A Year After Pledging Openness, Apple Still Falls Behind On AI


Justin Sullivan / Getty Images

Apple senior vice president of worldwide marketing Phil Schiller describes the AI in Apple’s facial recognition feature during Apple’s iPhone X event on September 12, 2017 in Cupertino, California.

“Apple is actually the NSA of AI.”

in which was how storied computer scientist in addition to also Stanford University adjunct professor Jerry Kaplan described Apple’s unconventional approach to artificial intelligence research in 2016. A year later, his assessment of the company is actually largely the same — despite Apple’s pledge to more fully engage with the research community in which drives innovation from the field.

“My observation is actually in which Apple tends not to be as heavily involved from the academic AI world as different companies in which are well-known for being involved,” Kaplan told BuzzFeed News This kind of week.

Two years ago, Apple’s penchant for keeping its artificial intelligence research secret was notorious throughout the industry. however since then, Apple has made a big show of ramping up its efforts in AI. The company hired Russ Salakhutdinov, a highly respected Carnegie Mellon professor, to be its first AI director, in addition to also This kind of publicly pledged This kind of would certainly engage more with academia. This kind of past July, This kind of debuted an official blog covering its progress in AI in addition to also machine learning.

Kaplan admitted in which he was not aware of these developments at Apple because he does not closely watch the company. however in which’s also the point. One year after Apple’s pledge to be more open with the AI research community, the company is actually still facing much of the same criticism in a crucial in addition to also extremely competitive sector of the tech industry. “different companies have very strong outreach, ties in addition to also interaction [with universities like] Stanford,” Kaplan explained.

BuzzFeed News interviews which has a dozen AI experts paint a picture of Apple’s artificial intelligence research in which shows the company is actually opening up a bit more — however there is actually still a disconnect between the academic AI community’s values in addition to also Apple’s way of doing business. The company’s obsessive focus on the AI applications in Apple products can make working for the company less desirable to some talented experts who have no shortage of options, researchers said. in addition to also in which’s bad news for Apple, which faces an uphill battle in attracting the people This kind of needs to become a true frontrunner in AI among the giants of tech.

“If Apple doesn’t publish, fewer talented people will join.”

Nowadays, the race among tech giants to lead in artificial intelligence — a term generally used for software in which allows computers to learn in addition to also improve tasks on their own — is actually heating up, in addition to also Apple clearly wants to be from the running. AI is actually a technology in which underpins self-driving cars in addition to also voice assistants like Siri in addition to also Amazon’s Alexa, in addition to also tech giants like Google, Facebook, Amazon, Microsoft are in a fierce competition to recruit the top minds in AI in addition to also offer the most advanced applications of the technology. (According to a recent report in The brand-new York Times, salaries for AI experts can range through $500,000 to $0 million annually.) For years, Apple was perceived as lagging from the field, largely because of its tight-lipped approach to research. today, as the availability of top AI talent has declined due to booming demand for expertise from the field, the company has recently sought to show This kind of can respond to the conventions of the AI community, where publishing frequently is actually a must.

“There is actually a stark contrast in how Apple deals with its AI research as opposed to different companies like Google, Microsoft, in addition to also Facebook,” Shreshth Gandhi, a research scientist at the Canadian biotech company Deep Genomics in addition to also a former machine learning graduate student at the University of Toronto, told BuzzFeed News. “Compared to its competition, Apple doesn’t seem to be doing enough to promote brand-new research in AI.”

“If Apple doesn’t publish, fewer talented people will join,” said University of Toronto machine learning student Yuhuai Wu, who studied under Salakhutdinov as a PhD candidate in addition to also who Apple has attempted to recruit. “My biggest concern is actually whether I can still be visible from the research community [if I worked at Apple], in addition to also whether I can do the research in which I want to do.”


