The last three years have been good for Michael Kratsios’ career.
In 2017, he left his post as chief of staff for controversial Silicon Valley investor Peter Thiel to become the White House chief technology officer. Last month, he became the fourth most powerful person at the Pentagon.
Kratsios, now acting undersecretary of defense for research and engineering, is in command of a $60 billion Pentagon budget and tasked with keeping the United States’ tech ambitions ahead of China.
“Fundamental to preserving our strategic superiority and our way of life is to retain America’s technological dominance,” Kratsios said in a virtual event on August 13, held by the Georgetown Center for Security and Emerging Technology.
With Kratsios in the Pentagon, a small piece of Thiel’s mindset is now installed in the cornerstone of the nation’s defense apparatus. The new undersecretary’s refrains aren’t dissimilar to what Thiel has called for in the past: fewer regulations on tech and closer collaboration between Silicon Valley and the Department of Defense.
He has also found ways to skirt regulation when testing new technologies. The Pentagon is testing 5G on military bases, leap-frogging the permits and partnerships that private companies like Verizon and AT&T have to navigate.
Kratsios is a relative tenderfoot to government work. His colleague, undersecretary of defense for intelligence Joseph Kernan, graduated from the U.S. Naval Academy a decade before Kratsios was born. But Kratsios’ relationship to the new cadre of government tech contractors, like Thiel-backed Palantir and Anduril, sets the stage for a new era of collaboration between the tech industry and the military.
As Kratsios said in his Georgetown remarks, his job is to make sure that the people working on A.I. collaborate with the people designing hypersonic missiles. And my bet is that Thiel might have some suggestions on where to find them.
And now, here’s some of the most interesting A.I. research of the week.
Algorithms that can change the mood
Two research papers tackle a similar problem this week. Google researchers take a portrait of a person, and try to get an algorithm to generate images of the same portrait with different lighting. Other Google researchers (with help from the University of California, Berkeley and startup Humen), do the same thing with images of a whole city by training an algorithm to generate images of the same building at different times of day with different light.
A.I. that responds to constructive criticism
Italian, French, and Chinese researchers have built an image-generating algorithm that can take written feedback and use it to better accomplish a task. If the algorithm generates an image of a bird, it can be told to change the color of the bird’s breast, and it will be able to do it. Researchers also demonstrated being able to add a beard and subtract a smile from the face of the actor Mario Lopez.
A faster way to make cartoons
Microsoft and two Hong Kong Universities have built a way to automate the creation of cartoons. The algorithm takes two images and then is able to generate the frames of animation in between them.
Read the original article at OneZero.