We Must Find a Grand Purpose for AI (original) (raw)

I met with Satya Nadella on the morning of April 20, 2017. I had come up to Microsoft to interview Satya for my forthcoming book, WTF? What’s the Future and Why It’s Up to Us, but so much of what I wanted to discuss was already in Satya’s own brilliant memoir, Hit Refresh. We have a shared optimism about technology, and a shared conviction that the challenge of Artificial Intelligence is to define for ourselves and for society what is truly human, and to build a world in which AI reinforces and augments human capability and experience rather than devaluing it. So that’s where we started our conversation

Tim: One of the things you said in your book is that the challenge will be to define the grand, inspiring social purpose for which AI is destined. You wrote: “In 1969, when President Kennedy committed America to landing on the moon, the goal was chosen in a large part due to the immense technical challenge it posed and the global collaboration it demanded. In a similar fashion, we need to set a goal for AI that is sufficiently bold and ambitious, one that goes beyond anything that can be achieved through incremental improvements to current technology.” I love that thought, and I wonder if you could expand on it.

Satya: If you start with the assumption that AI’s benefits have to be about augmenting human capability – if I even look at the place where even Microsoft’s own engineers are most passionate, most driven about using AI – it is in assistive technology. It would be amazing to say, “Wow, someone who’s got visual impairment can now see ... or someone who’s got dyslexia can read.” That would be a real breakthrough! And, it would be so meaningful to be able to say, “AI is finally that technology that truly brings inclusiveness,” especially in a world where we’re talking about AI as possibly creating so much displacement. We could instead direct AI to create true inclusion in our society and full participation in our society.

A lot of this comes even from my own son’s experience. Because I have a special needs kid. He’s locked in. And I always think, “If only he could speak!” So, when somebody talks about brain-machine connection, I say, “Wow, now think about what that could do!”

We need to set a goal for AI that is sufficiently bold and ambitious, one that goes beyond anything that can be achieved through incremental improvements to current technology.

In a world where we are going to create this technology of which we can even say, “It’s the ultimate technology,” which is this path to super intelligence, the design principle that guides us is everything, the moral philosophy that guides us is everything.

In some sense the biggest path-dependent decision we have to make as a human race is going to be, “Who gets to design the super intelligence?”

Tim: One of the chapters in my own book is called “Our Skynet Moment.” My argument is that one of the aspects of artificial intelligence that we don’t talk about enough is that it might be a collective intelligence.We’re building collective intelligence mechanisms with search engines and social media and financial markets. And we have already demonstrated what we have told those artificial intelligences to do, which is to optimize profit over people. So we already have set Skynet [the hostile AI of James Cameron’s _Terminator_] in motion, and we have to wake up and stop doing that…

Satya: That’s right. Humans, in spite of all the cognitive biases we have, have a big advantage. As Herbert Simon said, we don’t optimize, we satisfice. That is actually helpful. There is a certain satisficing that is driven because of the moral grounding that we have, collectively. Now, when you say, “It’s an optimization problem,” which is what AI is, then we may have a problem. Then you have to wire in that moral grounding; so when people talk about symbolic grounding. I don’t think that that’s the breakthrough; the breakthrough we need is moral grounding of that symbolic grounding.

The design principle that guides us is everything, the moral philosophy that guides us is everything.

Tim: You also talk about building intelligence that augments human abilities and experience. You say you want to focus on how human gifts such as creativity, empathy, emotion, physicality, insight can be mixed with AI.

Satya: Let’s say there’s abundance of quote unquote “artificial intelligence.” Then the scarce commodity is real intelligence. Then you say, “Well, what is real intelligence?”

Real intelligence [will be found in] the most human of qualities, which are empathy, and compassion, and those will be the most important things. Until we truly, truly figure out how to embed that into some machines, we will have to amplify that with [and for] humans.

We will have to think through what education should be for humans. With AI, maybe you don’t really need to learn Fourier Transforms. You might as well go and do mindfulness, you may want to develop compassion... So that’s little bit of what I’m claiming. There is all this obsession that everybody needs to be STEM-educated, which is true. But, are we overly emphasizing STEM?

Tim: There’s a concept that has played a role throughout my career, which Clay Christiansen called the Law of Conservation of Attractive Profits, which is that when one thing becomes commoditized, something else becomes valuable. I first heard this from Clay around 2004. He and I were giving talks at the same conference. I was giving my talk about how Microsoft displaced IBM: PC hardware became a commodity, and software became valuable. My argument was that open source was going to commodify software and something else was going to become valuable. Eventually I realized it would be big data. And Clay was talking about his idea of the Law of Conservation of Attractive Profits, and I realized, “We’re talking about the same thing.” This law totally applies to AI as well. As the mechanistic parts of human cognition get commodified by AI, the uniquely human parts will be even more valuable.

Satya: That’s a great way to say it.

Tim: So in the future economy, there are huge opportunities based on caring, but also creativity. The creative economy isn’t just art and culture. We don’t eat food. We eat food mixed with ideas. Every time products become a commodity, we find new ways to make them valuable by mixing them with human creativity.