One former computer vision AI engineer for Apple still remembers the aura of secrecy surrounding his job. He described how engineers only knew as much about any given project at the company as they needed to know. “This kind of’s very common where, in a team of 10 people, 5 of us are disclosed to a project while 5 of us are not disclosed to a project,” the engineer said. “So you could be sitting beside This kind of person, working together, in addition to also he has some information in which you don’t have, while you have some information he doesn’t have.” in which, the ex-Apple engineer said said, “created a bit of a bad feeling, because you don’t know what the big picture is actually. You don’t know — ‘why am I doing This kind of?’”

To be fair, This kind of was about two years ago, when things at Apple were at peak secrecy. however This kind of underscores how locked down Apple’s culture had been — in addition to also how much the company has tried to open up in subsequent years. In January, Apple joined the Partnership on AI, a group dedicated to developing best practices for AI research, along with Facebook, Microsoft, in addition to also different tech companies. Last October, This kind of hosted BayLearn 2017, a Bay Area machine learning symposium. When Apple brought Russ Salakhutdinov aboard, the announcement made waves throughout the AI community as yet another buzzy AI hire made by a top tech company, following brand-new York University’s Yann LeCun joining Facebook in 2013, in addition to also Geoffrey Hinton of the University of Toronto joining Google from the same year. (Salakhutdinov joined Carlos Guestrin, a notable University of Washington professor whose machine learning company Apple acquired in August 2016.)

At the iPhone event This kind of past September, people watching for clues on Apple’s AI progress were treated to a slew of AI-infused features from the brand-new iPhones. At the event, Apple executives in addition to also presenters explained the brand-new iPhone X’s features using terminology familiar to the AI-literate: the company’s brand-new A11 Bionic chip includes a so-called “Neural Engine” designed for artificial intelligence processing tasks like mathematically modeling the human face for the iPhone X’s brand-new FaceID authentication feature. Crucially, the chip uses machine learning to evolve recognition of face over time, in addition to also even if you grow a beard or wear glasses.

No different company is actually as well-positioned to pull off features like This kind of, said co-founder of Deep Genomics in addition to also deep learning expert Hannes Bretschneider — especially because they inform the user experience, or how consumers interact with Apple products, which is actually the company’s forte. “Apple actually thinks about the product first, then they find the tech to enable those products,” he said. (Eddy Cue, Apple’s senior vice president of internet software in addition to also services, echoed the sentiment in a 2016 interview with Wired: “We are driven by a vision of the end result,” he said.) in which strategy includes acquiring companies in which possess the technology Apple aimed to enable in its products, such as the eponymous startup behind its voice assistant Siri in addition to also a company called PrimeSense, which manufactured Microsoft’s movement-sensing game system Kinect, before being acquired by Apple, in addition to also which is actually likely behind the dot grid system used from the face recognition tech within Face ID. in which strategy has its pros in addition to also cons: Acquiring companies for their technology can be expensive, however the tech sees a more immediate application. Meanwhile, funding basic research pushes the far-out edges of AI technology — however applications may not end up materializing until 5 or 10 years down the line, if at all.

In July 2017, Apple also launched a machine learning blog to better align itself with the practices of the AI research community. however from the 5 months This kind of’s existed, the blog has published only seven entries on Siri, face detection, handwriting recognition, in addition to also text labeling — all of which reveals a product-centric focus. What’s more, the entries list teams at Apple instead of individual researchers, in contrast to different peer-reviewed journals from the field.

“in which blog is actually completely useless.” 

“in which blog is actually completely useless,” an AI professor of an elite university, who asked to remain anonymous because they did not want their name attached to criticisms of an influential tech company, told BuzzFeed News a few weeks ago. “There are absolutely no details, for example, in Apple’s post about AI in handwriting recognition. This kind of amounts to bragging in addition to also This kind of is actually impossible to actually learn anything through This kind of. This kind of feels like they realized most big-name institutions have blogs in addition to also created one, however didn’t do This kind of in a way in which adds any value. I would certainly contrast This kind of with Google’s post about neural networks for language understanding, which has many more details in addition to also points to public code along with walkthrough explanations.”

