Episode #73: Did A.I. Just Make My Life’s Work Obsolete? (transcript)
Quinn: Welcome to Important, Not Important. My name is Quinn Emmet.
Brian: And my name is Brian Colbert Kennedy. This is the podcast where we dive into a specific topic or question effecting everyone on the planet right now or in the next 10 years or so. If it can kill us or turn us into a gentler robocop we're in. Our guests are scientists, doctors, politicians, engineers, astronauts, and even a reverend. We work together toward action steps that our listeners can take with their voice, their vote, and their dollar.
Quinn: I was just thinking one day be fun if we actually talked to an artificial intelligence so we could add that to the list.
Brian: We have to.
Quinn: But would we even know we were talking to them?
Quinn: That's the question. This is your friendly reminder that you can send us questions, thoughts, ransom notes, and hate mail on important ... where do we ... on Twitter. On important not imp or you can email us at firstname.lastname@example.org. You can also join thousands of other smart people and subscribe to our free weekly newsletter at importantnotimportant.com. If you'd like for other human beings to hear your voice on our podcast, you can send us a voice message through the link in your show notes if I remember to put it there.
Brian, tell them what we talked about today.
Brian: This week's episode is asking if AI just took our jobs, except imagine our jobs were very, very important to like cancer.
Quinn: Right, right, right. Not these jobs.
Brian: Not these jobs. Not this job.
Quinn: Let me tell you something. These jobs, AI would of won the Webby.
Brian: Are you kidding? Of course they would of.
Quinn: Not even close.
Brian: We're just losers.
Quinn: Our guest is Dr. Mohammed AlQuraishi. He is lovely, incredible thoughtful, so intelligent. Not at all burned by what went down, and we'll get to that. So jet lagged, and still just delightful.
Brian: Really wonderful.
Quinn: Such a conversationalist, and so hopeful about the future. Which I've got to tell you, holy shit is that like a fucking oasis in the desert these days.
Brian: Yeah, keeping hope alive about future that is rough.
Quinn: I mean that is a cross to bear these days. Well, hopefully people enjoy it as much as we did, and feel like thank God there's fucking people out there who are still doing it. Instead are just like, fuck it.
Brian: Yeah. He's doing it, and he's hopeful so that's good.
Quinn: Gave me a little hope. I'm into it. All right. Let's go talk to Mohammed.
Quinn: Our guest today is Dr. Mohammed AlQuraishi, and together we're going to ask. Hold on a fucking second. Did AI just make me and my life's work obsolete? Mohammed, welcome.
Mohammed: Thank you for having me.
Quinn: For sure.
Brian: We're pumped. Thanks for being here. Let's get it going by just telling everybody who you are, Mohammed and what you do.
Mohammed: Sure, so I'm a assistance biology fellow at Highway Medical School. I work on computational biology specifically on something called the protein folding problem.
Quinn: Before we really dig into this. My reference to protein folding, my first understanding of it was when I feel like it was in the 90's and it was when, is that when they first launched that thing where you could use your desktop while it was sleeping to process. Was that fake, or was that ...
Mohammed: No, no it wasn't fake. I think it was late 90's probably. No, no it was the real deal. Was the real deal.
Mohammed: Yeah, yeah. They used PlayStation 3's at the time. I think that was the hot thing.
Quinn: Oh yeah, man.
Quinn: Oh yeah. To be clear, there was no downtime for my PlayStation 3. My desktop where I was supposed to be doing computer for your homework for sure. But PlayStation 3, no, no. All right, Brian let's do this.
Brian: Let's do it indeed. So Mohammed what we're going to do is just go over some context for our very fun question here today. Then, get into some questions that are very action oriented. Get to the core of why we should all care about you and what you're doing. Does that sound good?
Mohammed: Yeah, sounds great.
Quinn: All right, so Mohammed, we like to start with one important question to set the tone of things. Instead of saying tell us your whole life story, not that we're not interested in that. Everybody else does it. Why are you vital to the survival of the species?
Quinn: I encourage you to be bold, be honest.
Mohammed: Very heavy question. Let's see. Well, I mean, to be perfectly honest I don't think what I'm doing right now is vital to the survival of the species. What I do think is true, and it's not just me. I think the work that I'm doing and other people like me are doing. I hope we'll be pretty important simply in a few decades from now.
