Recent podcast
Check out my latest podcast with Joseph Aoun on Becoming Robot Proof in the age of Artificial Intelligence. There’s some good advice in there from a seasoned educator, leader, and author of the book, Robot Proof.
What’s a Good Question?
I snuck in a brief visit to Kashmir en route to my boarding school reunion. The picture above is the spot where I congregate for an annual family ritual on a specific full moon every summer, rain or shine. It’s a great place to recharge.
On this visit, I was invited to address a group of high school kids at a local high school – the Delhi Public School in Srinagar. On my way to the session, I noticed a sign on the wall that said something to the effect that computers are fast but stupid, while humans are slow but intelligent. I decided to anchor my remarks around that message.
That’s quite an old sign, I began, and asked whether it was still true. Drawing on recent work in AI, I proposed that machines have become quite intelligent, even creative.
There was a great vibe and excellent Q&A during the session. There were some of the usual questions, like whether AI would displace humans. But I thought the most interesting ones concerned whether the replacement of humans by machines is a good thing, and how a machine could possibly be more intelligent than its creators.
I thought the second question was an excellent one. It provided a natural way to unpack the components of intelligence, and to consider in which of these components a machine could exceed the capabilities of its creator in the not too distant future. After all, machines are already seeing, reading, and listening quite well, and could start interpreting this information in new ways that might exceed the capacity of humans.
The first question, whether AI automation of decision-making in important areas of our lives will be a good or bad thing, is a big juicy question of our time. What’s the answer?
Noisy Humans, Noisy India
To prime my answer, I began with my example from my conversation with Daniel Kahneman about the US justice system that I mentioned in a previous newsletter:
“…if you take a crime for which the average sentence is seven years in prison, the difference between two randomly selected judges is four years. So that's the average difference. If you pick two judges at random, 50% of the time the difference between two judges will be bigger than four years. This is really shocking. That's noise.” (Brave New World, Episode 21 with Daniel Kahneman)
Well, I asked rhetorically, is this the kind of justice system we want? Most certainly not, most of the audience nodded along, much to my satisfaction. At this point, I deftly avoided the question of whether machines would be better than noisy humans just because they’re less noisy!
But the question was particularly appropriate in a place like India, and especially Kashmir, where decision-making is particularly susceptible to high variance depending on context. What I mean is that rules there are more like guidelines, and things are negotiable at every level.
I asked students to ponder their own experiences with the local bureaucracy. I shared a personal experience from the day before, where I had gone to a government office to sort out some official paperwork relating to my dad’s house. The clerk was quite attentive, taking notes. I thought, wow, we might get something accomplished. Then he received a call, closed the file abruptly, and scurried out with a parting “send me the remaining details on WhatsApp.”
Waiting for a human to pass some sort of judgment was a waste of a few hours. Imagine the frustration, loss of time, and corruption in such a system. Imagine the efficiency that can be realized from fixing it with AI. And that’s without factoring in the consequences of these noisy decisions.
The majority conclusion of this clever high school audience was that noise isn’t working out very well for India. While rules exist, they are applied inconsistently. In fact, most people don’t even know the complex patchwork of rules. Corruption is rife. Courts are massively backed up. Machines would sort out a lot of the simpler cases, and humans could do the exception-handling.
My sense is that increasing digitization of records, which are primarily paper-based at the moment, will catalyze machine-based decision-making to replace a lot of noisy human judgment. I think India is the poster child for why eliminating human decision-making will lead to a more efficient and fair society. Administrative systems need to be swift and just. Employing humans for appeals makes a lot more sense than using them to enforce rules inconsistently.
I think most of my audience agreed, even accounting for the noise in my judgment.
Knowledge
This shrine in Kashmir is devoted a divinity that symbolizes knowledge and learning. The color of the water keeps changing. It was pink, which I’ve seen before. It is believed that when the water is pink, it augurs well for the future!
Let’s come back to my conversation with Joseph Aoun. Joseph argues that a primary role of higher education in the age of AI is to make humans “robot proof,” that is, unreplaceable by machines. But how? Joseph argues for acquiring three kinds of literacies: computer literacy, data literacy, and human literacy. The last of these comprises creativity, innovation, the ability to work in teams, leadership, and a global outlook. These are the kinds of broad skills that machines will lack for the foreseeable future. Joseph makes the case that narrow and deep knowledge, which has worked well in the past, is a recipe for replacement by machines.
Aoun calls for an integration of “experiential learning” and classroom learning. How is this kind of learning different from practical or vocational training? How can universities enhance human literacy?
More on this in my next post.