Of Apple’s most recent post on face detection, published last Thursday, the AI professor said: “Not bad. There’s still no code released, however not bad. Certainly better than the different stupid one. They’re improving!”

Three months ago, Apple won a best paper award at the 2017 Conference on Computer Vision in addition to also Pattern Recognition — one of the most influential conferences from the field of AI. in which’s an impressive accomplishment on its face, however what Apple has published formally in peer-reviewed journals pales in comparison to different tech giants. In 2016, Facebook published 125 articles on the free-to-access academic journal arXiv, with in which number likely to grow to about 0 articles in 2017, according to a company spokesperson. A Microsoft spokesperson said the company published 847 AI papers in 2016, in addition to also the tally was at 394 by August of This kind of year. A Google spokesperson, meanwhile, told BuzzFeed News in which the company does not contain the numbers for 2017 compiled yet, however in which This kind of published 133 papers in 2014, 171 AI papers in 2015, 203 in 2016, in addition to also in which This kind of expected 2017 to follow the same general trend. Apple declined to tell BuzzFeed News the number of AI papers This kind of has published in recent years. however in addition to its computer vision research paper, there are only three others with Apple researchers listed as authors on arXiv in which BuzzFeed News was able to find, which brings the unofficial tally to four.

Apple also noted in which This kind of presented three peer-reviewed papers at the International Speech Communication Association’s Interspeech conference in Sweden This kind of past August. According to AI researchers, This kind of is actually more reputable, because there are no rigor standards to publishing on arXiv. “Anyone can put a paper on arXiv in addition to also there is actually no review — peer or otherwise — to assess quality,” Georgia Tech AI researcher Mark Riedl said. “There are only a few papers in which I can think of in which are considered groundbreaking in which were never submitted to or accepted into a highly-competitive peer reviewed conference.” however two researchers, including Riedl, told BuzzFeed News in which the acceptance rate at these competitive conferences matters. “If the acceptance rate is actually less than 50 percent I’d say This kind of’s not bad, less than 25 percent is actually great,” said one AI professor. In 2015, ISCA said its overall acceptance rate was 51 percent.

More to the point, Apple still publishes very infrequently compared to different tech giants. “To be serious in AI, you have to publish at the main peer-reviewed artificial intelligence, machine learning, computer vision, in addition to also natural language conferences,” said Bart Selman, a Cornell University AI professor. Selman enumerated the conferences: NIPS, ICML, IJCAI, AAAI, CVPR, in addition to also ACL. “different tech giants publish dozens of papers at those venues every year. So, Apple incorporates a long way to go.” Of those conferences, Apple has only presented at ICML in addition to also CVPR. (Salakhutdinov did appear at NIPS in 2016 to announce in which Apple planned to start publishing, however the company didn’t present research at the conference.)

“I think This kind of kind of secrecy is actually antithetical to the open research culture from the AI community.”

Apple’s business-first approach to AI development is actually fundamentally at odds how the AI research community is actually used to exploring brand-new ideas in addition to also brand-new science, some researchers told BuzzFeed News. “One of my friends had a recruitment interaction with Apple, in addition to also they refused to give him even a vague idea of the kind of work he might be doing if he joins,” Gandhi wrote in an email to BuzzFeed News. “I think This kind of kind of secrecy is actually antithetical to the open research culture from the AI community where most research by universities in addition to also companies alike is actually shared,” Gandhi said.

AI researchers in which BuzzFeed News spoke with acknowledged in which there’s a big debate over the merits of publishing consistently versus having your research be deployed in millions of Apple devices the entire world over. “Many folks believe in deployment over publishing,” said the anonymous AI professor. “I think both are fair.”

however as Chris Nicholson, CEO in addition to also founder of deep-learning startup Skymind points out, AI is actually a field dominated by academics in addition to also researchers, in addition to also “those people like to publish. Publishing is actually like breathing to them: You do This kind of or you die,” he said. “So if you try to recruit AI researchers by promising lots of money in addition to also zero peer recognition, you won’t get very far. There are some people who will never join Apple for in which reason.”