There's sort of two angles to this. One aspect is this question of synthetic biology of being able to engineer things, engineer organizations that do stuff like clean up their environment, clean up the ocean, or generate energy. That sort of thing. I think we're very far away from that right now. But I suspect with the kind of technologies that we're developing today will be sort of foundational to enable the synthetic biology. I think it's unclear right now where things like climate science or ultimately climate change will be sort of impacted by synthetic biology. But it's conceivable that it's 40 years from now it will prove to be one of the key technologies for that. That's the short term.
If you're interested, I can give you the long term version as well.
Quinn: Let's do it. I would love to hear the long term of why you're vital to the survival of the species.
Mohammed: I mean, pretty big question.
Quinn: Brian's already chipping away a monument to you, please continue.
Mohammed: Well, no. Actually, very clear and not just me. I'm speaking of the feel of the whole. I think there's a lot of interesting things going on in the kind of work I do. And what other people do, which is building machine learning models for molecular biology. What I think is particularly exciting looking further out maybe in a century now is that on the one hand you have artificial intelligence creating machines that are very intelligent. The other hand you have sort of bioengineering and biotechnical perhaps allowing humans to integrate with machines or sort of be augmented with machines.
And what's exciting to me about this is that for the entirety of our history, our existence, and that's 20,000 years ago what have you. We've basically been very limited by our biology. We see, we hear, we smell. That's been it. What I think is very exciting about the future is we may be able to rewrite the we experience the world. We may be able to redefine what it means to be human in sort of an exponential sense. That to me is ... Again, it's a very futuristic thing. It's not just survival, it's about looking forward to the future and think about what kind of species we can become. That's very exciting to me.
Quinn: It sounds, and we've talked about this before a lot like how Brian spends his days.
Brian: I'm constantly just sitting around thinking how can we redefine what it means to be a human being.
Quinn: That's awesome. Well, in a world where we desperately need more forward thinking and more practical action towards forward thinking. I mean we like to say the whole fucking climate change thing is such a ticking time bomb. But if we can not necessarily solve it, because that's a hell of an ask at this point. But slow it down and maybe at some point start to reverse. There is so much incredible stuff going on simultaneously right now that's laying the foundation for truly some paradigm shifting advances when it comes to technology or like you said biology. Space for example. The whole like we can send a reusable rocket into space a week after we used it, which sounded insane 10 years ago, but now we do it every week. Is going to be all for shit if everything's on fire. God, man if we can just get there with some of this stuff, might just be pretty amazing.
Mohammed: No, absolutely. And I mean many of these technologies are sort of double sided. I mean, take Michelle Leonard, think artificial intelligence. On the one hand there's a fear, a very viable very that may obsolete human jobs well sort of with ultimate chaos. On the other hand, I mean I think if we can make it work, it could allow us to solve all of these various problems that include climate change.
Mohammed: The question is going to be what wins? Can we do things in time to sort of ultimately sort of have a good outcome. That's by no means a guaranteed thing. I do think that there's a lot of exciting things that are happening. I think the future can be bright if we're smart enough and cooperate to get there.
Quinn: I love it. I mean I feel like we should just end it right there. But excited to dig into the rest of this. Okay, so sometimes this little contextual minute I do is pretty wonky. Sometimes it's more philosophical. I'm going to attempt to really define, because I feel like so much of the issue these days is like artificial intelligence is incredible. Even people who are [inaudible 00:09:34] in on it totally misunderstand where we are or what the even means, much less people who aren't paying attention to it day-to-day. This is all to say, Mohammed. Please just jump in and as rudely as you can interrupt me and tell me how wrong I am as I give everybody the back story on what it means and where we are.
Quinn: Does that work for you?
Mohammed: Of course.
Quinn: Please, okay, so just to again lowest common denominator here is the goal. So that everyone's on the same page for our conversation. Just revisiting some definitions and nomenclature around AI or artificial intelligence. First, AI means artificial intelligence. What does that mean? More or less layman's terms, capability for intelligence of some subject matter from something that is organic. Right. Not a human, or mammal, or a fish or something that's otherwise alive. How does that artificial system get intelligent? Quote unquote. Well, for a long time, it kind of meant we give it a shit load of free programmed rules, right.
