AI in 2020
This article is a reupload of an article posted to our old blog, posted here for completeness. Readers may find it useful as a recap of some of the headline AI events of 2020.
Undoubtedly you have some strong opinions about 2020. It’s not easy to see the positives, while surrounded in all directions by headlines detailing the raging pandemic, catastrophic wildfires, contentious elections, and all of the misery that lies in between. But in the background, AI research has been ticking along. The following are some of the biggest AI landmarks that you might have missed amongst the cacophony this year…
AlphaFold Solves 50 Year Old Problem in Biology
For decades, scientists have been trying to find a method to reliably determine a protein’s structure just from its sequence of amino acids. You see, proteins are a complex thing…
And there are over 200,000,000 known proteins, with 30,000,000 new proteins discovered every year. If we could work out the structure, such as the one pictured above, from any chain of amino acids, it would aid research in practically all facets of biology. This is called the protein folding problem.
DeepMind has gone and done just that. The deep learning system, AlphaFold 2, is an astounding breakthrough to one of the biggest problems in science. AlphaFold 2 can predict the shape of proteins to within the width of an atom, and will drastically improve scientists ability to design drugs and understand disease.
AlphaFold was used almost immediately to predict the structures of six understudied proteins in the SARS-CoV-2 virus genome.
For more information about AlphaFold, see the following links:
Warwick AI Interview: AI Protein Folding with Martin Lotz
DeepMind — A solution to a 50 year old grand challenge in biology
Technology Review — DeepMind protein folding AI solved 50 year old challenge in biology
Tesla Overtakes Toyota as Worlds Largest Car Manufacturer
It wasn’t long ago that electric cars were a non-option in most peoples minds. The short range, high price tag and frankly difficult-to-look-at exterior didn’t do them any favours.
Tesla has changed that. Innovation in all the aforementioned areas has rocketed Tesla to a market value of $210bn, overtaking Toyota in mid-2020. This comes in part due to the industry-leading innovation that we’ve seen from Tesla over the last few years, including but not limited to the long-awaited Roadster 2, with it’s mind-numbing 0–60 in 1.9 seconds, 1000km range and a frankly ridiculous 402kmph top speed.
This is big news not only for the EV scene, but for AI. Teslas fleet of over a million connected vehicles provides a constant torrent of data that is being used to continuously improve the neural networks behind its autopilot system.
It’s beginning to look like car companies that want to succeed are going to have to become AI companies too — those with the data and the intelligent systems will be the ones that race ahead, leaving archaic, disconnected, fuel-based cars in the 2020's.
Boston Dynamics Continues Making Strides in Smart Robotics with Spot 2.0
2020 saw the release of Spot 2.0 by Boston Dynamics, the home of smart robots. The release enhanced mobility, added new APIs for developers, and additional payload support, among other improvements.
Boston Dynamics said in a press release, that “Using these enhancements, users and application developers will have access to a broader variety of autonomous behaviors on Spot”.
The sale of Spot to private companies, and focus on adding APIs and support for developers, as well as a new focus on moving the development of BD Robots outside of their laboratories and towards a more distributed development approach, signifies an important shift toward intelligent, consumer robotics.
Spot is on sale for $74,500 USD, in an attempt by Boston Dyanmics to achieve its aforementioned ambition of making the development of the robot and it’s software a more distributed process. Thousands of developers using Spot for their own niche problems will result in much faster developments than a single team in the Boston Dynamics lab could manage.
Spot has found use already in a Boston hospital, interacting with covid patients, and measuring their vital signs. It’s also being employed in Singapore to encourage social distancing.
For reference, Spot 2.0 is the dog-like robot that appears around the 1 minute mark:
Read more about Boston Dyanmics:
CNBC: Spot the $75,000 Robot Dog is For Sale
