[ad_1]
Unbeknownst to them, in DeepMind, they were already studying an extensive scientific article defining their system. Nature, According to John Jumper, who leads the AlphaFold project. DeepMind sent it to this address Nature on May 11th.
At the time, the scientific community knew little about DeepMind’s timeline. That changed three days after Baker’s prepress became available, on June 18, when Demis Hassabis CEO of DeepMind went to Twitter. “We have worked diligently on the role of our entire methods (which we are reviewing) along with open source code and providing wide free access to AlphaFold for the scientific community,” he wrote. “Very soon!”
On July 15, Baker’s RoseTTAFold paper was released on the same day, Nature released by DeepMind unedited but revised AlphaFold2 manuscript. At the same time, DeepMind made AlphaFold2 code freely available in github. And a week later, the team release an a huge database among the 350,000 protein structures predicted by their method. The revolutionary tool for predicting proteins, and a large number of its predictions, were eventually in the hands of the scientific community.
According to Jumper, the reason for not releasing DeepMind’s paper and code is trivial until more than seven months after the CASP presentation: “We weren’t ready to publish open source or very accurate paper that day,” he says. Once the paper was submitted in May, and the team was working on the review process, Jumper says they tried to get the paper out as soon as possible. “Actually, we were pushing as fast as we could,” he says.
It was published through the DeepMind group manuscript NatureAccelerated Article Preview workflow, which the magazine frequently uses for Covid-19 articles. In a statement to WIRED, a spokesman said Nature he wrote that this process is intended “as a service to our authors and readers,” especially so that the reviews reviewed by experts who are worth noting and time-sensitive are available as soon as possible.
Jumper and Pushmeet Kohli, heads of the DeepMind science team, questioned whether Baker’s role had taken their time into account. Nature publication. “From our point of view, we helped and presented the paper in May, so it came out of our hands somehow,” Kohli says.
Organizer Moult CASP believes the work of the University of Washington team can help DeepMind scientists convince their parent company to make their research freely available in the shorter term. “The sense I have from knowing – they are great scientists – is that I would like to be as open as possible,” says Moult. “There’s a tension there, because it’s a trading company, and in the end you have to make money somehow.” The company that owns DeepMind, Alphabet, has the fourth largest market in the world.
Hassabis characterizes the release of AlphaFold2, both for the scientific community and for Alphabet. “All of this is open science and we give this to humanity, without tying the strings together, the system, the code and the database,” he told WIRED in an interview. Asked if there has been any discussion about keeping the code private for commercial reasons, he said: “The good question is how we value it. Value can be given in many different ways, right? One is, of course, commercial, but it is also prestige.”
Baker is fast praising the DeepMind team for their detailed release of their role and codes. In a sense, RoseTTAFold says DeepMind was a hedge against the possibility of acting in the spirit of scientific cooperation. “If they could have clarified less and decided not to [release the code], then at least it would be a world-based starting point, ”he says.
[ad_2]
Source link
