Apple is hiring a rising star in the world of deep learning to serve as its first director of AI research. Ruslan Salakhutdinov, an associate professor at Carnegie Mellon University in Pittsburgh, will assume the new position, which is meant to help the company make sure that Siri and its other products make use of the fundamental breakthroughs coming out of academic AI research. Salakhutdinov will talk about his research at EmTech MIT 2016, an MIT Technology Review conference held this week.
Salakhutdinov researches very large neural networks used in a technology called deep learning, which lets a computer learn to perform a difficult task by consuming copious training examples. He will continue to work part time at CMU and will hire a team of researchers to work with him at Apple.
The move is part of Apple’s effort to catch up in the race with Google, Facebook, Amazon, and other tech companies to gain an edge through recent advances in AI and machine learning. In particular, it is becoming increasingly important that software be able to learn for itself how to perform a task, using techniques like deep learning. The latest AI techniques are used in Google’s new smartphone, Pixel, for example. Its voice helper, called Assistant, will go head to head with Apple’s Siri, using machine learning to try to understand users’ spoken queries.
Salakhutdinov’s work on cutting-edge machine-learning techniques could help drive progress in AI, and it could be applied in many different areas, including computer vision, natural language processing, and even robotics.
Speaking recently, Salakhutdinov said that there are three big areas where AI is progressing: giving computers better language understanding; enabling them to learn through repetition and positive reinforcement; and developing ways for machines to learn from unlabeled data. He also highlighted the work he’s doing on teaching machines to learn from unstructured data on the Web, something that could conceivably help make a product like Siri more intelligent.
“We’re working on the idea of trying to use external knowledge bases,” he said. “If I ask you something about a particular thing, can your system basically go to Wikipedia, read a few different articles, learn some facts about the world, and provide you with the right answer?”
Salakhutdinov’s recent work has also focused on enabling machines to learn from different sorts of data, and on ways for things learned in one context to be applied in a completely new one—areas known, respectively, as multimodal learning and transfer learning. He also collaborated on a project that showed, by taking inspiration from cognitive science, how computers can learn from relatively small amounts of data (see “This AI Algorithm Learns Simple Tasks as Fast as We Do”).
In recent years, competitors such as Google and Facebook have hired leading figures in deep learning to lead their AI efforts. Apple has a much less prominent AI research effort than competitors, and some have suggested that its traditional secrecy has made it difficult for the company to recruit the best researchers in the field.
Deep learning has gained prominence in recent years, after proving spectacularly good at enabling machines to recognize objects in images and spoken words in audio. Google and Facebook use the technology to automatically caption images. And it has made Siri and other voice-activated products better at recognizing words. Parsing the meaning of those words remains a grander challenge, however (see “AI’s Language Problem”).