This week - open source protein-folding AIs; the designer babies are now commercially available; a fully robotic pizzeria in Paris; and intro to AI safety; and more!
More Than A Human
There is a company in the US that offers genetic screen checks of embryos and IVF not only for couples at risk of passing on genetic diseases but also for healthy couples. The company has backing from some big names in Silicon Valley but scientists are not fully sold on their methods. Plus the whole thing raises ethical questions about designer babies.
After two years of consultation, WHO release Ethics and governance of artificial intelligence for health report which proposes six principles to ensure AI works for the public interest in all countries while minimalising the risks and maximising the opportunities for AI in public health.
The need for more powerful and energy-efficient makes computer designers look beyond silicon chips. One of the new emerging ways of building computers is photonics which uses light for computing and, theoretically, could accelerate deep learning by several orders of magnitude.
When quantum physicist Mario Krenn first applied a machine-learning algorithm to quantum physics, it generated new experiments no one thought about before. Now the updated version of that algorithm, with more computing power and better output, is helping to push the frontiers of quantum physics.
Robert Miles published an updated and cleaner version of his Introduction to AI Safety talk given back when we could meet in person. Anyway, Robert does a good explaining the basics of AI safety and why should we care about it.
The new pizzeria ‘Pazzi’ is staffed entirely by robots, from order-taking to prepping the dough to putting the pizzas in boxes.
Cool idea but as someone pointed out in comments, brushed motors and gas leaks is an explosive mix.
DeepMind published the source code for their protein folding algorithm AlphaFold 2 on Github for free for everyone to download, then follow the setup instructions and run. DeepMind also released a paper in Nature explaining how AlphaFold works.
While everyone was waiting for DeepMind to release the source code for AlphaFold, some researchers took the matter into their own hands and created their own protein-folding AI - RoseTTAFold - and made it publically available for everyone to download and try out.
This article describes how new ways of growing food will make farming as we know it obsolete, cut down the impact of agriculture on the environment and, as George Monbiot put it, "soon be able to feed the world without devouring it".