There has never been a technology with so much potential as AI. And this makes it crucial for nations all over the world to pay attention to it. Thus, different stakeholders – individuals, governments, startups, nations – are competing in order to develop and control general AI. The race is on and the question is who will be the first.
We spoke to Olga Afanasjeva and Will Millership from GoodAI who have recently launched the “Solving the AI Race” round of the General AI Challenge about the hope they have for the competition and how it is a great opportunity for new participants to take part in the AI revolution.
Olga explains that if AI is controlled by a small power group, the benefits will more likely be for just a few people. The GoodAI’s vision is to create a universe which is beneficial for all. “Even a small group of people developing something can make a very big change in the world,” Olga says. “At the moment, we are experiencing AI boom and getting the funding for a startup in AI is much easier than a few years ago. It’s a great possibility to enter into AI with a clear goal in mind, be it various applications or basic research.”
A lot of tools (check out the list of some useful links at the end of the article) are easily available for developing specialized AI applications, but the steps for developing a general purpose AI system are not as clear. Will believes that very few companies are focusing on general AI solutions. “It is because of the high risk,” he explains. “No one can tell when general AI will be fully developed, but at the same time the return on investment can be unprecedented.”
“Focusing on a long-term goal of general AI can be harder as a developer or an entrepreneur might find themselves demotivated by a clear vision of short-term results. Some investors can also be reluctant to invest in something that is uncertain,” Olga continues.
To lower down the risks of the general AI development for just a small group of people and in order to engage developers from all over the world, GoodAI is running the General AI Challenge. The first round ended in 2017 with four finalists who were jointly awarded USD 7,000. To see the summary of the round, click here.
On January 18th, the second round of the Challenge is launched with the working title AI Race Avoidance. The goal is to address the potential pitfalls of a race of transformative AI (general or specialized), where key stakeholders, including the developers, may ignore or underestimate safety procedures, or agreements, in favor of faster utilization and the fruits of the technology might not be shared by the majority of people to benefit humanity, but only by a selected few.
Olga explains that the vision of the Challenge is to solve the crucial milestones on the way to beneficial general AI. “The challenge will provide an opportunity to scale up our own research process,” she says. “We are interested in building general-purpose algorithms and solving the risks associated with AI. It is probably unlikely that the global race for AI can be totally avoided. A competition is a good stimulus after all, and what we can do is to develop robust mitigation strategies for the risks associated with the race dynamics.”
“We believe our upcoming round of GoodAI’s General AI Challenge can help us solve this through citizen science. We are not looking for one ultimate solution, but for a range of studies, solutions or proposals, that would help us better understand the actors and the dynamics of the race from various perspectives, and suggest concrete steps of what can be done to ensure beneficial future for all.”
Over the coming months, the General AI Challenge team will be working with all the advisors and partners to finalize the specifications of the round. You can read the report from the AI Race Avoidance workshop in Tokyo, which will act as a good starting point for participants. More specifications of this round of the Challenge can be found here.
The challenge is open to people who are already working in AI, as well as to the general public, or anyone with an interest in AI. Will says that they want to look not just for the AI experts, but for people who haven’t been looking in this area at all. “We want to encourage the interdisciplinary approach. People who have never done AI might have a step up to offer a completely new perspective on it. This will affect society as a whole. It would be great to listen to ideas of policymakers, economists, and people from across a variety of disciplines who might have a lot to say on this.”
The contender has to register on the website and submit a report (a white paper) along with a summary. An interdisciplinary panel of experts will then narrow down the applications received and vote for the top submissions. The USD 15,000 prize money will be shared among the finalists.
Why not give it a shot? Be ambitious!
List of tools for AI:
https://github.com/oxford-cs-deepnlp-2017
