Bringing AI to everyone.
Part III of our series, “Real Perspectives on Artificial Intelligence” features Rohit Prasad, the Head Scientist of Amazon’s Alexa AI.
I recently read a note by President Reif of MIT regarding the school’s new College of Computing. He stated that its students will be “navigating the algorithmic future.” That is a provocative concept — how do you envision such a future?
It’s definitely a provocative statement — I agree with it. We are living in the early years of the golden age of AI. AI algorithms in hardware and software are becoming more and more indispensable and ubiquitous. They are part of the social fabric.
In some cases, I think algorithms will recede into the background where they assist with decision making. In other cases, I think algorithms will be embodied in physical form factors — self-driving cars, robots; these are the manifestation of AI in the physical world. Alexa, as it currently stands, is an example that blends these two schools of thought.
The pervasive adoption of AI will be so profound that it will have a deep impact on all types of education. The barrier to programming AIs will also be lowered with the ongoing push in democratizing AI. Just a fascinating future from a student perspective.
To close on this — being in this field for a long time — I’ve noticed the conversation in industry is completely shifting. Leaders in enterprise are not asking the questions: what is AI? Is it real? Does it have any benefits? They’re asking: how do I apply AI right away? Or how can I go faster with AI? It is just inconceivable for me to imagine a future where AI will not touch every person in the world
So have we closed the perception gap — the gap between the capabilities of AI and the perception of those capabilities?
That gap has closed so much in the last decade, especially in the last five years. This isn’t an issue we should worry about anymore. Take Alexa as an example — if there was a gap between perceived capabilities and actual abilities, then customers wouldn’t use it.
Five years back when we were working on launching Echo and Alexa, even the academic community was extremely skeptical about the possibility of far-field speech recognition in a mainstream consumer product. There were lots of studies in labs, but to have a product that millions of customers with different accents and noisy households could interact with from a distance…that was unheard of.
So why did this change? We’ve all dreamt about the Star Trek computer, and we’ve all thought about making it a reality for a long time. The expectations were always high. The reason we were able to bridge that gap of perception versus capabilities was because Alexa did a few things quite well early on and then became more useful, more accurate, more conversational, and more natural in understanding and decision-making. Alexa is also striving to be a fast self-learner and is another reason for bridging the perception gap in human-machine interactions.
I’m sure you’re a little biased here, but which AI innovations excite you most?
I get most excited by AI innovations that transform our daily lives in a positive way. I also cherish AI advances that make complex tasks simple.
I am super excited by AI’s role in medicine; I’m very optimistic that AI will have a profound impact on personalized medicine. There are already early signs of that happening. Self-driving cars will be revolutionary, but it’s still very early. And the integration of AI into robotics, in both industrial environments and other forms, will be transformational.
Of course, my favorite is conversational AI, which means you interact with the AI as you do with a human. It is, to me, the hardest AI challenge — you are seeking to model human intelligence and then transcend it by adding superhuman capabilities.
As you noted, you’ve been in the industry a while. What’s surprised you the most regarding the evolution of AI over those years?
The first is Alexa related — if you asked me five years back, “In five years, will you see billions of customer interactions every week with Alexa, and more than 100,000 skills on the Alexa service?” I would say, “nope.” I had planned for success but not of this kind. We’re now comfortable speaking to a device at a distance and even anthropomorphize the AI. Such adoption is just mind-boggling. I may have a bit of proximity bias, but I really think it is a testament to the AI community that there has been such mainstream adoption.
The second positive surprise is the democratization of AI, where it is not a privileged technology anymore. Developers can integrate AI in their environment via open-source ML frameworks or AI tools from cloud providers such as AWS and even build an AI experience on a voice assistant using tools like the Alexa Skills Kit without worrying about how AI works. Such democratization has been the most fulfilling, positive surprise to me in the past 20 years.