Artificial intelligence may seem like a futuristic concept, but we’re already experiencing it in real ways in our lives, whether we know it or not — in areas including speech recognition, spam filters and even loan processing. And AI is only going to get more sophisticated from here.
That was one of the messages from Oren Etzioni, CEO of the Seattle-based Allen Institute for Artificial Intelligence (AI2), founded by Microsoft co-founder Paul Allen. Etzioni spoke with us for this week’s episode of the GeekWire radio show and podcast.
Our conversation comes amid a boom in everyday AI, from self-driving cars to a computer that has mastered the game of Go. Microsoft put its stake in the ground with an AI-driven vision that CEO Satya Nadella calls “Conversation as a Platform,” with virtual agents working on our behalf.
Etzioni takes a much more optimistic view of AI than some of his peers. “The existential risk is just way overblown,” he says. “It’s much more likely that an asteroid will strike the Earth and annihilate life as we know it than AI will turn evil. It’s just improbable and hundreds of years in the future.”
Listen to the show below, download the MP3 here, and continue reading for an edited transcript of this week’s show.
Todd Bishop: Oren, in your current position, you really have a sense for the state of artificial intelligence. I think a lot of people out there see it in their daily lives in a very primitive form. They’re using Siri on their phones. They’re watching Google’s DeepMind beat a world champion Go player. The potential of artificial intelligence is there in a rudimentary form. Where are we now today in terms of the state of artificial intelligence, and where do you think we’ll go over the next three to five years?
Oren Etzioni: I do actually think that people are using it more than they realize. In addition to something like Siri, Google Search algorithm uses AI and machine learning all the time. Speech dictation on our phones whether it’s Android or iPhone has gotten tremendously better and that’s using deep learning behind the scenes to improve what’s called a speech recognition. Loan processing these days is often done in a highly automated fashion using machine learning. Credit card fraud, spam detection in Gmail.
As a matter of fact, AI is becoming more invisible and integrated into our lives. Of course, that can be a little bit scary to people. They say, “Wait a minute. Are we going to have Skynet, Terminator?” And so on. The first thing I want to do is even though AI is more and more being used, I want to reassure people that Skynet and Terminator are not around the corner for many, many reasons.
In fact, what we have are limited targeted use cases for AI, say spam versus not spam. Bing’s predictions in various context whether it’s the NFL or other things. What we don’t have is what’s called artificial general intelligence. A program that can do frankly what my five year old can do very well. My five year old is a lot smarter than any AI program and that includes AlphaGo which recently beat the world champion.
AI at the movies
TB: You mentioned the Terminator obviously. Is there one movie that’s come out recently that’s really captured accurately either the state of AI or where it’s headed? Did you watched any of the more recent ones?
Etzioni: I tried to catch all of them, because A, it is fun, and B, I feel like I need to know how we’re being presented in popular culture. I would say that probably my favorite movie was Her and it’s because it’s much more about human relationships. The AI was just used as a substrate. I would say that generally, Hollywood and a lot of the hype doesn’t really capture what’s going on with AI.
The truth is that behind any AI program that works is a huge amount of, A, human ingenuity and, B, blood, sweat and tears. It’s not the kind of thing that suddenly takes off like Her or in Ex Machina. This very complex nuanced conversations in its own will. It’s very different, number one. Number two, these movies are what’s called dystopian, right? They’re pretty negative. Transcendence, Johnny Depp takes over the world and we’re trying to turn him off and we can’t. It’s really not like that at all when you’re in the trenches.
John Cook: There is a negative perception around AI and even some leading technology folks have come out against it or saying that it’s actually potentially harmful society. Where are you coming down on those discussions? How do you explain this in a way that maybe has a more positive beneficial impact for society?