Satya: In fact, that is essentially a lot of our obsession, right? When I look at the Minecraft generation, or you can even take the Snapchat generation… to me, the most inspiring thing about that generation, it’s just a wild look at the creation, the stories, or what have you. In Minecraft, both boys and girls are creating these worlds that don’t exist except in their imagination.

That’s the generation I want to build to. Even in the latest release of Windows. We took a product that is used by 100 million people every day – Microsoft Paint – and said “Let’s make that an absolute amazing creative canvas for 3D creation.” Because, in a world of Augmented Reality and mixed reality, that’s going to be the thing.

Tim: You also talk about the ability to reason about large amounts of data and do pattern recognition more quickly. There’s a kind of creativity that gets expressed in the development of AI itself. Now, when people were winning the Kaggle challenges with Random Forest algorithms, the creativity was in feature selection, and now it’s in higher level hyperparameters, because now you use all the data. How do you see AI augmenting that creative process? The creative approach to problem solving.

Satya: When I look at even the code completion tools that we now have, or that we’re building – if you look at some of the AI that we have added to the traditional tools, what used to be spelling and grammar and phrase in Word – now it’s complete writing assistance. We’re at the point where you could say, “I want to write like Faulkner,” and it’ll start saying, “Here is what you may want to do.”

Tim: Yeah, AutoTune for writing.

Satya: It is. I didn’t go to a writing school, I went to an engineering school, but yet I have the dream of becoming a great writer and I need this sort of AI to help me.

And the reason I got to that, and that one is progressing well, is that I looked at Xiaoice in China, which is our conversational bot, and the conversations that people were having with it and the threads and the number of questions in a particular session. More of it was driven around people’s need to converse. And people were expressing their deepest thoughts. And I was saying, “Wow, this is where this notion that AI can in fact draw out empathy.”To have real world situations that you’re talking about, sort of some issue you’re having in real life, and have the bot reasoning with you to help you have more empathy for the other person’s condition is pretty surreal.

Tim: We’ve also talked about before about the idea of AI becoming a third runtime. Let’s talk a minute about that and how that actually happens.

Satya: The last time I think we talked, I said “The PC or the phone operating system was the first runtime, the second one was the browser and the web, and the third is quote unquote this personal agent that knows you, knows the world, and knows the context and is helping you.”

Then we said, “Well, really if you take both thinking and perception and ultimately action, all three of them, it’s there.” There is this agent that essentially embodies the augmentation of your thinking, augmentation of your perceptions, speech being one, vision being the other, and can take action.

That’s the ultimate runtime, right?

Can we get to a quantum computer? The advances are real – I mean, we are deep in it.

And that’s why in my book, interesting enough, I tried to write about these three things: 1) What’s the ultimate computing experience? The ultimate computing experience is the real world superimposed with computing. That’s mixed reality. 2) That will require AI breakthroughs in perception, cognition and action. 3) Then I go off on a little bit of a tangent and say, “Well if you want to ideally create that kind of computing power, you’ve got to break free of Von Neumann architecture and classical computing. Can we get to a quantum computer?” So those are the three things that I write about.

Tim: I was a bit surprised to see quantum computing get so much airtime in your book.

Satya: The advances are real – I mean, we are deep in it. But, the question is how quickly will it come true. I don’t know. But it’s no longer, “Wow, is it possible?”

Tim: Another great line in your book: “We want not just intelligent machines, but intelligible machines.” This is one of the big concerns that people have about AI, that we won’t actually understand how they come to their conclusions. Now, of course, this is true of humans, too. Because we tend to focus only on the rational side of human thought, we don’t realize how much of what we do is completely unintelligible to us as well. So, it’s sort of the last cry of the rationalist in some sense.

“We want not just intelligent machines, but intelligible machines.”

Satya: Sure. But I came at it two ways. One is the accountability. “What happens when you’re a designer and you create AI and you need to be accountable?” And the entire Tay experience really influenced our thinking around that. The fact that it ran amok because, of essentially people targeting it, and feeding it training data that could essentially hijack the algorithm. Whose accountability is that? I would claim it is ours. It is our responsibility.

Tim: Well, obviously this is Facebook and the fake news problem. And I’ve had really interesting conversations with security researchers about how could you fool self-driving cars. How would you hack them by creating things that they would misinterpret in their visual field? There are going to be all kinds of hostile interventions. And we’ll need more robust AI designs.

Satya: When people say, “Oh man, these DNNs (Deep Neural Nets) are magical. They just do stuff, we don’t even know” ... I agree there’s truth behind that statement, but the choices around the number of layers, the weights – there are a lot of decisions humans make in how these things learn. And we shouldn’t abdicate so soon to these algorithms that we have created, that we are not responsible for the decisions they make. I think then I said, “Okay, what is the moral equivalent of literate programming in AI?”

