Antitrust inquiries into U.S. tech giants could upend the development of artificial intelligence.
Last week, lawmakers interrogated the CEOs of Apple, Alphabet, Amazon, and Facebook about whether their companies have become too powerful. House members drilled in on Big Tech’s acquisitions (they allegedly stifle innovation) and their collection of huge amounts of data (it gives the companies a huge advantage over rivals in developing A.I. and improving their products).
But even if lawmakers agree Big Tech is too big, they are in a quandary about what to do about it. Should they break the companies up? Fine them? Do nothing?
As antitrust expert Dakota Foster recently explained in a paper for Georgetown University’s Center for Security and Emerging Technology, what Congress decides could be critical to the federal government. Research and technology from major tech companies both directly and indirectly benefits the Pentagon. Tech giants often open source their A.I. research, which means that the federal government can use the findings for free. The companies also sell cloud computing and A.I. services to government agencies.
By taking action, lawmakers risk hurting the ability of tech companies to develop A.I., Foster warned Fortune. They may end up cutting spending into A.I. research, and thereby achieve fewer technological breakthroughs.
In the past, the government would have been easily able to dump Big Tech in favor of contractors like Lockheed Martin and Raytheon, Foster explained. But in recent years, the tech giants have leapfrogged the defense industry’s A.I. skills, she said.
Members from both political parties are concerned that slowing progress by Big Tech in A.I. may benefit China. That country is investing heavily in A.I., with the goal of becoming the world’s leader 2030.
Whatever the case, the federal government shifting to using smaller U.S. tech companies as an alternative to Big Tech isn’t particularly realistic. The data used by the upstarts for A.I. projects isn’t as complete as what the tech giants have. Furthermore, the small fry lack the money to pay for the tremendous amount of computing power required for A.I. projects. At the same time, they face a tougher time attracting the necessary talent, she explained.
Still, smaller businesses could get some lift if Big Tech had its wings clipped. As Foster said, “there’s definitely an argument to be made” that the mere presence of Big Tech is curtailing innovation by smaller companies.
“If Facebook was founded today would it become the Facebook we now know?” Foster asked. “Some people say it’s just not possible given the [tech giants’] presence.”
Jonathan Vanian
@JonathanVanian
jonathan.vanian@fortune.com
A.I. IN THE NEWS
Siri’s legal problems. Apple is being sued by Chinese tech company Shanghai Zhizhen Network Technology, also known as Xiao-I, over allegations that the Siri voice-activated digital assistant “violated a patent the Chinese artificial-intelligence company owns for a virtual assistant whose technical architecture is similar to Siri’s,” The Wall Street Journal reported. Shanghai Zhizhen is seeking $1.43 billion in damages and wants Apple to "stop sales, production and the use of products flouting the patent—a category that includes virtually all the U.S. company’s devices,” the report said. Apple told the Journal that it disputes the allegations and it looks forward to "presenting the facts to the court."
Rite Aid waves goodbye to facial recognition. Rite Aid has ended the use of facial recognition technology at hundreds of its stores throughout the U.S. after a Reuters investigation revealed the extent of the drugstore’s use of the controversial technology. The Reuters article said that Rite Aid “deployed the technology in largely lower-income, non-white neighborhoods” in New York and Los Angeles and that some of the facial recognition tools the company used for over a year were created “from a company with links to China and its authoritarian government.” From the article: Among the systems used by Rite Aid was one from DeepCam LLC, which worked with a firm in China whose largest outside investor is a Chinese government fund. Some security experts said any program with connections to China was troubling because it could open the door to aggressive surveillance in the United States more typical of an autocratic state.
This agriculture startup just raised millions. Farmers Business Network, a startup specializing in data aggregation and analytical tools for farmers, has raised $250 million from investors like BlackRock, Kleiner Perkins, and GV (formerly Google Ventures), Bloomberg News reported. Farmers using the startup’s service share information like the size of their field or how much they paid for chemicals in order to get a better picture of pricing trends in agriculture. “The new capital will be used to expand its seed and crop protection business, which aims to help farmers reduce their cost of production,” the report said.
A.I.’s call of the wild. Researchers from Cornell University are using deep learning to analyze elephant calls that were recorded in the wild, NPR reported. The team of scientists have accumulated about “a million hours of tape,” which, as you can imagine, is a “very, very slow, very tedious” task to analyze, said Peter Wrege, a Cornell University behavioral ecologist. From NPR: We feed these models hundreds of examples of both audio clips with and without elephant calls, and then these deep learning models are, basically, over time able to learn specific features that the people training these models don't fully know ourselves.
