Proteins are the building blocks of life, involved in virtually every biological process. Understanding how proteins interact with each other is crucial for deciphering the complexities of cellular functions, and has significant implications for drug development and the treatment of diseases.
Generative models have had remarkable success in various applications, from image and video generation to composing music and to language modeling. The problem is that we are lacking in theory, when it comes to the capabilities and limitations of generative models; understandably, this gap can seriously affect how we develop and use them down the line.
One of the main challenges has been the ability to effectively pick samples from complicated data patterns, especially given the limitations of traditional methods when dealing with the kind of high-dimensional and complex data commonly encountered in modern AI applications.
Now, a team of scientists led by Florent Krzakala and Lenka Zdeborová at EPFL has investigated the efficiency of modern neural network-based generative models. The study, published in PNAS, compares these contemporary methods against traditional sampling techniques, focusing on a specific class of probability distributions related to spin glasses and statistical inference problems.
“SoftBank was founded for what purpose? For what purpose was Masa Son born? It may sound strange, but I think I was born to realize ASI. I am super serious about it.” — Masayoshi Son.
SoftBank CEO Masayoshi Son laid out his vision for artifical super intelligence, or ASI, that he said would be 10,000 times smarter than humans.
Nearly all the neural networks that power modern artificial intelligence tools such as ChatGPT are based on a 1960s-era computational model of a living neuron. A new model developed at the Flatiron Institute’s Center for Computational Neuroscience (CCN) suggests that this decades-old approximation doesn’t capture all the computational abilities that real neurons possess and that this older model is potentially holding back AI development.
Building symmetry breaking into neural networks.
Discovering Symmetry Breaking in Physical Systems with Relaxed Group Convolution.
Rui Wang, Elyssa Hofgard, Han Gao, Robin Walters, Tess Smidt MIT June 2024 https://openreview.net/forum?id=59oXyDTLJv.
Modeling symmetry breaking is key…
Elon Musk says he wants Tesla’s humanoid robot to be considered a friend. Musk also joked that the company wanted to make the robot “good-looking.”
Speaking at the Cannes Lions International Festival of Creativity, Elon Musk discussed Tesla’s ambitions for its humanoid robot.
Posted in food, life extension, robotics/AI, sustainability | 1 Comment on Conscious AI
Plus, Turing also showed that achieving universality doesn’t require anything fancy. The basic equipment of a universal machine is just not more advanced than a kid’s abacus — operations like incrementing, decrementing, and conditional jumping are all it takes to create software of any complexity: be it a calculator, Minecraft, or an AI chatbot.
Likewise, consciousness might just be an emergent property of the software running AGI, much like how the hardware of a universal machine gives rise to its capabilities. Personally, I don’t buy into the idea of something sitting on top of the physical human brain — no immortal soul or astral “I” floating around in higher dimensions. It’s all just flesh and bone. Think of it like an anthill: this incredibly complex system doesn’t need some divine spirit to explain its organized society, impressive architecture, or mushroom farms. The anthill’s intricate behaviour, often referred to as a superorganism, emerges from the interactions of its individual ants without needing to be reduced to them. Similarly, a single ant wandering around in a terrarium won’t tell you much about the anthill as a whole. Brain neurons are like those ants — pretty dumb on their own, but get around 86 billion of them together, and suddenly you’ve got “I” with all its experiences, dreams, and… consciousness.
So basically, if something can think, it can also think about itself. That means consciousness is a natural part of thinking — it just comes with the territory. And if you think about it, this also means you can’t really have thinking without consciousness, which brings us back to the whole Skynet thing.
face_with_colon_three year 2021.
The solar aircraft is made by a Spanish-American aerospace startup called Skydweller Aero. Based in Oklahoma City, the company raised $32 million in its Series A funding round, led by Italian aerospace firm Leonardo.
“For us, if you’re flying 90 days with one aircraft, that’s two takeoffs and landings versus … hundreds,” Skydweller Aero co-founder John Parkes told Aviation Today. “Being able to fly thousands of miles, persist over an area for 30–60 days and fly back is a differentiator. It’s a huge cost savings to the US government when you look at the whole cost of doing a lot of the national security missions that we have.”
Tesla’s development of AI-powered self-driving cars has the potential to revolutionize transportation, disrupt labor, and create significant value in the market.
Questions to inspire discussion.
What is Tesla’s approach to AI-powered self-driving cars?
—Tesla’s approach involves convergence of AI hardware for cars and robots, with a focus on maximizing hardware and software interplay for real world AI products.
