Some companies are spending over $5 million on artificial-intelligence projects, showing just how serious businesses are getting about the cutting-edge technology.
The finding comes from a recent survey of over 300 companies by Figure Eight, a business that was recently acquired by data company Appen that helps customers label their data. The results provide a glimpse into how quickly companies are adopting A.I. and some of the related complications.
Over half the respondents said they oversee an annual A.I. budget of at least $51,000. Meanwhile, 13% said they control A.I. budgets of $251,000 to $500,000, an amount that reflects a modest commitment and likely small staff.
Five percent said they’re spending over $5 million. That would imply a huge operation that rivals spending on corporate tech basics like some cybersecurity products.
Although the survey didn’t get into how that money is being spent, Figure Eight vice president of marketing Sid Mistry speculates that the budgets include capturing and labeling data, building and running machine-learning models, and then improving the models’ performance over time.
The survey also found that data scientists and machine-learning engineers are surprisingly unhappy. About 30% of data scientists said they were either unsatisfied or somewhat satisfied with their jobs.
This is despite the fact that data crunching is one of the most in-demand jobs. Annual salaries average around $120,000, which is well above the norm.
Mistry speculated that data scientists may be growing increasingly frustrated because of the greater number of business leaders parachuting into A.I. projects. Amid the growing hype over A.I., more companies are considering A.I. core to their survival and are giving once largely autonomous projects greater scrutiny.
“Three years ago, I don’t think people would have figured out how to spell A.I.,” Mistry said.
EYE ON A.I. NEWS
Deepfake concerns. Artificial intelligence experts briefed members of the House Intelligence Committee last week about the dangers of deepfake videos that look real but are heavily altered by computers to show people saying things they never said. One expert recommended that lawmakers partner with social media companies to “rapidly refute smears” that may arise from deepfakes, CBS News reported.
Facebook’s hiring spree in London. Facebook will open a third office in London that will focus on technology, with a bulk of its new hires specializing in unspecified A.I. roles, reported VentureBeat. Facebook will employ about 3,000 people in London in all three of its offices by the end of 2019, the publication said.
Let others decide. Amazon’s chief technology officer Werner Vogels told the BBC that he doesn’t believe it’s his company’s responsibility to ensure that other use its facial-recognition technology accurately or ethically. “It’s in society’s direction to actually decide which technology is applicable under which conditions,” Vogels said. “It’s a societal discourse and decision–and policy-making–that needs to happen to decide where you can apply technologies.”
Self-driving tech doesn’t come cheap. Hyundai invested less than $30 million into the autonomous vehicle company Aurora, reported TechCrunch, citing unnamed sources. The investment marks an expansion of a partnership between the two companies that was announced in Jan. 2018, the report said.
DEEP LEARNING AS “MAGIC”
François Chollet, a prominent A.I. researcher at Google, describes the fallacy of thinking that deep learning by itself will lead to computers that think like humans, otherwise known as artificial general intelligence (AGI). He writes on Twitter: “Looking at what you can do with deep learning today and thinking “let’s scale this up to AGI” is basically the equivalent of watching a magic trick and thinking ‘wow magic is real! im gonna start a magic company that uses magic spells to generate infinite value.’”
EYE ON A.I. HIRES
Wells Fargo picked Debra Chrapaty to be the banking giant’s chief technology officer. Chrapaty was previously Wells Fargo’s group chief information officer for core platform services. She was also a chief information officer at Zynga, a general manager at Cisco, and a corporate vice president at Microsoft.
Blue Apron named Irina Krechmer to be the meal-kit company’s chief technology officer. Krechmer was previously the vice president of engineering at XO Group.
Embark, a startup specializing in self-driving trucks, hired Zeljko Popovic to be lead its efforts in robotic perception. Popovic was previously a high-ranking member of Tesla’s Autopilot team.
EYE ON A.I. RESEARCH
A.I. stumbles in countries with many low-income households. Facebook researchers published a paper detailing how image-recognition tools from companies like Microsoft, IBM, Google, and Amazon fail more often when examining objects in pictures from low-income households in countries outside the U.S. In one example, the researchers found the image-recognition tools failed to recognize toothbrushes in photos of taken in households without bathrooms.
Eye on A.I. Researchers from Intel published a paper about using deep learning in video-conferencing systems to automatically alter people’s eyes so that they appear to be staring at their computer cameras instead of their computer monitors. The researchers used the same technique (generative adversarial networks) used to develop deepfake videos to create their eye alterations.
FORTUNE ON A.I.
Deepfake Video of Mark Zuckerberg Goes Viral on Eve of House A.I. Hearing – By Bernhard Warner
How Facial Recognition Will Change Driving – By Jeff John Roberts
Uber Finally Has Ride-Hailing Competition in London – By Katherine Dunn
A.I. goes James Bond. Unspecified countries may have created a fake persona on LinkedIn intended to fool Washington political insiders as part of an espionage campaign, The Associated Press reports. What makes this spy story unique, however, is that the photo of the LinkedIn persona, “Katie Jones,” who connected with several Washington insiders, appears to have been completely generated by computers. The news agency talked to A.I. researchers who suspect that the photo was developed by the generative adversarial network (GAN) techniques used to create deepfake videos. “This is a typical GAN,” one researcher said. “I’ll bet money on it.”