EYE ON A.I. TALENT
ByteDance, which owns the video app TikTok, said that Ma Wei-Ying would no longer be the head of the company’s A.I. lab, Reuters reported. Ma, who previously joined ByteDance from Microsoft, will join an unspecified A.I. research institute at Tsinghua University and become a “technological consultant for ByteDance,” Asia-Pacific news outlet KrASIA reported. Microsoft is currently pursuing an acquisition of TikTok from ByteDance, which has come under fire from President Donald Trump over national security reasons.
VIDA Diagnostics, Inc. hired Todd Johnson to be the health tech startup’s chief technology officer. Johnson was previously the CTO of Evident, a company specializing in dental technology.
EYE ON A.I. RESEARCH
Night night, neural network. Researchers from institutions like Los Alamos National Laboratory and Drexel University published a paper about techniques they developed around so-called spiking neural networks, a kind of deep learning system. They describe their techniques as akin to the way sleep restores brain functions in humans. Researchers from companies like Intel have been experimenting with spiking neural networks for the development of neuromorphic computer chips, which, theoretically, would be powerful A.I. chips that conserve power by only activating when they receive the appropriate signal.
As Popular Mechanics describes in a story about the new paper, the researchers developed their “resting” techniques to help stabilize the spiking neural network during unsupervised learning:
In their paper, they describe incorporating "epochs of sinusoidally-modulated noise that we hypothesize are analogous to slow-wave sleep."
That means to achieve some semblance of artificial sleep, the researchers had to inject noise into the neural network. They experimented with a few different types of static, but ultimately settled on Gaussian noise waves, which include an array of different frequencies and amplitudes.
Watkins and her colleagues believe their new algorithm, which induces the noise, can mimic the slow wave sleep that rejuvenates the biological neurons in our own brains. "It was as though we were giving the neural networks the equivalent of a good night’s rest," she said.
The researchers presented their paper during the Conference on Computer Vision and Pattern Recognition in June.
FORTUNE ON A.I.
Congress wants to curb Big Tech. It could end up crushing startups instead—By Patricia Nakache
The U.S. and EU’s key data-protection deal is dead. How one of the world’s biggest data brokers is adapting—By David Meyer
Facebook and Amazon grilled over history of aggressive competitive practices at antitrust congressional hearing—By Danielle Abril
Google’s upcoming Grace Hopper subsea cable will span the Atlantic Ocean—By Jonathan Vanian
BRAIN FOOD
Drug discovery is very difficult and not even GPT-3 can help. Drug and pharmaceutical expert Derek Lowe wrote in his “In the pipeline” commentary about the limits of machine learning to aid the task of drug discovery. As he explains, machine learning doesn’t work well when it's applied to big problems that lack appropriate data, and he cites the classic “garbage in, garbage out” adage to highlight his point.
Unfortunately, when it comes to drug discovery, there’s “just a lot of things that we don’t know, and sometimes we are destined to find out about them very painfully and expensively,” he writes.
He explains that while research group OpenAI’s GPT-3 language technology has been receiving much hype as of late, the technology, while genuinely novel, is likely not going to uncover clues in DNA, for instance, if someone actually gave it DNA data to analyze. The text that OpenAI was trained on was ultimately created by humans, thus imbuing the information with some form of “human purpose and some level of intelligence,” he explains, inferring that this helps the language tech pick up on patterns within text data. But biological systems like the human brain or even the tiny fruit fly are “fundamentally different” and more complicated than anything people have ever created.
He writes:
But what happens if you feed a pile of (say) DNA sequence information into GPT3? Will it spit out plausible gene sequences for interesting new kinase enzymes or microtubule-associated proteins? I doubt it. In fact, I doubt it a lot, but I would be very happy to hear about anyone who’s tried it. Human writing, images that humans find useful or interesting, and human music already have our fingerprints all over them, but genomic sequences, well. . .they have a funkiness that is all their own. There are things that I’m sure the program could pick out, but I’d like to know how far that extends.
And even if it really gets into sequences, it’ll hit a wall pretty fast. There’s a lot more to a single living cell than its gene sequence; that’s one lesson that have had – should have had – beaten into our heads over and over. Now consider how much more there is to an entire living organism. I’m all for shoveling in DNA sequences, RNA sequences, protein sequences, three-dimensional protein structures, everything else that we can push in through the textual formatting slot, to see what the technology can make of it. But again, that’s only going to take you so far.
