The first man vs. machine showdown happened on a checkers board.
In 1961, Arthur Samuel, the father of “machine learning,” taught an artifical intelligence (AI) program to beat the fourth-ranked checkers player in the country.
Samuel had both academia and private tech to thank for his success. His career took him from MIT to Bell Telephone Laboratories to the University of Illinois to IBM, where his checker program demonstration raised the company’s stock by 15 points. At his last gig, he taught PhD students at Stanford.
Today, machine learning experts are training programs to take on much bigger challenges than checkers—from trading stocks to predicting social unrest.
Pairing Data with Talent
But data scientists would be smart to follow Samuel’s lead, and break down barriers between academia and private tech. Only by pairing the resources and real-world data of the private sector with the talent at colleges and universities can we unlock the full potential of AI and machine learning.
Private tech companies know that some of the greatest AI minds of our time are walking around university campuses. According to Carnegie Mellon University’s Dean Andrew Moore, AI students are “worth somewhere between $5 million and $10 million to a company’s bottom line.”
It’s no wonder, then, that big companies are heavily recruiting PhD candidates. Over the past 10 years, nearly 20 percent more data science PhD students have taken industry jobs. Collaborating on data science projects is a great way to connect university students to real-world opportunities and private companies to the talent they are looking for.
Solving Real-World Challenges
By sharing their resources and data libraries, researchers and companies are much more likely to discover AI solutions that improve human lives. A machine-learning model at Stanford sorted through 50 million images in two weeks—a task that would take a human 15 years.
Researchers are using such unprecedented computational power to solve real-world problems. The cancer center at University of North Carolina at Chapel Hill, for example, is using a commercially available AI system to quickly sort through research papers and identify treatment options that doctors might not have considered. Unlike humans, the system can digest all 8,000 medical papers that are published daily.
(Next page: The best AI resources; contests and incentives)
The Best Tools on the Market
Academics and companies also offer each other access to the best AI resources and equipment. Google’s Cloud program, for example, provides researchers with 180,000 “petaflops” of extra computational power—roughly equivalent to 180,000 MacBook Pros—as long as those researchers share their findings with the rest of the tech community.
Some private companies are even tapping into international data science talent. The Facebook Artificial Intelligence Research lab recently gave four GPUs—or graphic processing units—to the Technical University of Berlin’s Dr. Klaus-Robert Müller. Originally created to perfect 3D imaging for video games, the GPU is an AI speed weapon that manages thousands of software threads simultaneously.
Müller is using his new GPUs to accelerate his research on breast cancer image analysis. For its part, Facebook gets to stamp its name on an AI project searching for one of the most elusive cures of our time.
Contests and Incentives
Data scientists are also teaming up to get more creative about the way they study AI. Microsoft is collaborating with academics—by challenging them to a video game contest. Microsoft has PhD. students train AI systems to thrive in Project Malmo, an AI mind game built over Minecraft.
The winners get $20,000 research grants and placements at Microsoft’s AI Summer School in the United Kingdom. Microsoft gets to see how top data-science academics go about solving AI challenges on a virtually limitless stage.
My company, Adobe, likewise recognizes that professors can help us develop tomorrow’s AI solutions. Since 2014, we’ve awarded more than 40 research grants of $50,0000 each to over 60 academics at universities in the United States through our Adobe Digital Marketing Research Awards.
Through our grant program, we’ve collaborated with professors across the country, such as UC Santa Cruz’s Lise Getoor to follow users across multiple devices by predicting which devices belong to the same user. That allows us to make sure customers only receive relevant content and aren’t subjected to the same repetitive ads on multiple devices. With access to Adobe’s 41 trillion marketing transactions, the research collaboration with Getoor enables us to personalize content across devices delivering a lift of 10 to 15 percent.
Adobe has also worked with Stanford University researchers to find out why specific digital coupons lead to customer purchases; with New York University to uncover what consumers’ offline mobile activity says about their preferences; and with a Rutgers professor to detect anomalies in demographic data and customer clicks.
AI and machine learning have the potential to solve some of society’s most pressing problems. They could result in fully autonomous vehicles that eliminate crippling traffic. They could enable oncologists and radiologists to detect, treat, and even cure cancer. They could simply help businesses advertise and market more efficiently, delivering a better experience to consumers.
If private tech companies boost their partnerships with brilliant academics, there’s virtually no limit to the new technologies they could create.