Quinn: And a ton of data. Then, the system in question that you've now built uses those rules to execute a program, or calculate some now relatively rudimentary equation. Is that correct?
Mohammed: Yeah, yeah. There are some nuance. I think decades ago-
Mohammed: People were very interested in trying to build these expert systems. This whole idea of rules. Can you encode what we know about the world into these systems and then make them intelligent that way? It's almost like teach them how to be intelligent. Sort of build them from scratch with intelligence. I think that's happened the last 15-20 years there's been this shift toward machine learning. The idea there is that you're not sort of hard coding the rules. But instead as you eluded to providing the system with data and it can learn about the world from this data. There's a shift.
Quinn: Right. People have used the example of ... I mean, I don't know why this didn't make more sense earlier. Of course, we didn't have the computing power for it. It says if kind of how a child learns. Right. Deep learning essentially uses these specialized algorithms. They use multiple layers or neural networks as they call them to from what I understand gradually extract more for instance if we're looking at pictures of faces extract more higher level features from a face. So the first layer might do the edges of the picture. The next the background, or the outline of the face, and the next the nose. Et cetera. Et cetera. That does that on a million other images until it finds every picture with blue-green eyes. Is that a terrible version of how it works?
Mohammed: No, that's reasonable. I mean, you hit a key feature of this, which is that it's trying to sort of build a representation. It's trying to build a way that it can think about the problem. The way it does that is it takes something complicated like a human face. It breaks it down into smaller components. Into eyes, and noses, and then edges and so on. It's a key thing is that it's not really learning. It's sort of eluding to learn. It's learning to organize it's own thoughts, or thoughts are maybe too strong of a word. It's own representation so that it's sort of amenable to be able to find patterns in the data.
Quinn: Got you. It seems to have, and one of these seems to be advancing pretty ... Well, they both are pretty quickly. One came about without people realizing. One is finding images of very tiny tumors and lung cancer by assessing millions of pictures of lung cancer. The other is to actually produce something. We're seeing this stuff with deep fix. Or these pictures someone will post a gallery of 30 pictures online say, hey, these aren't actually real people. These are all drawn by artificial intelligence because of what it's learned.
For our case, what's interesting, from what I understand is a deep learning can either be supervised or unsupervised. I think from what I understand, and again that is such an important point from what I understand it's super limited. It applies to everything from ice cream to artificial intelligence. This is sometimes where there's some revelation at the end, and the scientists who operate this intelligence code. This is cool, we don't totally know how it got to this conclusion. Kind of like a black box as they say. So in these, we don't really teach it rules. We just like you said throw a fuck down of similar related data until it finds cat or tumor.
In some cases, and again, this is cutting a huge number of corners. I like how you said there's some nuance. Which yeah, I mean of course. It might be able to start to find those tumors faster or more accurately than say a trained radiologist. Or sometimes faster or years earlier than a pair of human eyes. Sometimes better than a fleet of trained radiologist. That's where we start to see things like we have this seemingly, this paradigm shift of a result. It gets published in a respectable scientific or medical journal. Then, a science focused blog sites it in a post. Then, the New York Times science section picks it up. Then, all of a sudden we have this mass media headline that says something like: AI Beats Radiologist at Lung Cancer Analysis with 94% Success Rate. Radiology's canceled, go home everybody because it changed when we cured cancer.
To be clear, that's not what's happening. It's not the full story. I love that you before we even started recording two days ago corrected me on just our intro to this, which is what I'm excited to dig into. Because it is a tool that we are working alongside at least for now. And it's also important to notice that these are again emphasize these are very specialized.
Mohammed: Yeah, absolutely. I mean ... Yeah.
Quinn: It can't drive a car. Right. The same one that finds a lung tumor can't drive a car, or set the temperature in your house, or water your flowers, or play go. Right.
Mohammed: No, simply not. In fact, I think there's almost a linguistic definition now where AI linguistic distinction. Where AI is considered fairly specific, fairly circumscribed technology that's really application focused while something like that AGI, artificial general intelligence. Is meant for something which is more aspirational that can do something like multiple tasks instead of going out in the world.