Etzioni: That’s a really great question and I would say there are three points to address. Point number one: this notion of existential risk. AI will take over. That just has no foundation. In fact, we haven’t seen AI systems do anything rather than very narrow task. Even if you think of one of the biggest success stories that’s unfolding which is autonomous cars, they’re autonomous in the sense that they’ll take you to your destination without hopefully crashing into anything. It’s not like a whole bunch of autonomous cars are suddenly going to band together and go takeover the White House or City Hall.
JC: Although people may want that Oren. Given the coming election cycle, I don’t know.
Etzioni: I didn’t say Congress. That’s a different story. If these guys were replaced by AIs, it would be already an improvement. We’re not going there. That’s number one, the existential risk is just way overblown. It’s much more likely that an asteroid will strike the earth and annihilate life as we know it than AI will turn evil. It’s just improbable and hundreds of years in the future.
The second point though is there is a real concern. I’m not that Panglossian, “Oh everything is great.” There is a very real concern that a lot of people have about jobs. Automation has consistently impacted jobs both by enabling outsourcing and offshoring and so on. I do think we need to, now, start discussing and studying the impact of AI in the labor force. I think that is an important discussion.
The third point is when you mentioned about benefits. Given the potential impact on labor, why even do AI. The answer is it has tremendous potential benefits. For example, at AI2, we’re building a program called Semantic Scholar that ultimately is going to help scientists, medical researchers and even doctors get the information they need to save people’s lives.
Another example is autonomous cars. The biggest reason we want autonomous cars is to prevent accidents. There’s more than 30,000 deaths on the highway each year. Many more people injured and maimed. Kids are out there driving distracted, texting and driving. There’s drinking and driving. All these things can be greatly reduced if we have AI in our cars.
TB: With self driving cars, would you get into one at this point?
Etzioni: I did get into one at this point. I took a Tesla for a test drive and I turned on the autonomous driving. Knowing AI software, my hands were inches from the wheel and I was pretty terrified. The salesperson was like, “Relax. Relax. It’s fine.” I said, “No, no, I’m not relaxing. I’m terrified.” Then a few minutes later, we’re on Aurora and she says, “Well, you might want to take the wheel now.” I’m like, “Why? I’m just getting comfortable.” She says, “Well, because I can see in the display that actually the AI has not perceived a large barrier between the lens and you could crash.” I was like, “Oh, okay.”
TB: At the Allen Institute, you’re really trying to push the boundaries. Can you give us a status report on how many folks you have and different milestones you’ve reached?
Paul Allen’s AI vision
Etzioni: Absolutely. It really starts with Paul Allen and his vision. Many people know him, of course, with the Seahawks and so on, but …
TB: I thought you’re going to say as the co-founder of Microsoft.
JC: Yeah. Really, I did too.
Etzioni: That little thing. Yeah.
JC: Maybe he has some AI inspired football team in the works here or something.
Etzioni: They’re using all kinds of technology in the Seahawks, but he’s also been really a huge visionary in scientific philanthropy. Of course, there’s the Allen Institute of Brain Science which is now hundreds of people. There’s the Allen Institute of Cell Science. We like to call ourselves AI2 because we came after Brain Science. That’s based on his passion for AI, artificial intelligence.
We started in January 1st, 2014. We’re now just about 50 people and we’re actually on a huge hiring spree. We’re hiring to grow to 75 people or so over the coming year. We’ve just gotten a bunch of extra space in our building in Fremont. We’re growing because … a lot of people say it’s the AI spring.
TB: When you’re hiring, what types of people are you hiring? The job market for traditional programmers, engineers is very difficult for folks who are trying to hire right now. Are you hiring in that talent pool or is that a different talent pool?
Etzioni: Hiring is very competitive. You’re absolutely right, but there are a couple of things that make us unique. First of all, somebody said to me, “Google has 30,000 brilliant engineers.” But they don’t have 30,000 interesting problems. We hire people who are sometimes fascinated by AI, fascinated by doing new things as oppose to maintaining software and solving technical problems that aren’t that interesting frankly at bigger companies.
