MIT begins new quest for intelligence
What is consciousness? How does the arrangement of billions of neurons make you you? What does it mean to think? These were some of the many “thought-provoking” questions posed at MIT’s IQ launch event last week, exploring the next stage in human and machine intelligence. To understand machines, much of the discussion drew parallels to the development of children. Similar to machines, children first learn through a language they don’t understand and need vast amounts of data (audio, visual, etc.) to build comprehension. However, children soon have the capability of understanding intention, an intuitive gift that far surpasses even the most advanced computers. Professor Josh Tenenbaum explained that despite all the hype, there is currently no such thing as “artificial intelligence”, only systems that are trained to do one specific thing. His work seeks to develop AI that “starts off like a baby, and learns like a child.” To do this, his team is leveraging techniques such as Bayesian program learning, whereby computational systems figure out what action sequence is most likely to produce a desired outcome. Goals determine rewards and physics determines costs. His vision is to one day “build a machine that grows into intelligence.”
MIT’s best and brightest took the stage over the course of the day, including a sequence of three MacArthur Genius Grant recipients. The groundbreaking research of Professor Regina Barzilay is using machine learning to help detect and cure cancer, while Professor Andrew Lo is using artificial intelligence to build personalized investment themes designed around individual risk tolerance levels. The latest inventions coming out of MIT labs were also on display. Those that caught the audience’s attention included Jibo, the first social robot for the home and one of Time’s “Best Inventions of 2017”, SMART, a fully autonomous self-driving wheelchair developed by MIT’s Computer Science and Artificial Intelligence Laboratory (CSAIL), and Sonic Vision, which uses wifi signal to detect motion and monitor human functions like breathing and heart rate with accuracy similar to that of an EEG.
No topic was off limits for our esteemed speakers. The discussion on intelligence and society was perhaps the most introspective. Joi Ito, Director of the MIT Media Lab, explained that there are unfair false positives hidden in the data that disadvantage people and urged us to look for causal inference. For example, the many hidden proxies for race including zip code. Prediction isn’t fair when the data isn’t fair as it only serves to reinforce more accurate models of bias and prejudice. He also asked us to redefine how we measure outcomes. For example, he explained that GDP, the most common definition of productivity, is an outdated concept from the industrial age that was never intended to be a measure of welfare. It doesn’t capture time spent with a child, but gives tremendous weight to meaningless tasks that will likely be replaced by machines. Megan Smith, former Chief Technology Officer of the United States and Assistant to President Barack Obama, suggested we need to democratize who creates with AI, so that it is not just in the hands of the few. Part of the future of work is including everyone. AI should help us all flourish.
Moving forward with the initiative, there are a few key developments worth noting: 1) Eric Schmidt, Alphabet senior advisor and former executive chairman and CEO of Google, will serve as a founding advisor for IQ and will be joining MIT as a Visiting Innovation Fellow. 2) David Siegel, Co-founder of Two Sigma Investments (a quantitative investment firm managing $45 billion), will serve alongside Eric Schmidt as a founding advisor to help shape the direction of the initiative. 3) SenseTime, China’s leading AI “unicorn” valued at more than $3 billion, will be one of the first companies to join the initiative as part of the new MIT-SenseTime Alliance on Artificial Intelligence.
As we look to the future, President Reif encouraged us to consider MIT’s vibrant past. Sixty years ago, Marvin Minsky asked the question: What is intelligence and how does it work? This ultimately led to new areas of research including artificial intelligence, cognitive science, neuroscience, and modern linguistics. This in turn inspired what we now know as computational neuroscience, neural nets, and empathetic robots. President Reif painted a picture of a nearby future where machines and people work together and encouraged us to ask us how can we use our grasp of human intelligence to build wiser and more useful machines, for the benefit of society. Perhaps the sentiment of the day was best captured by Laura Schulz, Principal Investigator of the Early Childhood Cognition Lab at MIT: “When you start on a quest, you don’t know quite where it will take you, but you have a sense of what you might try to find. My hope is that this it is the kind of quest that brings us back to ourselves and a deeper understanding of who we are.”
Please click here for a video recording of the event: http://web.mit.edu/webcast/iqlaunch/