Mohammed: That's really much more of a doing right now. That the success that we've had recently are much more sort of in the narrow AI. That's ... that's I describe as very specialized.
Quinn: So, but the ones that are narrow, some of them are just making incredible jumps. Right. Sometimes it might seem as if if might, or some people might say, it makes your life work seem a little bit obsolete at your profession. Mohammed, why don't you just tell everybody how you got roped into this. What exactly happened with your profession and deep mind to finish getting us up to speed?
Mohammed: Sure. Like I said earlier, I work on this thing called protein folding which is a field that has to do ... I mean we can go into what that means later on. But basically it's this field that has to do with trying predict the shape of proteins. This is a field that's been going on for decades. I would say, maybe around 5 or 6 years ago, people started to think about machine learning, using artificial intelligence very seriously in applying it to this problem. That's impacted the field. That's made some progress.
What's happened very recently, last December in fact is there was this competition that happens every two years where people get together and try to do these sort of predictions. For the first time, there was an entry from an industrial lab, or from a research group called Deep Mind that's part of Google or Alphabet. They decided to participate in this competition. They in fact, did quite a lot better than anybody else in this competition.
What's interesting about it is that this group's expertise was primarily artificial intelligence. They had one or two people who had some biochemistry experience. By and large their expertise was largely in machine learning. While all the academic groups were almost the opposite. They all had sort of deep expertise in biochemistry. They were not necessarily machine learning experts. They're just learning their ropes in a way in machine learning. That sort of inversion I think was something of a surprise. That suddenly if you have deep talent in machine learning, you could tackle a new problem, or tackle a problem that you're not familiar with and make a lot of progress even more so than groups that had been working on it for a very long. I think that surprised a lot of people and perhaps made them a little uneasy.
Quinn: I think we've shared something about that actually when that news was coming out in our newsletter. It was very interesting. You eluded to this. Could you just take a quick pause and tell us exactly what protein folding is. Why it's important and I guess why you work on it?
Mohammed: Sure. Just again, going to the basics. Proteins-
Quinn: -That's where we live. The very basics.
Mohammed: It's a good place to live. Proteins are basically the kind of molecular machines in our body. Anything that does anything in your body is typically made out of proteins. So they're like little nano machines that stretch, that flex, that extend, that twist, and so on and so forth. The protein folding problem has to do with predicting the shape of these proteins. Every protein is typically made out of sort of a chain of molecules. We know where that chain is typically for any given protein. It's easy to sort of know the sequence that makes up a protein. How that sequence turns into a three-dimensional shape, that's a very difficult question. That's something that we've been trying to do for almost half a century.
We don't really yet have an algorithm that does that reliably. We could experimentally go and determine that structure, that shape, but that's very expensive. Takes probably a hundred thousand dollars to do it per protein.
Mohammed: Being able to develop an algorithm that can do that would be very useful, because it would allow us to in principle determine shapes of all proteins. Which then would allow us to do things like stimulate the cell perhaps, or begin to stimulate the cell. Which is what I started out with in the beginning talking about synthetic biology and trying to engineer biological organisms.
Quinn: Okay, I think that's super helpful.
Quinn: So this is ... feel free to just literally call me a dumbass.
Brian: Are you talking to me too, or just Mohammed? Can I call you ...
Quinn: No, Brian, that does not apply to you. I told you.
Brian: Got it.
Quinn: With this professional. I mentioned at the very beginning when my Windows 95 computer and my PlayStation 3 were cranking on protein folding like when my PlayStation 3 was the original deep mind.
Quinn: Were we trying to still accomplish the same thing?
Mohammed: Yes, yes but there was actually ... There's quite a bit of difference.
Quinn: This is the understatement of the century.
Mohammed: Well, it's a bit of a technical distinction. That's why I'm hesitant to say it's a huge difference. The difference basically is with the thing that you're talking about from whatever 20 years ago. Which was called folding at home.
Mohammed: There what that was really not only predict the shape of a protein, but to actually simulate the folding process itself. In your body, each protein is initially made in sort of an unstructured form. Then, it sort of dynamically, it literally over time takes on it's final three-dimensional shape.