That draws a bunch of people to us. Then a bunch of people are mission oriented. Our motto is AI for the common good. We’re looking for how AI can benefit society. We’re looking for the cutting edge, not just tweaks on deep learning or what have you. That’s drawn another set of people from the research community. We have researchers and engineers working shoulder to shoulder which, again, makes it I think a lot of fun for both.
TB: One of the projects that you did essentially was trying to get AI at the level of an 8th grader. Is that right? Can you explain that project and where you are with it?
Etzioni: I like to point out that I’m not trying to put kids out of work or even create a cheating tool for them to pass …
TB: Either of which would be fine.
Etzioni: Good point. Right, we’re not after child labor here. The issue is the following. People don’t have rational expectations of AI. Some people are terrified. The fears are overblown. There are all kinds of predictions like lawyers are going to be out of work in three years. Again, some people might not object to that either. We decided to rationalize the conversation, we’re going to take 8th grade science test as is and give them to the machine.
Rather than just giving them to our computer, we had a competition where we announce a $50,000 prize and we called for anybody in the world who wanted to submit an entry to build up their program on training questions and then be tested. Several thousand teams joined. 800 of them submitted results and the winners were announced in February. The interesting thing is that none of these were able to achieve higher than 60% on the 8th grade test.
TB: So basically a D?
Is AI smarter than an 8th grader?
Etzioni: Yes, exactly. Today, the state of the art in AI cannot do better than a D on 8th grade science, because that involves understanding complex questions and involves mixing scientific knowledge with common sense knowledge. It involves making sense of language and diagrams. It’s surprisingly difficult. There’s a really, really interesting paradox there. It turns out that what’s easy for kids, not getting a D on your science test, getting even a C, turns out to be hard for the machine. What’s hard for people like being the world champion of Go turns out the machines have surpassed that.
TB: Is that because it involves language and language involves so many variables whereas Go, there’s a constrained domain that the machine is attacking?
Etzioni: That’s exactly right. Go, pardon the pun, is black and white.
TB: This is the legendary Chinese game that sort of their version of checkers in some ways but much, much more complex.
Etzioni: Checkers is played on an 8×8 board and Go is played on a 19×19 board. It has more possible games. They say many more possible games than atoms in the universe. I’m not quite sure how they figured out how many atoms there on the universe, but they say there’s about 10 to the 80 atoms in the universe and about 10 to the 170 games of Go. That’s more than insults and hyperboles by Donald Trump.
TB: We got a political theme going through here. What did you think? This was the AlphaGo program powered by Google that beat one of the world’s top Go players. Was that a milestone for humanity because humans created that AI? Or was it a milestone for machines because machines were triumphant? Where do you come down on that?
Etzioni: I do feel like it’s a triumph of human ingenuity. On the one hand, we had Lee Sedol who’s an absolute genius. On the other hand, we had brilliant researchers and engineers at Google and Google DeepMind with a huge amount of effort building the program that did that. They built on previous effort. I like to say that AlphaGo’s overnight success was 50 years in the making, because they played in Korea, so the games were always going on at night, but they built on technology and research that goes back to the 50s and 60s and 70s. It’s really a confluence of ideas.
TB: When you look at that, do you think that there’s further potential for humans to just have no chance against the machines? At some point, any machine could be any world champion. Am I thinking of that correctly?
Etzioni: Humans have no chance in these discrete, ultimately artificial board games. The machines are going to win. When you come to something more nuanced, whether it’s broadcasting or radio show and interviewing someone or using natural language as simple as understanding a sentence. Reading a biology textbook and answering the questions in the back of the book. It turns out the machines are very, very far behind there. Language, knowledge, reasoning, very hard.
TB: If you were to look at AI2, your organization, the Allen Institute for Artificial Intelligence, what would success look like in three years?
