Lifeboat Foundation

This post is also available in: עברית (Hebrew)

Experts at the Technical University of Munich (TUM) have pioneered the world’s first ethical algorithm for autonomous vehicles, which could see autonomous driving become the norm globally.

The researchers’ ethical algorithm is significantly more advanced than its predecessors, as it fairly distributes levels of risks instead of operating on an either/or principle. The algorithm has been tested in 2,000 scenarios of critical conditions in various settings, such as streets in Europe, the US, and China. The innovation could improve the safety and uptake of autonomous vehicles worldwide.

A new algorithm can organize hundreds of atoms into pristine patterns—including a honeycomb lattice, a fractal called a Sierpiński triangle, and a lion’s head.

Black box optimization methods are used in every domain, from Artificial Intelligence and Machine Learning to engineering and finance. These methods are used to optimize functions when an algebraic model is absent. Black box optimization looks into the design and analysis of algorithms for those problem statements where the structure of the objective function or the limitations defining the set is not known or explainable. Given a set of input parameters, black box optimization methods are designed to evaluate the optimal value of a function. This is done by iteratively assessing the function at multiple points in the input space so as to find the point that generates the optimal output.

Though gradient descent is the most used optimization approach for deep learning models, it is unsuitable for every problem. In cases where gradients cannot be calculated directly or where an objective function’s accurate analytical form is unknown, other approaches like Evolution Strategies (ES) are used. Evolution strategies come from evolutionary algorithms, which refer to a division of population-based optimization algorithms inspired by natural selection. Basically, Evolution Strategies (ES) is a type of Black Box Optimization method that operates by refining a sampling distribution based on the fitness of candidates and updating rules based on equations.

In a new AI paper, researchers from Deepmind, have introduced and developed a new way to use machine learning to learn the update rules from data, called meta-black-box optimization (MetaBBO), to make ES more flexible, adaptable, and scalable. MetaBBO works by meta-learning a neural network parametrization of a BBO update rule. The researchers have used MetaBBO to discover a new type of ES called learned evolution strategy (LES). The learned evolution strategy LES is a type of Set Transformer that updates its solutions based on the fitness of candidates and not depending upon the ordering of candidate solutions within the Black box evaluations. After meta-training, the LES can learn to choose the best-performing solution or update solutions based on a moving average.

Imagine a universe with extremely strong gravity. Stars would be able to form from very little material. They would be smaller than in our universe and live for a much shorter amount of time. But could life evolve there? It took human life billions of years to evolve on Earth under the pleasantly warm rays from the Sun after all.

Now imagine a universe with extremely weak gravity. Its matter would struggle to clump together to form stars, planets and—ultimately—living beings. It seems we are pretty lucky to have gravity that is just right for life in our universe.

This isn’t just the case for gravity. The values of many forces and particles in the universe, represented by some 30 so-called fundamental constants, all seem to line up perfectly to enable the evolution of intelligent life. But there’s no theory explaining what values the constants should have—we just have to measure them and plug their numbers into our equations to accurately describe the cosmos.

An artificial pancreas originally developed at the University of Virginia Center for Diabetes Technology improves blood sugar control in children ages 2 to 6 with type 1 diabetes, according to a new study. Details of the clinical study and its findings have been published in the New England Journal of Medicine.

Trial participants using the artificial pancreas spent approximately three more hours per day in their target blood sugar range compared with participants in a control group who continued relying on the methods they were already using to manage their blood sugar.

The Control-IQ system, manufactured by Tandem Diabetes Care, is a diabetes management device that automatically monitors and regulates blood glucose. The artificial pancreas has an insulin pump that uses advanced control algorithms based on the person’s glucose monitoring information to adjust the insulin dose as needed.

The Eqs. (3a) and (3b) suggest two important features of the location of neutrons and the spin by switching the choice of the post-selection: (i) The first lines indicate that the neutrons are found to be localized in different paths by switching the choice of the post-selection; they are found in the path I and II by applying the post-selection \({|{\Psi ^{+}_f}\rangle }\) and \({|{\Psi ^{-}_f}\rangle }\), respectively. (ii) The lines of the second part of the equations indicate that the spin in the different paths is found to be affected by switching the choice of the post-selection; the spin in path II and I is affected by applying the post-selection \({|{\Psi ^{+}_f}\rangle }\) and \({|{\Psi ^{-}_f}\rangle }\), respectively. Note that, in both choices of the post-selection, neutron and spin are localized in different paths, i.e., the location of the cat itself and its grin are interchanged by switching the choices of the post-selection. Since measurement of the locations of the neutron and the spin in the interferometer can be carried out independently of the delayed-choice process, the picking of a direction for post-selection, the influence of the delayed-choice on the preceding measurements can be investigated. We would like to point out that the experimental proposal in a recent publication³⁵, contains a delayed choice scenario, too. The difference to the experiment presented in this report is that the authors of³⁵ suggest a setup where two properties of the same system, represented by two non-commuting observables, are separated. In contrast to that, we deal in our experiment with the separation of one property from the system itself, hereby constituting the phenomenon of disembodiment. Further we would like to point out that in their Gedanken-experiment the effect of a change in the pre-selection is discussed that in our view has no retro-causal implications.

The experiment was carried out at the S18 silicon-perfect-crystal interferometer beam line at the high flux reactor at the Institute Laue Langevin. A schematic view of the experimental set-up is shown in Fig. 2.

Year 2014 face_with_colon_three If black holes have infinitely small sizes and infinitely density this also means that string theory would also solve the infinitely small problem because now we know that infinitely small sizes exist and if that exists then so does infinite energy from super string essentially filling out the rest of the mystery of the God equation. This means that computers could be infinitely small aswell saving a ton of space aswell.

If you’ve wondered how big is a black hole? then you’ve come to the right place! Learn about the sizes of black holes and the multi-layered answer.

AI or bust. Right now, AI is what everyone is talking about, and for good reason. After years of seeing AI doled out to help automate the processes that make businesses run smarter, we’re finally seeing AI that can help the average business employee working in the real world. Generative AI, or the process of using algorithms to produce data often in the form of images or text, has exploded in the last few months. What started with OpenAI’s ChatGPT has bloomed into a rapidly evolving subcategory of technology. And companies from Microsoft to Google to Salesforce and Adobe are hopping on board.

What started with ChatGPT has bloomed into an entire subcategory of technology with Meta, AWS, Salesforce, Google, Microsoft all racing to out innovate and deliver exciting generative AI capabilities to consumers, enterprise, developers, and more. Exploring the rapid progress in the AI space.

We don’t learn by brute force repetition. AI shouldn’t either.

Despite impressive progress, today’s AI models are very inefficient learners, taking huge amounts of time and data to solve problems humans pick up almost instantaneously. A new approach could drastically speed things up by getting AI to read instruction manuals before attempting a challenge.

Continue reading “An AI Learned to Play Atari 6,000 Times Faster” »