternative models while also using orders of magnitude less search.” https://www.nature.com/articles/s41467-024-50966-x

2. Learning integral operators via neural integral equations: “From physics to biology, and beyond, our approach addresses a critical gap in ML for nonlocal operators.” https://www.nature.com/articles/s42256-024-00886-8

3. How to build a lensless camera https://learnandburn.ai/p/how-to-build-a-lensless-camera

4. Ask Claude: Amazon turns to Anthropic's AI for Alexa revamp https://www.reuters.com/technology/artificial-intelligence/amazon-turns-anthropics-claude-alexa-ai-revamp-2024-08-30/

5. Dell’s AI Server Business Now Bigger Than VMware Used To Be https://www.nextplatform.com/2024/08/30/dells-ai-server-business-now-bigger-than-vmware-used-to-be/

6. ReKep: Spatio-Temporal Reasoning of Relational Keypoint Constraints for Robotic Manipulation https://rekep-robot.github.io/

7. You Only Cache Once: Decoder-Decoder Architectures for Language Models — Substantially reduces GPU memory demands, yet retains global attention capability. https://arxiv.org/abs/2405.05254

8. Smaller, Weaker, Yet Better: Training LLM Reasoners via Compute-Optimal Sampling https://arxiv.org/abs/2408.16737

9. Microsoft Research introduces AutoGen Studio, a low-code interface for rapidly prototyping AI agents. https://microsoft.github.io/autogen/docs/autogen-studio/getting-started/

10. Human brain organoid bioprocessors now available to rent for $500 per month https://www.tomshardware.com/pc-components/cpus/human-brain-organoid-bioprocessors-now-available-to-rent-for-dollar500-per-month

11. Machine Learning Potentials: A Roadmap Toward Next-Generation Biomolecular Simulations https://www.arxiv.org/abs/2408.12625

12. AI for Bio: State Of The Field https://www.lesswrong.com/posts/CGegZveogJRGCK3LA/ai-for-bio-state-of-the-field

13. Principles for the AGI Race https://www.lesswrong.com/posts/aRciQsjgErCf5Y7D9/principles-for-the-agi-race

14. Anthropic CEO Dario Amodei says big, powerful AI models will spawn and orchestrate smaller models to assist with tasks, creating a swarm intelligence that will decrease the need for human input https://www.youtube.com/watch?v=7xij6SoCClI

15. Instead of using an LLM to solve a problem set, flip the table and ask the student to TEACH the LLM how to solve the problem set. https://www.youtube.com/watch?v=om7VpIK90vE

16. WavTokenizer: an Efficient Acoustic Discrete Codec Tokenizer for Audio Language Modeling https://arxiv.org/abs/2408.16532

Miscellaneous:

1. David Reich – How One Small Tribe Conquered the World 70,000 Years Ago — "Human history has been again and again a story of one group figuring ‘something’ out, and then basically wiping everyone else out." https://www.youtube.com/watch?v=Uj6skZIxPuI

2. “Climate Tipping Points are Real, Stop Denying It” https://www.youtube.com/watch?v=HiO3M_cPeSo

3. The Terrifying Power of Mathematics https://www.math.columbia.edu/~woit/wordpress/?p=14113

4. The Netherlands plans to limit ASML Holding NV’s ability to repair and maintain its semiconductor equipment in China, a potentially painful blow to Beijing’s efforts to develop a world-class chip industry. https://www.bloomberg.com/news/articles/2024-08-29/the-netherlands-to-put-more-curbs-on-asml-s-china-chip-business [no paywall: https://archive.is/J9Eqk]

Links for 2024-08-27

AI:

1. “Perplexity is the first generalized AI chatbot I’ve found useful enough to integrate into any part of my daily workflow, much less across multiple domains.” https://www.lesswrong.com/posts/CcJ2YWPqcMsWg8vGt/perplexity-wins-my-ai-race-1

2. AI-powered coding pulls in almost $1bn of funding to claim ‘killer app’ status https://www.ft.com/content/4868bd38-613c-4fa9-ba9d-1ed8fa8a40c8 [no paywall: https://archive.is/SY7PJ]

3. Chip starvation driving Chinese DL flight to cheap/specialized/edge uses, away from SOTA leading-edge LLMs https://www.wsj.com/tech/ai/chips-or-not-chinese-ai-pushes-ahead-31034e3d [no paywall: https://archive.is/lhdhG]

4. "Is Xi Jinping an AI doomer? China’s elite is split over artificial intelligence" https://www.economist.com/china/2024/08/25/is-xi-jinping-an-ai-doomer [no paywall: https://archive.is/7E6Ea]

