Why companies are spending more on A.I.
Companies are spending more on big artificial intelligence projects to help identify business trends, among other things.
Just over half of businesses said they would spend $500,000 to $5 million on A.I. initiatives this year, up from 34% in 2020, according to a survey released on Tuesday by data labelling firm Appen.
While some companies have halted their A.I. projects over the past year due to the COVID-19 pandemic, the Appen survey suggests that some businesses are actually spending more. One reason may be that companies are now more frequently training their machine learning models, a costly expense because of the computing power required.
The survey found that 87% of companies plan to update their machine learning models—essentially, formulas that make predictions based on data—at least each quarter in 2021, up from 80% last year. Meanwhile, 57% of businesses said they plan to update their models even more often—monthly—versus 45% last year.
Appen marketing vice president Sid Mistry explained that the pandemic is likely responsible for the increased data training. A year ago, for example, grocery delivery company Instacart told Fortune it had started training its A.I. systems more often after customer behavior changed during COVID-19; for example, shoppers bought a lot more toilet paper and store supplies ended up running low.
Paul Cormier, CEO of IBM-owned business software firm Red Hat, also told Fortune that it had to overhaul its sales forecasting and budgeting models because the historical and analytical data it had used previously was now “out the door.”
“Think about a company like DoorDash,” Mistry said about the food delivery business that had a huge influx of orders during the pandemic. “Your model from January is extinct in May or June.”
The increased frequency of data training also leads to increased spending on ancillary activities like data collection and data labelling.
Here are some other highlights from the survey, which was based on 501 responses from executives to corporate data scientists:
• Most companies view “responsible A.I.,” a term referring to mitigating societal harms caused by A.I., through the prism of “risk management” rather than fairness or ethics. The finding implies that businesses are more worried about A.I.’s potential legal risks than altruistic principles.
• The No. 1 reason companies invest in A.I. is to “support internal IT operations.” The second is to “improve understanding of corporate data,” followed by “improve productivity and efficiency of internal business processes.”
• Small-to-medium sized companies are more willing than larger ones to share data with others “with the right security and privacy in place.” This is significant in that it shows bigger companies are less inclined to share because of fears that doing so will benefit competitors.
Join Fortune's MPW Community
Nominate a rising female executive to attend the Fortune Most Powerful Women Next Gen Virtual Summit June 23-24. Complete the form here.
Disrupting traditional industries with quantum computing
A quantum revolution is upon us. Quantum computing will help enable organizations to tackle problems that the finest classical computers of today are incapable of. Organizations looking to build a first mover advantage and lead the field in years to come should invest in building quantum capabilities. Learn more
Nvidia goes deep on maps. Nvidia said it would acquire DeepMap, a startup that helps create high-definition maps for companies developing self-driving cars. Nvidia said that DeepMap’s technology will help “bolster” the company’s own autonomous vehicle technology that it sells to automakers. The chipmaker did not say how much it paid for DeepMap.
A new A.I. task force emerges. The White House announced the creation of the National Artificial Intelligence Research Resource Task Force, which will advise lawmakers about A.I. policy and research issues. The task force will “write the road map for expanding access to critical resources and educational tools that will spur AI innovation and economic prosperity nationwide,” The White House said. Some of the task force’s members include technical experts like deep learning pioneer Fei-Fei Li of Stanford University; Google Cloud A.I. general manager Andrew Moore; Oren Etzioni, the CEO of the non-profit Allen Institute for AI; professorJulia Lane of the NYU Wagner Graduate School of Public Service; and Deputy U.S. Chief Technology Officer Lynne Parker.
Amazon fresh. Amazon introduced its cashierless check-out system at an Amazon Fresh store in Bellevue, Wash. The online retail giant said the Bellevue Amazon Fresh store is the first full-size grocery store that has an automated check-out system, which lets people walk out of stores with items without needing to physically scan each product. People can still choose to shop the traditional way with human clerks.
Another A.I. acquisition. U.K. biotech startup Exscientia, which uses A.I. to help design drugs, acquired the company Allcyte, Fortune scribe Jeremy Kahn reported. Allcyte uses A.I. to help screen and match patients with different cancer treatments. From the article: Andrew Hopkins, Exscientia’s founder and chief executive officer, said that his company is buying Allcyte because of the data its technology can provide about whether drug candidates are likely to be successful prior to human clinical trials.
Target hired Brenda O’Kane as vice president of infrastructure and production engineering, tech publication The Information reported. O’Kane was previously a vice president of technology at Disney.
Google picked Christopher Phillips to help lead its product and engineering teams that work on Google Maps, among other services, TechCrunch reported. Phillips was previously a chief product and technology officer and executive vice president at SiriusXM.
A.I. comes to chip design. Google researchers published a paper in Nature about the use of deep reinforcement learning—an A.I. technique that teaches computers to learn by trial and error—to help design computer chips, a process referred to as chip floorplanning. It’s a big deal because chip designing is complex, requiring human expertise in computing and electronic engineering. The paper shows that reinforcement learning can be a useful tool to more quickly develop computer chips than current, conventional methods.
From the paper: “Despite five decades of research, chip floorplanning has defied automation, requiring months of intense effort by physical design engineers to produce manufacturable layouts. Here we present a deep reinforcement learning approach to chip floorplanning. In under six hours, our method automatically generates chip floorplans that are superior or comparable to those produced by humans in all key metrics, including power consumption, performance and chip area."
A.I. and Super Mario Brothers…the movie that is. The rise of deep fakes, which are images and videos covertly manipulated by A.I. to fool people into thinking they are real, has alarmed politicians, companies, and citizens who are worried that the technology poses serious threats to society. But the technology also has some legitimate uses, particularly for filmmakers.
It turns out that the technology used to create deepfakes can also be used to help restore older movies to look livelier and more colorful to accommodate today’s 4K televisions and the like. A.I. researchers have previously used this technology, known as generative adversarial networks (GANs) to restore movies from the early 1900s.
More recently, filmmaker Garrett Gilchrist used the GAN technology to help restore the oddball 1993 movie Super Mario Bros., in which the late Dennis Hopper plays an agitated President Koopa. In an interview with Slate, Gilchrist describes how A.I. was a useful tool in the colorization process, used to enhance or even add new colors to the strange film.
From the article: I was really into using artificial intelligence on this edit. I started using artificial intelligence synthesis to do Remini on the faces, which is going to be controversial, or to do Topaz, or to colorize, to make the work-print VHS, which looks like shit, look more like the Blu-ray, just to synthesize these two sources. I did that for a bunch of shots in this, and I think they look really nice.
Thanks for reading.
If you liked this email, pay it forward. Share it with someone you know.
Did someone share this with you? Sign up here. For previous editions, click here.
To view all of Fortune's newsletters on the latest in business, go here.