Etzioni: We’ve defined very specific goals for ourselves. That’s one of the ways that we’re different from just a university who’s pursuing things for curiosity’s sake. Our goals over the next three years including increasing the number of people who are using Semantic Scholar or AI-based search engine for science, launching Semantic Scholar for neuroscience which we’re going to do in 2016 and expanding to other areas of biology, other sciences. Then in addition, we’re looking at improving the scores on these tests like the science tests.
Can machines think or feel?
TB: You’re just back from South by Southwest. Why would an artificial intelligence person go to South by Southwest at this point?
Etzioni: We were invited to a panel on can machines think.
JC: Can they?
Etzioni: It’s a complicated question, but the short answer is no, they cannot. Really cannot. Actually, then we expanded, “Can machines feel?” Again, the answer is no. People have overblown perception of what machines can do in this day and age. For rather esoteric topic, a room which was packed by 300 people and there are many panels and discussion groups and meetups about AI. I was amazed.
TB: It just goes to show this is part of the popular culture at this point. If not, a part of our lives to the extent that they will be in 5 or 10 years I guess.
Etzioni: Very much so. Again, things like the match against AlphaGo helped to reinforced that. I really want to remind people. I wrote an article saying, “AI will empower us. Not exterminate us. It’s important to understand that fears about existential risk from AI are way overblown.”
JC: Speaking of that, one of the critics of AI and we’ve seen this just here in the last few months come out with a real big non-profit effort with OpenAI is Elon Musk. What are your feelings on Elon and his approach to AI and what they’re doing at open AI versus what you’re trying to accomplish at the Allen Institute?
Etzioni: I have to say that hearing him use a very strong language like, “With AI, we’re summoning the demon.” That’s almost religious imagery. It’s a bit disconcerting to me as somebody who’s been working in the field for 30 years. I’m not trying to summon the demon. I’m trying to use AI to make the world a better place. To help scientists. To help us communicate more effectively with machines and collaborate with them and so on and so on.
I really can’t account for those remarks. The fact that he founded an institute and invested in AI companies, of course, his company, Tesla, uses AI integrally in their cars I think shows that at least his real attitude about AI is a bit more complex.
JC: Is OpenAI’s starting point just different than the Allen Institute’s starting point? Are they starting from a negative base versus a positive base? How do you see the differences between the two organizations?
Etzioni: Their technical focus is very much on what’s called deep learning or neural networks which is these pattern recognizing algorithms. Very loosely inspired by the brain. We think a lot more about how do we get AI programs to exhibit a modicum of common sense to understand natural language. There are big technical differences. Both are set up as non-profits, but recently, they have been back peddling a little bit saying, “Well, maybe if we invent something, we’re going to need to keep it.”
They say for the safety of humanity, I have to admit, I’m a little bit skeptical of that comment, because they say, “Well, if we find something valuable, we’re going to patent it.” If you’re going to patent it, then you’re putting the information out there. Is that really for the safety or maybe other considerations that are at play?
TB: Like commercial licensing?
TB: Is that something that the Allen Institute would ever do? I realize it’s a non-profit, correct?
Etzioni: We’re a non-profit. We do have a very small incubator and actually recently spun off a company out of their kid AI which has recently gotten a bunch of funding and some great momentum for their natural language technology. Fundamentally, we’re a non-profit focused on creating data and tools for science.
JC: Oren, you’ve spent a lot of time in this field. You’ve seen a lot of advances. What is the one problem in society that you would love to have fixed by artificial intelligence?
Etzioni: I think that as often is the case with technology, it’s a mixed blessing. It will solve a problem, but it will introduce new ones. The problems that I would love to see solved are jobs that are either dangerous like trying to deal with the Fukushima nuclear disaster or working in a coal mine. It’s generally unhealthy and dangerous. There are all kinds of jobs that in an ideal world, we’d let the machine do those. Also jobs with a tremendous amount of drudgery.
