Hey guys! Hope that you are all doing well!
Sorry I haven’t been regular recently, I had to prepare for an important exam for my Ph.D.
Today, I wanted to talk about the latest news in Artificial Intelligence (AI). So let’s start!
1- Google just gave control over data center cooling to an AI
Yes, you read it right, Google just applied an AI to several of its data centers to automatize its cooling system. It is estimated that the AI could help to save up to 40% of energy consumption, and thus generate millions of dollars in energy savings [1]. The system is still supervised by a manager in action, just to make sure that the algorithm does not aim for something too risky. The AI algorithm used to achieve this prowess is called “Reinforcement learning”, an algorithm that is used to train bots in video games (DOTA 2, GO, Mario games etc …) but can be reused in many different domains: Financial trading, web navigation, neural network architecture search etc …
My opinion about this: Reinforcement learning once more shows how useful it can be for the humanity. This algorithm that automatizes cooling systems could become more democratized and thus applied in different areas such as Blockchain data centers or Hospitals. This is very promising.
2- DeepMind just developed an algorithm to spot eye disease
DeepMind is an Artificial Intelligence company that is owned by Google which is mostly famous for having created an AI (AlphaGo) that was able to beat Lee Sedol at Go. Here, they developed an AI Algorithm in collaboration with Moorfields and the University College London that can recommend the correct referral decision for over 50 eye diseases with 94% accuracy [2].
My opinion about this: DeepMind has been investing a lot of ressource into health research. Mustapha Suleyman who is one of its founder made it really clear that some of the long term goals of DeepMind was Ethical AI and Health Research. DeepMind’s results are actually very promising.
3- OpenAI Five is already at the Internationals (Vancouver, Canada)!!
OpenAI Five, the AI that was created by the OpenAI engineers and scientists (using Reinforcement learning) is going to participate in the Dota 2 Internationals Competition that is currently happening!! OpenAI engineers and scientists just arrived in Vancouver yesterday [3]. There team has been developing that AI for months now. They even held a match again Dota 2 professionals a few weeks ago, winning 2 by 1. If you are curious about how to watch their games, just go to the Twitch OpenAI channel [4] or follow Greg Brockman on Twitter [3].
My opinion on this: To my knowledge, this is unprecedented in the history of MOBA games. It is amazing that OpenAI has been able to get this far in just a year. This is a major event not only for the AI and the gaming community, but surely for the entire world, as the algorithms that are used to train OpenAI five can be reused into so many other domains and make a positive impact to many people’s lives.
4- OpenAI just trained a humanoid hand to manipulate objects with unprecedented dexterity
OpenAI just trained a human-like robot hand called Dactyl to use its hand to solve real world tasks [5], using the Reinforcement learning algorithm. The task was clear: Dactyl had to reposition a block into a different orientation (e.g. change the position of the block to put a new face on top). This had been a long difficult problem in robotic control.
My opinion on this: In my humble opinion, more applications like this will be made to create robots that can achieve repetitive tasks.
5 - A team from the University of California Beckerley just developed a more “corteous” self-driving car.
Usually, self-driving cars tend to take extreme rational decisions, which can end up being either too conservative or too aggressive, and thus not taking into account the other driver’ s psychology [6]. This team from the University of California Beckerley just developed a self-driving AI algorithm that can have a more corteous behavior towards humans and anticipate their actions for instance when changing lanes or turning left.
My opinion on this: This is a very good news to improve self-driving cars compatibility. Car accidents are the number one cause of death in the world so anything that can be done to make the self-driving cars the safest possible is a fantastic move in my opinion. I think that if the self-driving cars were to use the Li-Fi (Internet with Light) technology, they would be much more safer and cheaper in my opinion than now.
Ok that was it for today guys! Hope that you all enjoyed reading the post! I will see you guys very soon for a new post! :)
References:
[2] https://twitter.com/i/moments/1029803254144819200
[5] https://blog.openai.com/learning-dexterity/
[6] https://arxiv.org/abs/1808.02633
Pictures
Picture 1 was taken from https://goo.gl/jTTKzv
Picture 2 was taken from https://goo.gl/saEkx7
Picture 3 was taken from https://goo.gl/2hmbJL
Picture 4 https://goo.gl/D8bVWH
Picture 5 was taken from https://goo.gl/UdHtX2
Picture 6 was taken from https://goo.gl/bEXyUa
juv79505 sincerely thank you for reading this article. Please feel free to comment below in case you have ideas, questions, suggestions or simply want to criticize this article. Also, note that all pictures used in this article were extracted from the google section “pictures labeled for commercial reuse”. Stay tuned for more articles on health, environment, artificial intelligence, video games, technology in general, books, graphic novels, geography, history, sports and much more!