5. AlphaFold2 knows some protein folding principles https://www.biorxiv.org/content/10.1101/2024.08.25.609581v1

6. Learning to Move Like Professional Counter-Strike Players https://arxiv.org/abs/2408.13934

7. Training-free Long Video Generation with Chain of Diffusion Model Experts https://arxiv.org/abs/2408.13423

8. Self-Consuming Generative Models with Curated Data Provably Optimize Human Preferences https://arxiv.org/abs/2407.09499

9. AI Spots Cancer and Viral Infections at Nanoscale Precision https://www.crg.eu/en/news/ai-spots-cancer-and-viral-infections-nanoscale-precision

10. Machine learning algorithm improves 3D printing efficiency https://www.engineering.com/machine-learning-algorithm-improves-3d-printing-efficiency/

11. Astrophysicists Use AI to Precisely Calculate Universe’s ‘Settings’ https://www.simonsfoundation.org/2024/08/26/astrophysicists-use-ai-to-precisely-calculate-universes-settings/

12. Building a nuclear fusor in your bedroom with Claude Sonnet 3.5 + Projects? https://x.com/hud_zah/status/1827057785995141558

13. Stuart Russel on "AI: What if we succeed?" (25 April 2024) https://www.youtube.com/watch?v=UvvdFZkhhqE

14. What’s Really Going On in Machine Learning? Some Minimal Models https://writings.stephenwolfram.com/2024/08/whats-really-going-on-in-machine-learning-some-minimal-models/

Neuroscience:

1. In language-processing areas of the brain, some cell populations respond to one word, while others respond to strings of words. https://news.mit.edu/2024/scientists-find-neurons-process-language-on-different-timescales-0826

2. Why do varied DNN designs yield equally good models of human vision? Paper shows that diverse DNNs represent images with a shared set of latent dimensions, and these shared dimensions turn out to also be the most brain-aligned. https://arxiv.org/abs/2408.12804

3. Possible evidence for the Law of General Intelligence in honeybees [PDF] https://gwern.net/doc/iq/animal/2024-penaherreraaguirre.pdf

4. Do Cleaner Fish (Labroides dimidiatus) Have General Cognitive Ability? A Reanalysis of Individual Differences Data and Consideration of Phylogenetic Context [PDF] https://gwern.net/doc/iq/animal/2023-woodley.pdf

5. New study may have cracked the code for a type of working memory in C. elegans https://www.biorxiv.org/content/10.1101/2024.08.11.607402v1

Technology:

1. Compute Moonshots: Harnessing New Physics https://www.mackenziemorehead.com/compute-moonshots/

2. Ever wanted to use Python and recreate the first Monte Carlo simulations that were made on an IBM 701 in the 50s? That's just 1 of the goal of "Learn molecular simulations with Python". https://mdcourse.github.io/

Politics:

1. “Not a single country in the entire world has a higher fertility rate than in 1950.” https://x.com/StatisticUrban/status/1828171338760196549

2. The staggering death toll of scientific lies https://www.vox.com/future-perfect/368350/scientific-research-fraud-crime-jail-time

3. Thoughts on seed oil https://www.lesswrong.com/posts/DHkkL2GxhxoceLzua/thoughts-on-seed-oil

4. Is Telegram really an encrypted messaging app? https://blog.cryptographyengineering.com/2024/08/25/telegram-is-not-really-an-encrypted-messaging-app/

Few realize that more and more large companies are now primarily AI companies.

Disney, for example, is no longer just about movies. The company is building humanoid robots that can move like humans: https://www.youtube.com/watch?v=Q2I7u0tjlJs

At Tesla, a lot of research is going into autonomous driving and humanoid robots.

X is now primarily focused on competing with OpenAI.

The same goes for Meta, Amazon, Apple, etc.

But analysts are still saying that AI is a bubble. As if all these companies are delusional and sinking billions of dollars into AI for nothing.

Amazon: “We estimate [our GenAI assistant] has saved us the equivalent of 4,500 developer-years of work.”

Source: https://x.com/ajassy/status/1826608791741493281

Automated Design of Agentic Systems

Presents Meta Agent Search to demonstrate that we can use agents to invent novel and powerful agent designs by programming in code. Meta Agent Search discovers superior agents compared to the baselines in every domain.

Project page: https://www.shengranhu.com/ADAS/
Code: https://github.com/ShengranHu/ADAS

"Importantly, we consistently observe the surprising result that agents invented by Meta Agent Search maintain superior performance even when transferred across domains and models, demonstrating their robustness and generality. Here, we present performance across multiple domains when transferring top agents from the Math (MGSM) domain to non-math domains. Agents discovered by Meta Agent Search in the math domain can outperform or match the performance of baselines after being transferred to domains beyond math."