Elon Musk and Sam Altman are worried that artificial intelligence will take over the world. Therefore, two entrepreneurs create a non-profit organization to a billion dollars, which maximizes the power AI and then share it with anyone who wants. At least that is the message of Musk, founder of electric-car maker Tesla Motors, and Altman, President of startup incubator Y Combinator, announced during the presentation of his brainchild OpenAI.
In an interview with Steven levy of Backchannel, timed to coincide with the launch of the company, Altman said that they expect this project for decades will surpass the human intellect. They want the risks associated with artificial intelligence, had been mitigated by the fact that this technology “will use every, and not, say, just Google”.
Naturally, Levi asked the question, their plans to freely share the technology to the fact that it falls into the hands of bad people who will become “doctors of Evil” local cut. But the risk, according to the founders of the OpenAI, is reduced due to the fact that the power of many outweighs the power of a few. “Just as people are protected from Dr. Evil at least that most people are good, and the collective power of humanity can contain bad elements, says Altman. — But we think it much more likely that many of the AI will stop random bad people”.
Will be years before we find out whether these conflicting arguments. Artificial intelligence superhuman level terribly far, if at all. “This idea contains intuitive appeal, says miles Brendaj graduate of the University of Arizona, which deals with human and social dimensions of science and technology, about OpenAI. — But any case it does not. At the point where we are now, none of the AI is not able to capture the world — and will not in the foreseeable future”.
But the creation of OpenAI involves more considerations than simply the ability to capture the world forces of superhuman intelligence. In the short term OpenAI can directly help Mask and Altman and their companies (of Y Combinator, has raised such monsters as Airbnb, Dropbox and Stripe). After smahivanija the leading AI developers from companies like Google and their employment in OpenAI, these two entrepreneurs will receive access to ideas that had no access previously. And together, flooding each of your company data, they can understand the meaning of these ideas. In our days one key to the development of AI engineers, while the second data.
If OpenAI will remain true to its mission to provide everyone access to new ideas, it will serve the purposes and powerful companies like Google and Facebook. With Musk and Altman and others, who pours more than a billion dollars in this enterprise, OpenAI demonstrates how the concept of competition in recent years. Increasingly, companies, investors and entrepreneurs hoping to compete with rivals by giving its technology. This is to talk about contradictions.
The advantages of openness
OpenAI is the culmination of an extremely busy month in the world of artificial intelligence. In early November, Google announced the source code (partial) of the software engine running the AI company — technologies for deep learning, which has proved to be extremely capable of image recognition, the definition of words, translation of languages and understanding natural language. And just before the announcement OpenAI Facebook revealed the source code of the project computer server that runs its services deep learning, and many of them solve the same problem that technology Google. OpenAI now and promises to share all that will be able to make, and special attention will be paid, of course, deep learning.
Yes, such a division is a sort of rivalry. If a company like Google or Facebook open in sharing the device software or hardware, it boost the progress of AI in General. And this, again, is in its interest. As a community improves, these open technologies, Google and Facebook can bring these improvements into your business. In addition, opensourcing — method of finding and recruiting talent. In the field of deep learning scientists — many of whom come out of academies — welcome the idea of the open division of their work, meet large numbers of people. “There is definitely the competition for scientists, said Altman WIRED magazine. — The people we hired… they like the fact that OpenAI is open and they can share their work.”
This competition can be straighter than you think. We can assume that Google opened the source code of its engine AI, TensorFlow, because I knew that there will be OpenAI — and Facebook shared the design of its servers to Big Sur in response to the actions of Google and OpenAI. Facebook said that it is not. Google did not respond to the request. And Altman does not want to discuss it. But said that Google knew what was going to appear OpenAI. And how could she not know? This project took Elijah Sutzkever, one of the leading developers of AI.
This does not diminish the value of the project Google open source. Regardless of the motives of the company, this code is available for all users, may use as they please. But it is worth remembering that in the modern world distribution technology is more than just generosity. The deep learning community is relatively small, and all these companies are fighting for talents, which will help them to take advantage of this incredibly powerful technology. They want to share, but also want to win. They reveal their trump cards, but not all. Open source will accelerate the progress of AI, but until that happens, it is important that no one company or technology does not become too powerful. Therefore, in OpenAI makes sense.