Apple has also reportedly fumbled its AI projects in areas where This kind of has been ahead of the curve. A recent Wall Street Journal article described how former Siri team members said progress on the voice assistant was slowed by a failure to set ambitious goals, changing strategies, in addition to also a dominating focus on developing the iPhone. “Siri is actually Apple’s biggest Achilles’ heel in AI,” said Bretschneider. “In terms of ability, This kind of’s long been outgunned by Google in addition to also Amazon.” Unlike Amazon’s Alexa, which allows developers to code custom “skills” across a wide range of applications, Apple opened Siri up to only seven types of apps, including payment in addition to also ride-sharing, during its developer conference last year.

different limitations are set by Apple’s emphasis on privacy. At Apple’s developers conference in 2016, the company made a big deal of being the first to apply a research technique called “differential privacy” at scale — basically a way for Apple to analyze user data at scale without revealing anything about an individual, like what a user is actually accessing on the web via their iPhones.“Apple’s goal there is actually to have This kind of both ways,” explained Bretschneider. “For a long time, This kind of just collected very little data through their users to begin with. [Using differential privacy] allows Apple to collect more data of their users today without changing their fundamental agreement.” While through the outside This kind of’s hard to know how effectively This kind of strategy is actually helping Apple develop its AI, Bretschneider said in which This kind of is actually apparent Apple’s AI efforts are “more targeted in addition to also limited,” whereas a company like Google, “essentially wants to be an AI company.”

Doubling down on its commitment to privacy, Apple also keeps most user data on the phone itself in addition to also deletes This kind of after a few months. however Eugenio Culurciello, a professor at Purdue University who works on machine learning hardware, said in which while AI processing on a chip is actually better than This kind of has been before, limitations on power in addition to also memory bandwidth still make a mobile device no match for cloud AI — which is actually what Google in addition to also Amazon use. (These companies keep data until users explicitly request for This kind of to be discarded.)

“AI at Apple is actually hobbled by the way they handle information.”

Essentially, Skymind’s Nicholson added, Apple is actually accepting a commercial disadvantage based on its business type. “AI at Apple is actually hobbled by the way they handle information,” said Nicholson. Apple takes the data privacy of its users very seriously — quite possibly the most seriously of any major company in Silicon Valley. This kind of likes to tout in which This kind of is actually because This kind of is actually a hardware business, not a company whose business ultimately relies on advertising, like Facebook in addition to also Google, which sell ads based on user data.

however This kind of also means in which This kind of’s harder for Apple to benefit through so-called “data effects,” where products improve based on the sheer volume of data you collect. “AI benefits through data effects,” Nicholson said. “This kind of needs massive amounts of data, in addition to also once you have in which, you can build a superior product, attract more users, expose your AI to even more data, in addition to also embark on a virtuous cycle. Apple’s not fully participating in in which cycle.” (Apple, for its part, said the right data is actually more important to the company than having the most data, in addition to also This kind of is actually satisfied with the data This kind of collects in privacy-preserving ways.)

Apple is actually a wildly successful company — a company in which, in turn, has become the entire world’s most valuable publicly traded company, the entire world’s top retailer in sales per square foot, in addition to also the entire world’s most profitable company ever. however as the rise in addition to also fall of the entire world’s biggest corporations tells us, no one player incorporates a permanent claim to No. 1. in addition to also if the pundits are correct, the fortunes of today’s most successful companies may well turn on AI, technology in which is actually already radically transforming essential human industries, through medicine in addition to also finance to labor in addition to also art.

“Apple has not shown the entire world This kind of’s a leader in AI,” said Nicholson. “in addition to also its strategy may mean in which This kind of won’t own the future of AI.”

Davey Alba is actually a senior technology reporter for BuzzFeed News in addition to also is actually based in brand-new York.

Contact Davey Alba at davey.alba@buzzfeed.com.

Got a confidential tip? Submit This kind of here.

Leave a Reply

Your email address will not be published. Required fields are marked *

*

two × five =