Links for 2023-06-02
NVIDIA's Neuralangelo can turn videos from any device into detailed 3D structures, fully replicating buildings, sculptures, or other real world objects or spaces virtually.
Learn more: https://blogs.nvidia.com/blog/2023/06/01/neuralangelo-ai-research-3d-reconstruction/
How big tech companies are investing in AI startups. https://news.crunchbase.com/ai-robotics/ai-venture-funding-nvidia-microsoft-salesforce-intel-google/
Full Automation is Unlikely and Unnecessary for Explosive Growth https://www.lesswrong.com/posts/sW8CG7TiPFG5Bxwjs/full-automation-is-unlikely-and-unnecessary-for-explosive
A value-based RL agent that achieves super-human performance in the Atari 100K benchmark. https://arxiv.org/abs/2305.19452
Humans in 4D: Reconstructing and Tracking Humans with Transformers https://shubham-goel.github.io/4dhumans/
Some Neural Networks Learn Language Like Humans https://www.quantamagazine.org/some-neural-networks-learn-language-like-humans-20230522/
TinyStories: Small Language Models That Still Speak Coherent English https://www.lesswrong.com/posts/dMoaBvcxpBE7LcES4/tinystories-small-language-models-that-still-speak-coherent
Musings on how the future of PCs & the internet will be Personal Language Models (PLMs). https://twitter.com/garybasin/status/1650294293280571392
Behavioral Game Theory for LLMs: “In the canonical iterated Prisoner's Dilemma, we find that GPT-4 acts particularly unforgivingly, always defecting after another agent has defected only once.” https://arxiv.org/abs/2305.16867
Scaling Data-Constrained Language Models https://arxiv.org/abs/2305.16264
PEARL: Prompting Large Language Models to Plan and Execute Actions Over Long Documents https://arxiv.org/abs/2305.14564
Geoffrey Hinton: “Very recently, I realised that maybe the digital intelligences are actually learning *better* than the brain. With 1% of the storage capacity, GPT-4 knows thousands of times more than us. That strongly suggests its got a better way of getting information into the connections.” https://twitter.com/peterhartree/status/1661860447186722817
“We found clean, CAUSAL evidence that the shingles vaccine prevents a good chunk of dementia cases. So, could a virus cause Alzheimer’s” https://www.medrxiv.org/content/10.1101/2023.05.23.23290253v1
Key development in the Ukraine Starlink saga: Pentagon signs a contract with SpaceX to provide Ukraine with critical terminals that enable battlefield intelligence, surveillance and reconnaisance, as well as artillery correction and counter-battery fire. [Bloomberg] https://archive.is/W0siX