Billionaire tech entrepreneur Tom Siebel: The pandemic will “clear the silliness out of the market”
This is the web version of Eye on A.I., Fortune’s weekly newsletter on artificial intelligence and machine learning. To get it delivered weekly to your in-box, sign up here.
Last week, I spent some time chatting with Tom Siebel, the billionaire entrepreneur who runs C3.ai, which sells big data analytics and A.I. software to some of the world’s largest businesses. It counts Royal Dutch Shell, Bank of America, 3M and the U.S. Air Force among its customers.
Siebel is known for being outspoken and, in our wide-ranging discussion about today’s state of A.I., he didn’t disappoint.
On pandemic data:
“No one here has any idea what they are talking about,” Siebel says of the U.S. response to Covid-19. And he includes many of the world’s leading epidemiologists and infectious disease researchers among “no one.” The problem, he says, is that the coronavirus response has become so politicized and polarized that it is impossible to find an objective analysis of the data.
Siebel enlisted C3.ai in creating the world’s largest Covid-19 “data lake”—essentially a giant compendium of datasets. The lake, hosted on AWS, contains 22 big datasets, including the Allen Institute for Artificial Intelligence’s set of papers that my colleague Jonathan Vanian highlighted in this newsletter a few weeks ago as well as data from the World Health Organization, World Bank, U.S. Census Bureau and Johns Hopkins University.
Unlike some similar attempts at pooling coronavirus data, C3.ai created what is called a “unified, federated” dataset. Using automated tools it originally developed to handle the massive datasets of customers, C3.ai spent about four weeks doing the hard work of making sure all 22 datasets were readable by machine, cleaned up and combined into a single knowledge graph that is searchable, with queries pulling data from across the entire lake.
The data lake, which C3.ai is making freely available to researchers, is already being used by teams at MIT and the Federal Emergency Management Agency to analyze supplies of protective equipment and coronavirus tests in different regions, and by researchers at Arizona State University to look at media coverage of the pandemic and its effect on social psychology.
Siebel says he hopes it will bring some ground truth to discussions of the pandemic. “So the next time someone says something they will know what the hell they are talking about instead of listening to this, candidly, idle speculation,” he says.
On the coronavirus wreaking havoc on A.I. models:
Siebel says this might be true for a few systems—like those trying to predict purchasing behavior based on historical data—but he is convinced Covid-19 is massively accelerating digital transformation, including the adoption of machine learning. (Jonathan and I have previously discussed the issue in this newsletter.)
Take the oil majors, many of whom are his customers. “They have two choices: They can reinvent themselves and figure out how to make money at $26 per barrel oil or go out of business, it is that simple.” He says C3.ai has seen its “pipeline of opportunities” quadruple compared to the same period last year, and he predicts it will triple again by next year.
At the same time, Siebel sees a big shakeout coming for A.I. startups—many of which, he says, are not real businesses. He predicts eight out of ten will go bust. “You clear the silliness out of the market,” he says. “All the Masa Sons are gone.” (Siebel is no fan of SoftBank CEO Masayoshi Son’s $100 billion Vision Fund, which he says was the “poster boy” for irrational excess in tech investing over the past five years. Vision Fund recently posted a loss of $18 billion in the year to March.)
On working with the Department of Defense and Big Oil:
Unlike some A.I. engineers and executives, Siebel has no issue working with Big Oil or the Pentagon. Many oil companies are trying to go green, he says, and machine learning can help them transition.
As for the military, he says it is the height of hypocrisy for any Google engineers and executives, “many of whom are my friends” but who owe “their big homes in Atherton” to the development of the Internet and GPS satellites, both of which the U.S. Department of Defense created and funded, to refuse to work with the military.
“We are proud to work with the governments that stand for democracy and individual liberty and happy to work with energy companies,” he says, and he doesn’t mince words. “If someone is not OK with that, fine, go work with Google or Facebook, and consort with the Chinese and manipulate 13-year-old girls into being anorexic and suicidal.”