I would say that AI has the potential to save a lot of people’s lives. Whether they’re coal miners. Whether they’re people who’ve been hit by a car, driven by a person who’s drunk. Driven by a person who’s destructed. I would love to see AI save lives. The concern that I have and other people share is about AI taking away jobs. That is a real concern that we need to be thinking about and discussing. How do we implement policies that address that.
TB: If you netted that out over the long term, what is your outlook on that? Do you think that enough new jobs will be created by technology to compensate for the jobs that will be taken away by robots?
Etzioni: Frankly, I think it’s very hard to make long term predictions and nobody really knows. In the industrial revolution, that’s exactly how it turned out. 98% or so of society was working in agriculture and now 2% or less are, but we found other jobs for that. At the same time, we used to have a lot of great factory jobs and a bunch of those have gone away. Again, partially due to automation. Partially due to offshoring. It’s a complicated picture.
However, whatever you think will happen in the long run, in the short run, there will be very real disruption. Jobs will be taken away and those people need to be taken care of. People have floated the idea of universal basic income, of negative income tax, of training programs. We have an obligation to figure out how to help people cope with the rapidly changing nature of technology.
TB: Which jobs are most at risk? Apart from the ones that are just super dangerous. Is there a job out there that is extremely large in terms of its population right now that’s probably going to go away?
Etzioni: I think jobs that are relatively rote are in the process of going away. Even if you think about clerical jobs, it used to be that a secretary would be typing letters and doing correspondence. Now, you can just send an email. I think the more rote the job, rote white collar jobs are going away. I think over the next 20 years, all the Uber driving gigs are going to dry up. I think there’s a lot there.
TB: As we’ve mentioned, you are also a serial startup founder. The companies that you founded include Farecast which became the Bing Travel Predictor, Bing Travel. Basically predicting where the airfares would go. Also Decide.com which was a consumer electronics pricing predictor. You have experience in the tech industry beyond just research in the university and the institute.
Is there one company out there today in artificial intelligence that you feel like has the lead? Is there one big tech company or even a small tech company that really gets artificial intelligence and is poise to be the leader in this coming wave of AI?
Etzioni: Google has made phenomenal investments in AI. The acquisition of DeepMind in London which was called the most expensive acquihire ever because they paid more than 400 million dollars to acquire a company with 100 PhDs but no products and no revenue. It comes to mind.
Certainly, Microsoft, for decades, has been developing and investing in AI. I think Amazon has recently … I don’t want to say jumped on the bandwagon, but started major investments. If you look at Amazon Echo which some people predict to be a billion dollar product or platform, there’s so many opportunities there to improve its natural language processing. Its ability to carry out dialog and more.
JC: Has that device surprised you in terms of the success that it’s had? It came out of left field it seems. I think it surprised a lot of people in the tech press release.
Etzioni: I was very surprised. We bought one just to study it and I was surprised how useful we found it. AI works really well when you couple AI in a raisin bread model. AI is the raisins, but you wrap it in a good user interface and product design and that’s the bread. If you think about raisin bread, it’s not raising bread without the raisins. Right then it’s just bread, but it’s also not raisin bread without the bread. Then it’s just raisins. Echo is a great example of high quality raisin bread. It’s got some speech recognition. It’s got some natural image processing. It’s a great speaker. We listen to music out in the kitchen.
TB: I should tease ahead to our next show. You know that Alexa, the personality behind the Echo, was one of the most popular guest ever on the GeekWire Radio Show and Podcast.
JC: We asked a series of questions of Alexa. That was a fun show.
TB: Absolutely. She is actually returning on next week’s show. We’re going to be talking with her about some of her new features, so everybody should stay tuned.
Etzioni: You guys keep saying Alexa, and all over America, Alexa is turning on.
TB: It’s kind of like the Xbox problem in my living room. I have to say, “Stop listening.” Because it turns on when I say Xbox. Okay, go ahead.