Tim: I want to move on to another topic, because I know we’re running out of time. One of the big themes that runs through my book is the importance of platforms, and how platforms have to work for all their participants. I’ve observed over time in the computer industry that companies play by a certain set of rules that make them great, and then they forget those rules as they become more and more dominant. I talk about how Microsoft began by expanding access to computing, and built a rich developer ecosystem, but then began to compete with that ecosystem. Developers then went on to other platforms. The same thing is happening now with Google and the web.

And you, again, hit this theme in your own book. We seem to be thinking very much along the same lines. You wrote, “When I became CEO, I sensed we had forgotten how our talent for partnerships was a key to what made us great. Success caused people to unlearn the habits that made them successful in the first place.” Can you talk a little bit about that?

Satya: My approach has been that any time you start looking at the world as zero sum, that’s the beginning of the end. Because that’s when you go defense versus offense. You start measuring your own surplus creation versus growing the market. This theme runs all through my book.

I even look at this and say, what the heck is the role of a multinational company? I mean, after all, I was born in a country that once was captured by a multinational company. And so that’s in the consciousness of anybody who ever grew up in India.

I think you cannot get that out of our heads, saying, “Wow, that’s what happens, when you say that’s the purpose of a multinational.” [Just to extract value.] So for example, to me, in Microsoft’s business model, it’s so important for us to say every country is going to be its country first, right? And so we would have to show up in, the UK, or in Germany or in the US or in India ... and say, what did we do for that country lately in terms of local opportunity? What is it that we did in terms of employment locally? What did we do for education outcomes? What did we do for small business productivity, large business competitiveness, public sector efficiency?

Any time you start looking at the world as zero sum, that’s the beginning of the end.

I’m a believer that capitalism can only survive if we have a world where surplus is getting created, and that surplus is more broadly getting distributed across all constituencies, versus narrowly. Or we could choose Marx’s late stage capitalism, and it’ll die.

Tim: Yup.

Satya: The reality is that our capitalist system rewards short-term surplus creation. That’s how billionaires get created. I’m not naïve to not understand that.

But, at the same time, one of the most fascinating things about Steve Ballmer’s USA Facts was that he himself came to the realization that as a billionaire, as he looks to say, “Oh, I’m going to give a bunch of money off to these philanthropies and non-profits.” he says, “Holy sh*t, most of these non-profits are actually funded by the government.”

That is what you’ve got to really start to think about, which is ... the public good. I’m a big believer in creation of surplus that’s more evenly distributed.

Tim: There’s one last thing I wanted to ask you, because I love this quote. You wrote, “We challenged ourselves: ‘At the end of the next year if we were tried in a court of law and the charge was that we failed to pursue our mission, would there be enough evidence to convict us?’” That charge should be brought for our entire society!

Satya: One of the things that I felt, was that in spite of all the success we had, we did not have that clear sense of purpose that was guiding the choices we were making, whether it was the products we build, the business model we constructed around it, what’s the ultimate reason that Microsoft exists?

And you can’t motivate anybody if there isn’t the real action behind the words. And that was the idea. In some sense, more than anything else in the last three years, that’s I think the corner that we have turned.

It’s a work in progress. We’ll never be done. It’s having that sense of purpose and having a cultural meme of learning, and then a growth mindset. Those are the two things that have been most helpful.

“We challenged ourselves: ‘At the end of the next year if we were tried in a court of law and the charge was that we failed to pursue our mission, would there be enough evidence to convict us?’”

I wouldn’t have said this three years ago. I would have been in the classic trap of the strategy of the day, and the brilliance of it, the product of the day. And what I realized is that everything has a growth curve, and then it sort of taps off.

What I realized is, you’ve got to be true to yourself. That’s why our creation story is such an important thing. A little bit being the Joseph Campbell fan, I look at it and say, “The creation myth of Microsoft is what should inspire us.”

One of the first things the company did, when Bill and Paul got together is that they built the BASIC interpreter for the ALTAIR. What does that tell us today, in 2017? It tells us that we should build technology so that others can build technology. And in a world which is going to be shaped by technology, in every part of the world, in every sector of the economy, that’s a great mission to have.

And, so, I like that, that sense of purpose, that we create technology so that others could create more technology.

Tim: But we also have to create technology that is empowering, not technology for its own sake. So, can you generalize that? You used that vision as a wonderful lever for corporate transformation. But we’re facing many of the same issues as a society. Can you generalize that thinking to ... how we should be thinking about our economy, our culture, our country, our world?

Satya: The way, at least, I thought about it is one of the most magical things that has happened in human history, it’s the United States. Think about the 200-year history of the United States. It’s unprecedented. Never before was there that level of egalitarian society, there wasn’t a constitution that existed that is like the American Constitution. The “blue collar aristocracy” never existed before it was conceived and brought to life in America. The fundamental ability to create more surplus, more wealth, more well-being for more people.

So here’s the real tragedy: That the American dream doesn’t remain a hope for the rest of the world. That would be the real tragedy.

Satya’s book, Hit Refresh, will be published on September 26, 2017. Satya has started a social media campaign asking you to share your own #HitRefresh story, explaining how you reinvented yourself by rediscovering your deepest values.