His own Apollo program
To some extent it is possible to suspect that Musk sees himself and the exchange splitting as a way to win. “As you know, I had some concerns about the AI for some time,” he told Backchannel. And his public shiver on the topic of AI Apocalypse are well known. But he is also working on Tesla, which will benefit from technologies that will develop OpenAI. Like Google, Tesla is building driverless cars that will be able to derive real benefit from the technology of deep learning.
Deep learning relies on the so-called neural network, an extensive network of software and hardware that try to be like the network of neurons in the human brain. Feed enough pictures of cats neural network, and it will start to get to know the cats. Download it enough dialogue, and she will learn to hold a conversation. Feed it enough data about what is facing the car when driving on a road, and how drivers react, and she will learn to drive.
Yes, the Masks could just hire AI researchers to work in the Tesla. And hires. But with OpenAI he can hire the best researchers (because it is a public company and because it is not limited to the usual business model of a traditional company or short-term interest). It can even pull scientists from Google. In addition, it can create much more powerful data pool, which will help in the work of these researchers. Altman says that companies Y Combinator will share their data with OpenAI, but this is not enough. Align these data with Tesla, and will begin to compete with Google — at least under certain conditions.
“In some ways this is good, and in some ways not, said Chris Nicholson, CEO of a startup Skymind deep learning, which was recently accepted into the Y Combinator program. — I’m sure there’s a lot of data about Airbnb accommodations, which Google can’t reach”.
Musk was an early investor in a company called DeepMind — a British company that describes itself as the “Apollo program for the AI”. And this investment has opened for him a window on the development of this wonderful technology. But Google bought DeepMind, and now the window is closed. Now Musk begins his own Apollo program. And he’s back on the train. Other investors OpenAI are in the same position, including Amazon, the Internet giant, which follows on the heels of Google and Facebook in the race the AI.
Pessimism and optimism
No, this does not detract from the value of the project Mask. He can be selfish or altruistic motives. But the end result will be extremely useful for AI all over the world. In the process of sharing its technology with the world, OpenAI will force Google, Facebook and others to do it if not forced. This is good for Tesla and all companies Y Combinator. But also it’s good for everyone who is already interested in using AI.
Of course, in exchange for its technology, OpenAI will also provide new ammunition for Google and Facebook. And Dr. Evil, wherever he was hiding. He can make good profit at the expense of OpenAI. But the biggest concern is not that Dr. Evil will turn this technology into a universal evil. And that technology get lost in the world. Deep learning will not end on driverless cars and understanding natural language. Leading scientists believe that with proper mixture of data and algorithms this understanding can be extended and on what people call common sense. And even superhuman intelligence.
“Fear of super-intelligence, which recursively improves itself, reaches escape velocity and becomes an order of magnitude smarter than any human, says Nicholson. — This will not happen soon. And some believe that will not happen at all. But if it happens, will be awesome”.
With this Mask and solgar and try to fight. “Development, enrichment, availability of technology to protect people, says Altman. — To do this — the best way to protect us all.” But at the same time, it shortens the path to superhuman intelligence. And although Altman and Musk can assume that each access to the super-intelligence allows you to detect a runaway AI can be quite the opposite. If companies will know that everything in the world aspire to have highly intelligent AI, they will be less to take care of the precautions.
How necessary these measures are, oddly enough, depends on how optimistic you look at the ability of mankind to increase technical progress. Based on their past successes, Musk and Altman have every reason to believe that the arc of progress will be pointing up. But others do not particularly believe that AI will pose a threat to humanity, as it is seen by Musk and Altman. “Thoughts about AI is the opium for technologists: they excite us and make paranoid,” says Nicholson.
In any case, “Google” and “Facebook” of our world quickly push AI to new horizons. And, at least in the details, OpenAI will help keep them — and everyone else — in check. “I think Elon and the group believe that AI is not stopped, says Nicholson. — So all they can hope for is to change its trajectory”.