Links for 2023-04-21
Align your Latents: High-Resolution Video Synthesis with Latent Diffusion Models
...we turn the image generator into a video generator by introducing a temporal dimension to the latent space diffusion model and fine-tuning on encoded image sequences, i.e., videos. Similarly, we temporally align diffusion model upsamplers, turning them into temporally consistent video super resolution models...
Text prompt for the displayed video: "A teddy bear is playing the electric guitar, high definition, 4k."
Project page: https://research.nvidia.com/labs/toronto-ai/VideoLDM/
Robust flight navigation out of distribution with liquid neural networks https://news.mit.edu/2023/drones-navigate-unseen-environments-liquid-neural-networks-0419
“We formulate a general tool learning framework: starting from understanding the user instruction, models should learn to decompose a complex task into several subtasks, dynamically adjust their plan through reasoning, and effectively conquer each sub-task by selecting appropriate tools. We also discuss how to train models for improved tool-use capabilities and facilitate the generalization in tool learning.” https://arxiv.org/abs/2304.08354
🦎 Chameleon: Plug-and-Play Compositional Reasoning with GPT-4: “Chameleon with GPT-4 achieves an 86.54% accuracy on ScienceQA, significantly improving upon the best published few-shot model by 11.37%; using GPT-4 as the underlying LLM, Chameleon achieves a 17.8% increase over the state-of-the-art model, leading to a 98.78% overall accuracy on TabMWP.” https://github.com/lupantech/chameleon-llm
“Utilizing internet videos of human behavior, we train a visual affordance model that estimates where and how in the scene a human is likely to interact. The structure of these behavioral affordances directly enables the robot to perform many complex tasks.” https://robo-affordances.github.io/
Text2Performer: Text-Driven Human Video Generation https://yumingj.github.io/projects/Text2Performer.html
Mental navigation and telekinesis with a hippocampal map-based brain-machine interface- mice can control a virtual environment simply by activating place cells https://www.biorxiv.org/content/10.1101/2023.04.07.536077v1
Moonshot proposal for leveraging AI to facilitate literal conversations between humans and sperm whales https://www.sciencedirect.com/science/article/pii/S2589004222006642
“NOTHING IS SAFER THAN AN EXTEMPLORATION OF FANTASMICALITY.” https://www.lesswrong.com/posts/jkY6QdCfAXHJk3kea/the-petertodd-phenomenon
Lawsuits Are the Hitman of the State: "How is this monolithic censorship possible in a politically diverse nation? Because every employer is afraid of getting sued for civil rights violation - and every worker is afraid of getting fired for putting his employer at risk of getting sued." https://betonit.substack.com/p/lawsuits-are-the-deep-state
"Chinese adults boost their average vocabulary score by 3.21 points and mathematical score by 3.83 points per decade" https://www.sciencedirect.com/science/article/abs/pii/S0160289623000338
Mirror-imaging in molecules can modify neuron signaling https://news.unl.edu/newsrooms/today/article/mirror-imaging-in-molecules-can-modify-neuron-signaling/
Reality is a Paradox – Mathematics, Physics, Truth & Love. https://www.youtube.com/watch?v=Osh0-J3T2nY
“My intuition tells me that categorical string diagrams, as described in Quantum in Pictures, is the future of human thinking…These diagrams capture the 'affordances' of a complex system (here it is Quantum computing) and distill it in diagrams that can be manipulated with certain rules that underneath are mathematically sound. The methodology leverages human intuition to feel its way toward solutions.” https://twitter.com/IntuitMachine/status/1645354785342738437
Yudkowsky abandons alignment research 🙀 [deepfake] https://twitter.com/YaBoyFathoM/status/1649103596930187290
China’s Foreign Ministry spokesman when asked about India overtaking China as the world’s most populous country: “We need to look at not just the size but also the quality of the population.” https://www.fmprc.gov.cn/mfa_eng/xwfw_665399/s2510_665401/202304/t20230419_11061886.html
An impressive demonstration of how subtle changes can really make a huge difference in how someone is perceived:
I wrote about the power of subtle clues before:
A few years ago, I was sitting in a doctor's waiting room when a teenage girl suddenly claimed that an older man sitting next to her had stung her.
Everyone present in the room looked away, embarrassed, or pretended that they hadn't heard anything. The man just showed his hands and denied the allegations.
A strange reaction, you might think. Yes, but only if she had been a typical teenage girl.
There wasn't anything obviously wrong with her. Yet everyone in the room knew she had mental problems even before she said a word. How? Hard to describe in words. Her posture. Her stare. Her movements. Many subtle, almost subliminal clues created an overall impression of someone who is mentally disturbed.
It's very difficult to appear normal if you are not neurotypical. If you're not an excellent actor who studied the fine nuances of normal behavior, you just risk sliding down into the uncanny valley and appearing deceptive and fake.
P.S. Women, due to being physically weaker, are especially attentive to such cues and err on the side of caution. This is probably the predominant reason for the Incel phenomenon. Those guys try everything to impress women but it is never enough so they conclude it must be the fault of women rather than their own fault. They don't realize that their problem with women is much more subtle and harder to fix.
P.P.S. Tom Cruise is an interesting example. He's popular but many people sense that he's a little bit crazy. Maybe he practiced the "Scientology stare" a little bit too much.