Roundup Welcome, friends. Here's your human-generated, totally not computer written, summary of this week's AI news, beyond what we've already covered. In short: Elon Musk steps down from OpenAI's board, Uber is looking to train new coders in machine learning, and there's a new AI conference.
FAT conference – The inaugural Conference on Fairness, Accountability, and Transparency, presenting research about bias in algorithms and machine learning kicked off today in New York.
It’s a positive sign that experts are conscientious of the possible negative impacts of AI, and provides an opportunity for computer scientists and social scientists to work together. What started as small workshop at the Neural Information Processing Systems (NIPS) conference in 2014, has now grown to its own event with about 500 attendees, and another 400 people on the waiting list.
Some of the topics covered so far include discrimination on online advertising, privacy, and interpretability and explainability of algorithms. We'll also be filing stories as they happen.
FAT is a two-day event, and for those that didn't manage to nab a ticket there’s a livestream here.
Farewell, Elon! - SpaceX and Tesla supremo Elon Musk has stepped down as chairman of OpenAI, the nonprofit boffinry nerve-center focused on reinforcement learning he cofounded two years ago.
This move is to - apparently – “eliminate a potential future conflict” as his other biz, automaker Tesla, becomes increasingly focused on AI in order to produce self-driving cars. He will, however, continue to donate funds to OpenAI and advise the organization, according to an official blog post.
Musk is well known for his out-of-this-world alarmist tweets about killer robots taking over Earth (maybe it’s to fuel his other company SpaceX, which wants to start a human colony on Mars), so maybe leaving the board will relax fears a little – but perhaps not.
OpenAI also announced that a gang of entrepreneurs, VCs and philanthropists are becoming donors, and there are a few new advisors too. You can find out more in the blog post link above.
MADDPG – More OpenAI news. It has released code for an algorithm that explores multi-agent environments in machine learning.
The algorithm codenamed, MADDPG, is based on work that was published last year in June. It explores how agents can be programmed to either compete or collaborate with one another on a specific task.
It’s tricky to train multiple players effectively. So MADDPG introduces the role of “actors” and “critics” for its agents.
“We treat each agent in our simulation as an “actor”, and each actor gets advice from a “critic” that helps the actor decide what actions to reinforce during training. Traditionally, the critic tries to predict the value(i.e. the reward we expect to get in the future) of an action in a particular state, which is used by the agent - the actor - to update its policy,” it explained in a blog post.
You can play around and create your own games here.
Uber AI residency – Uber is launching its own AI residency program, a one-year research gig geared towards applying machine learning to its ride-hailing business.
“Residents will have the flexibility to pursue a range of different directions in research and application. Some projects might involve fundamental AI research, pushing the frontiers of the field by developing new algorithms for learning and control, while others might include devising and training new models to help more efficiently transport people and things in the physical domain, improving real world user experiences,” it explained in a blog post.
AI is a relatively new and emerging field, and talent is scarce. Many researchers and developers are lured by the interesting work and generous salaries, but there aren’t enough of them around to cope with the current boom in AI.
So companies are now introducing these swanky AI residencies in the hopes of training and maintaining new devs.
Google, Facebook, OpenAI and Microsoft have similar programs. Uber haven’t set strict requirements and said it welcomes “candidates of all academic and geographic backgrounds”. Work visas will be assessed on a case-by-case basis for those who are accepted. ®