Earlier this year, Open AI, an organization fueled by investments of technology bigwigs like Elon Musk (founder of electric car Tesla) and Sam Altman (Y Incubator president), was publicly revealed for the first time, at the end of a major AI conference in Montreal.
Even before the announcement, but also during the conference, tremendous amounts of money had been offered to Open AI’s brilliant body of researchers. It came very specifically from Silicon Valley companies that were fearful of losing their top talents to a company that was being put together by two of the most visionary minds in their market.
Instead of matching the money offers, however, Open AI did something more appealing: They offered the opportunity to research the future as opposed to being concerned with products or financial results. Further, they offered the freedom for sharing all that research with anyone interested. When we talk about giving away for free what may become the greatest paradigm-changing technology in this century, it becomes pretty major.
In fact, one of those who has received these absurdly immense money offers says that it actually sounded to him as an attempt to prevent Open AI’s mission, and that it has turned him down.
The new organization’s unselfish undertaking was clearly something much more worth being part of. Here’s the paradox: the more companies try to retain professionals like these, the more they want to be free to share. Instead of keeping knowledge to themselves, they want to improve the world by sharing it. In this case, their single goal is to make Artificial Intelligence better. That is indeed a 21st century paradox.
Later on, Open AI launched its first product dedicated to perform benchmark testing on artificial intelligences. Called Open Gym AI, the software will be used to understand and measure how a machine can handle learning in order to use the numbers obtained to create a kind of ranking for electronic brains. This is Open AI’s first move in their mission to become a great force pushing AI to as far as it can go. From the way the company has been created to its goals, it’s easy to see the dimension of the revolution that lies just ahead of us. When we talk about revolution it’s more than a technological change, but rather the change in the ways technology is created.
And when it comes to AI, the change is real. We already see companies making use of what is now called Deep Learning in systems that recognize faces in photos, voices in smartphones and in the way our searches on the internet are currently responded. This technology is the germ of a new era for automated devices that will not only be able to understand human thinking and perform human actions, but moreover, will be able to learn on the go. We are talking about thinking. Just like humans do.
Open AI was bred as Sam Altman’s effort to create a new AI laboratory that could function outside of the control of the big technology corporations. In order to do so he started by hosting an informal dinner at a Silicon Valley hotel. He invited an eclectic but prodigal group of AI enthusiasts for this meeting in order to share his goal: to help humanity build AI in a safe way.
After Open AI Gym was launched it becomes clear how he’s been successful in sharing his intention. According to his partner Musk, extremely intelligent machines can be much more dangerous than atomic bombs. So the first software created by Open AI aims exactly at that point, by testing the ability of a machine to analyse what it must absorb and what it must ignore. That’s clearly not far from our concepts of good and evil, that we all learn by growing up in society.
That should prevent blunders like the Microsoft AI.
Among the AI enthusiasts who were at the meeting, many had very specific interests. Musk needs AI for his Tesla vehicles to self drive and for his other company Space X to keep humans alive in space. But like we’ve seen, Musk’s enthusiasm doesn’t make him blind to the danger of self learning systems getting out of control.
Open AI’s first obstacle was exactly the fact the most people qualified enough to contribute to its goal was already working for Google and other giants. A young but experienced company builder named Greg Brockman was the first one to join the team. He was so moved by the hotel meeting he embraced the task of building the laboratory as a full time project and thus approaching other masterminds in the AI field.
In order to break any resistance knot, he invited his ten favored experts to wine and dine in a winery in Napa. According to Brockman, that Saturday in Napa was the great catalyst to the project. The vibe and the wine helped. After the get-together he asked the ten guests to join the company and gave them three weeks to think. Nine were in before the deadline and they all stuck with the company despite the humongous money offers.
AI is real but it is still young. So a lot of the researchers that had left the academic field to join the tech giants soon realized that researching alone and in secret would get them to fall behind.
So even Google is being forced to share its AI resources and Open AI is making its statement by pushing the sharing as far as they can. And they have been doing so mostly by example. Openness, however, has limits. Open AI is a non-profit but it isn’t charity, and despite not paying its employees the same huge amount of money as Google or Facebook, it still has its own obligations like any other company. Open AI has sister companies, more specifically the ones its directors own, like Tesla. So in order to compensate some huge disparities it offers stock options in those companies.
It’s a hard effort to avoid Open AI to become a consulting company for its sister companies however. But Brockman makes sure he does all he can so that it doesn’t happen. But he acknowledges that not everything at Open AI is done in the open. According to him, the processes and the ideas need to be nurtured until they mature before getting published. He also says that there’s no set in stone commitment to patent-free. That means that if in the long term they realize that patenting some of their work is the right thing to do, they wouldn’t hesitate. Despite not planning on patenting anything in the short term, Open AI could engage pre-emptive patenting, a strategy that prevents others from securing patent.
Besides openness and to be able to do research outside the market, Open AI is concerned with the threats of powerful artificial super intelligence. And that has everything to do with openness because bad actors could snatch technologies that aren’t proved to be safe. Open AI is working in a way that it doesn’t happen, by ranking artificial intelligence and providing it with a sort of consciousness, it aims, already via its first software, to save us from rebel automatons with malicious tendencies in the future.