Siebel is known for his wariness over doing business in China, citing concerns over “state-sanctioned IP theft.” Google has made multiple attempts to gain footing in China, including through its controversial Project Dragonfly. (And yes, frequent Facebook use has been linked to eating disorders in young women—though the company is cracking down on related content.)
On A.I. ethics:
Siebel thinks it is hypocritical for companies like Google and Facebook to tout any adherence to A.I. ethics. These companies, he says, use machine learning to monetize people’s personal information and, through social media, feed addictive behavior and mental-health issues. He notes their platforms have been used for political disinformation and election meddling. What’s ethical about that?
And on government regulation:
“I think if governments do not get involved in regulating A.I., we will be very sorry, because we are going to have to live there,” he says. “I am not a big government guy at all. But there is a role for government here.”
And with that, here’s the rest of this week’s A.I. news.
A.I. IN THE NEWS
Layoffs at Cruise as the self-driving car industry faces a reckoning. Self-driving company Cruise Automation, which is majority owned by General Motors, is laying off around 160 employees, or 8% of its workforce, Bloomberg reports. The San Francisco company has had to push back its timeline for putting its self-driving cars on public roads—which it originally said it would do by 2019—several times. The New York Times has a long, sobering look at the way the coronavirus is creating a serious shakeout among self-driving car startups, with only handful likely to survive.
Nvidia rolls out its most powerful chip yet. The chip giant has unveiled a potent new graphics processing unit designed to speed up the time it takes to run A.I. operations. Nvidia's new A100 packs 54 billion transistors on a single chip and runs some A.I. workloads 20 times faster than the company's previous top-of-the-line GPUs. Nvidia's GPUs, which are also used for graphics-intensive animation and computer gaming applications, have been mainstays of machine learning and A.I. workloads but they face increasing competition from A.I.-specific chips. My Fortune colleague Aaron Pressman has the story here.
More chip news. TSMC will open a chip factory in the U.S. Taiwan Semiconductor Manufacturing, one of the world's Big Three chipmakers, announced it will spend $12 billion to build a chipmaking plant in the U.S. The factory, which TSMC says will each month produce 20,000 wafers embedded with its latest, smallest, 5-nanometer transistors, has about a fifth the capacity of its newest factory in Taiwan. The announcement comes as the Trump administration has voiced increasing concerns about the resilience of technology supply chains and their dependence on China. The Wall Street Journal has the story.
And yet more chip news...Sony says it has created an image sensor with built-in A.I. The Japanese electronics titan says it created an image-sensing module, composed of a sensor, a logic processor and memory, all on a single chip, with built-in A.I. firmware that enables the sensor to run computer vision applications without having to actually output an image, according to Bloomberg News. Sony says this will make image processing much faster and more secure. A few days after debuting the chip, Sony said it had signed a deal with Microsoft to start rolling the new A.I.-capable chip out to customers in "retail and logistics businesses" where it could be used to optimize "warehouse and factory automation, quantifying the flow of customers through stores and making cars smarter about their drivers and environment," according to a second Bloomberg story.
Facebook uses more A.I. to combat hate speech and Covid-19 misinformation and launches a meme-detection challenge. The social media giant touted a number of advances in its use of artificial intelligence to help police content. I reported on these developments for Fortune here. The innovations include one of the largest pre-trained language models ever deployed to date. Facebook has also made freely available a dataset of 10,000 hateful memes for researchers to use to train algorithms to automatically detect. The company is running a competition with a $100,000 prize pool to encourage people to work on the problem. So far, the company's own best effort, an algorithm that trains on both images and language simultaneously, scores 63% accuracy compared to 85% for a human reviewer.
EYE ON A.I. TALENT
Google has hired Paritosh Desai to work on improving the ads served up through its "Discover" and Gmail applications, according to a post on Desai's LinkedIn page. Desai had been senior vice president, chief data and analytics officer at Target.
Hewlett Packard Enterprise has appointed Kumar Sreekanti as chief technology officer and head of software, according to a company blog post. He replaces Mark Potter, who is retiring.
