Issue #102

This week, Google is holding Google I/O conference, so the Artificial Intelligence section is dominated by the news from Google. Apart from that, OpenAI announced a big research breakthrough in machine learning. We will also pay a visit to MIT robotics lab, talk about CRISPR, augmenting humans and more!


More than a human

Genetically engineered humans will arrive sooner than you think. And we’re not ready.

Michael Bess speaks with Vox about bioengineering and augmenting humans. Why it will be a huge thing, the pros, cons, fears and hopes.

 

Bionic Man

Here’s a talk by Dario Farina which he gave during the Imperial Festival 2017. He will be showing off the latest advances in robotic limbs, before discussing the obstacles to further progress, and imagining a future of not just amazing bionic limb performance, but also ethical dilemmas such as those raised by the possible demand for bionic limb over own healthy ones.

 

Artificial Intelligence

Google.ai

Google published this website to showcase their AI projects and tools they use. Right now there isn’t anything fancy there. Just a list how Google uses machine learning in their day-to-day operations, what tools they use and links to relevant articles about AI scattered across Google research blogs and news articles.

 

Build and train machine learning models on our new Google Cloud TPUs

Google announced that they will offer the second generation of Tensor Processing Units in Google Cloud. It means that if you have a machine learning system you can leverage Google’s infrastructure and run your algorithm on the same hardware as Google does their services.

 

Google’s AI Invents Sounds Humans Have Never Heard Before

Researchers from Google Magenta, a team of AI researchers and Google Brain’s off-shot focusing on making machines capable of doing art, showed their latest project called NSynth. It’s a neural network that synthesises sounds which it learned a massive database of sounds and can recreate the sound of any instrument and even build new sounds from scratch. You can listen to them in the linked article.

 

An Algorithm Summarises Lengthy Text Surprisingly Well

Guys from Salesforce created an algorithm that generates quite good TL;DR versions of articles. It can be pretty handy when you are overloaded with information from everywhere, but still need to know what’s going on around you.

 

Robotics

Checking Out New Robots Inside MIT

I wish I have such a lab. So. Many. Awesome. Robots!

 

Robots that Learn

Researchers at OpenAI have created a robotics system, trained entirely in simulation and deployed on a physical robot, which can learn a new task after seeing it done once.

 

Microrobots inspired by nature

I like to see how nature inspires robotics. Here, for example, a team of researchers from South Korea have designed microrobots that mimic the rowing action of the cilia present on the single-celled Paramecium; an organism common in stagnant water and ponds. The result – much faster movement than conventional microrobots.

 

3D printed, soft, four-legged robot can walk on sand and stone

Engineers at the University of California San Diego have developed the first soft robot that is capable of walking on rough surfaces, such as sand and pebbles. The 3D-printed, four-legged robot can climb over obstacles and walk on different terrains.

 

Biotechnology

CRISPR Creator Jennifer Doudna on the Promises—and Pitfalls—of Easy Genetic Modification

Jennifer Doudna, one of the creators of CRISPR, takes eight possible uses for the technology she helped created and gives her opinions about them in a couple of sentences.