JC: Back to the topic on jobs. That’s been obviously a big political hot-button issue. You’ve mentioned some of the areas that might be severely impacted. Just talk about manufacturing, because manufacturing is one that I think maybe comes to mind quickly. You have a political candidate here in Donald Trump who’s talking about bringing these jobs back. Is that a reality or should we let those jobs go? Where do you come down in that issue?
Etzioni: I come down on the side that income and equality, the way it is today, is a major problem and that we do need to make sure that people have meaningful activities and sustainable lives. Of course, when they have disposable income, that drives economic growth and so on.
I am completely sympathetic to the people who say, “We need to think about jobs.” Now, the question is what’s the solution. Is socialism the solution? That’s one candidate. Is xenophobia the solution? I don’t think so, but I think the problem is real and we need to think of enlightened ways to address it.
An artificial intelligence utopia
TB: If you were to paint the perfect world of artificial intelligence, what would that look like? I’m not asking you what it’s going to be like, but if you could create it, what would AI be doing on our behalf? What would we still be doing ourselves as humans? I’m curious where this could be headed in your vision of the world.
Etzioni: That’s a fantastic question, because so much though people talk about is dystopian and you’re asking what’s an AI utopia. An AI utopia is a place where people have income guaranteed because their machines are working for them. Instead, they focus on activities that they want to do that are personally meaningful like art or where human creativity still shines, in science. They’re engaged in those activities because of the interaction. Another one would be, of course, interaction between people and not because they need to make a buck.
JC: That’s been a promise forever with technology that it was supposed to make your life easier, but we find that it actually you end up working harder. A great example of that is email or a smartphone where now you’re connected all the time, so you end working more than a tool that was supposed to save you time. I’m just wondering if it’s just human nature to end up working harder on these things even though the technology is supposed to make life easier.
I think you’re absolutely right John that technology is a two edge sword. There is another side to the coin. Again, Todd was asking about a utopian vision. That’s the utopia. Is that the reality? I don’t know. I do think that an American society in particular, we do make that choice. By the way, we might be working harder than ever even in this utopia, but it would be on things that have real significance to us as oppose to pay the rent.
TB: Oren, as you said, you have 25 open positions. You’ve already hired 50 people approximately. What would be your pitch to folks out there to join AI2? Why does your organization matter in the world?
Etzioni: There are three points I’d like to quickly make. One is we have a mission to use AI for the common good. We’re trying to make the world a better place. Number two, we are absolutely at the cutting edge of technology, at the cutting edge of research and it’s incredibly interesting to work there. Three, we have a friendly and supportive environment. We think a lot about how to create a culture that’s welcoming to everybody, welcoming to women which has always been a challenge in technology. We really care about the people, not just the tech.
TB: How many hours a week does the typical AI2 researcher work? 80, 100? Is it a startup?
JC: They should have robots fixing this problem for them!
Etzioni: I think the way I would answer it is that AI2 employees don’t work one hour more than they want to. We have a tremendous amount of freedom and flexibility and people do work hard, but it’s because they’re engaged.
TB: How frequently do you end up conversing with Paul Allen and how intimately involved is he in the institute?
Etzioni: I have to say I’ve been completely blown away by Paul Allen. For somebody who has so much going on and so many different arenas. We are in constant touch. He is both an inspiration and a constant visionary force to drive us forward. He’s very engaged and have a lot of impact. Just the other day was our scientific advisory board meeting. He came in for the full day and mixed it up with AI experts from all over the world. He’s a hero of mine frankly.
TB: Of course, you’re designing the robots to eventually replace the Seahawks players.
Etzioni: We are not doing that, but we are designing AI that’s going to show up in his computer museum.
JC: Really? In what way?
Etzioni: All kinds of ways to show off the latest and greatest in AI technology as part of the history of computer science.
JC: I think it’s only fitting Oren that in our conversation here that something in the last sentence set off Siri on my phone. It just automatically started listening to what you are hearing. I don’t know. Something eerie going on here with my phone.
TB: He’s the AI whisperer.