Bank of America has hired Diane Daley as head of its enterprise data governance group, where she will be responsible for helping to shape policies governing the bank's use of A.I., according to a Business Insider story. Daley had been a managing director and head of finance and risk infrastructure at Citigroup.
Italian artificial intelligence company Axyon AI has appointed Francesca Campanelli to the new role of chief commercial officer, according to a story in Funds Europe. Campanelli had been head of product development at Quaestio Capital Management.
EYE ON A.I. RESEARCH
How about a nice game of chess?
The Defense Advanced Research Projects Agency (DARPA) is running a competition, called Gamebreaker, to find an A.I. system capable of analyzing a number of different open-world video games. It wants that A.I. to be able to suggest the best ways to upset the balance between two competing players of equal skill—changing the design of the game so that even if two opponents have equal footing, one is much more likely to win. The agency says it wants to develop this game-design A.I. in order to apply to new war-fighting simulations that DARPA is also currently developing to train military officers with.
DARPA announced Gamebreaker back in February, but last week it put out a press release announcing the nine teams participating and the games they will be working on. The companies include a number of large defense contractors, such as BAE Systems, Lockheed Martin and Northrop Grumman, some of them in partnership with academics from MIT, Purdue and UC Santa Barbara. There are also small defense contractors and video game companies involved. A lot of the teams have chosen to work on StarCraft II, which, although set in a science-fiction world, has obvious parallels with real-world war-fighting games. (And where DeepMind has already had success using reinforcement learning to create an A.I. system that can play better than most humans.) Some of the other games being analyzed include FreeCiv, the car-racing game TORCS, and the complex real-time war game Command: Modern Operations.
“We received many incredibly rich proposals from a diverse base of universities and small and large companies,” said Lt. Col. Dan “Animal” Javorsek, the Gamebreaker program manager in DARPA’s Strategic Technology Office. “Since the teams aren’t competing against each other, we’re looking forward to a collaborative effort as performers develop their A.I.s across an array of games.”
FORTUNE ON A.I.
Nvidia introduces new graphics chips aimed at speeding A.I. apps—by Aaron Pressman
Microsoft and FedEx team up to make deliveries more predictable—by Jonathan Vanian
Honeywell lasers in on tech—by Robert Hackett
How Uber is swerving to survive post-pandemic—by Adam Lashinsky
Some of these stories require a subscription to access. Thank you for supporting our journalism.
Will Covid-19 make us more or less comfortable with A.I.?
It is pretty clear at this point that the pandemic is likely to accelerate digital transformation plans for many businesses, and increase reliance on automation, A.I. and robotics. But are consumers and customers ready?
In partnership with Survey Monkey, virtual-assistant software company Interactions LLC recently surveyed 1,000 Americans and found that, for the most part, Covid-19 has made people more accepting of chatbots and robots.
- 45% of consumers said they would prefer to have their inquiry handled by an A.I. system if it is faster than waiting for a human.
- Nearly 40% of those surveyed said they were now more comfortable with the idea of a robot delivering an online order.
- And 33% said they were now more comfortable with the idea of robots working in groceries to stock shelves or clean the store.
- That's about the same number of people who said they were now also more comfortable with the idea of an A.I. system helping doctors or nurses.
But these issues remain divisive, and the pandemic has made a sizable minority less comfortable with A.I.-driven tech.
- 23% of people said the pandemic had made them less comfortable with the idea of delivery robots.
- 25% said they would not place an online order if they knew a robot would be delivering it instead of a person.
- About a quarter of people also said that the pandemic had made them less comfortable with the idea of riding in a self-driving car.
"As the coronavirus continues to have a radical impact on business and society, it’s ushered in a fresh urgency for technological progress,” Jim Freeze, chief marketing officer at Interactions, said. “It’s well understood that A.I. can play a powerful role in operations. At the same time, it’s crucial that consumer comfort is kept front and center as organizations navigate the new normal.”
What, if anything, will companies do to make sure they are serving that substantial minority as they push automated tech forward? Will they keep some human-powered systems in place for those uncomfortable with the virtual alternatives? Or will they abandon those customers altogether?