Really what folding at home was doing and it still is doing is trying to simulate that entire process. What I'm describing now is sort of a short cut. It's saying, give me the sequence, and I'll give you the structure. I'm not going to worry about how we go from a to z. I'll just give you the final result. It's a technical distinction, but it's an important one.
Quinn: Wait does folding at home still exist?
Quinn: That's so exciting.
Quinn: I mean I have more than a PlayStation 3 now, but I did love my PlayStation 3.
Brian: That was a good one.
Mohammed: In fact, I think this is kind of an interesting Factoid, but I think collectively it is a larger super computer in the world.
Quinn: That's so cool.
Quinn: Does it become sky net at some point? Hopefully not [crosstalk 00:22:30].
Brian: Yes, right.
Mohammed: That's exactly why they're very specialized type of algorithm, so it's probably one of those kind. That's probably one that's kind of.
Quinn: Probably he says. Good, good this is great. This is all going so great.
Mohammed: All of us sleep safely at night.
Quinn: I don't. It's fine. Don't worry. I know too much.
Brian: Hey, so can we ... let's talk about misconceptions and hype a little bit. If you were writing these news article headlines. How would you frame what happened that day and what's happening every day?
Mohammed: I would probably say something like substantial progress made in the protein structural prediction problem by industrial research lab.
Brian: Yeah. I mean that sounds so much more correct, and also at the same time you understand why whoever's writing for the New York Times wouldn't write that.
Mohammed: Of course.
Quinn: So I want to get into, and again like we've talked about. There are more and more of these specializations every day. There's so many businesses across every spectrum that are building a version of artificial intelligence into their business. It feels like everybody from the postal service to Uber to protein building in a sense.
Mohammed: Yeah, I mean I should say I actually do think that this is one of the fairly extraordinary and almost generational moments in science. I mean, there's only a handful of these things that I think you can point to see in the 20th century. These kind of transformations. I think this is one of them. On the one hand, I think there's a lot of hype. And on the other hand, you don't see these sorts of things more than once or twice in your lifetime. I mean, in the broader thing. Not talking about just protein folding. I'm talking about what you just referring to. The fact that machine learning is just revolutionizing so many different fields. It's really quite ordinary.
Quinn: Look, I have no doubt it's up there with the printing press as far as the liberation of data. But when we're talking about practicalities and we try not to ... We try to really talk about how something is effecting somebody now or in the next 10-20 years. That way they can kind of take action. To be clear, as we've eluded to and I want to dig into more how it's more of a tool for you than it is replacing you. There are a very large number of very common jobs that are going to be replaced by either a version of artificial intelligence or automation. In many cases, artificial intelligence isn't a helpful new tool for a mortgage loan officer. It might of started off that way, and now it's actually the new mortgage officer. Bye, bye Jeff. Right.
A decade from now, all indications are that the first sort of autonomous driving as much as we've realized how much further we really are consumer driver is probably going to be the long haul truck driver. Right.
Quinn: Because highways are so much less complicated. The problem is truck driver is the most common job in most states in America. From FedEx drivers to actual long haul truckers. It's something like 4 to 6 million jobs. My question is then turning it back is why is your profession different? Why does AI remain a tool while you're still running the show?
Mohammed: To be clear, in the limit of infinite time, we'll all be obsolete.
Quinn: Sure. We're all turning into like beams of light at some point. No doubt.
Mohammed: Yeah. There's nothing for the mentally different about my field then say any other field. I would say there is maybe two general things there that may play a role in some fields versus other fields. One is just how kind of technically challenging, or how much sort of scientific or just technical expertise you need. Our machines are good at certain things like battery recognition. They're not yet good at reasoning, logic, those sorts of things. If a field requires a lot of backing a technical scale. It's just machines just can't yet do it. It's as simple as that. It's not clear that it can do it in the next 10-20 years. They might. But there's no clear path that we could see from where we are today to where that is. That's one class of jobs that I would say is at least for foreseeable future that some what immuned.
The other is one that requires a lot of human interaction. You mentioned geology earlier. I mean, a part of geologist's job is to interpret data. All of it has something to do with interacting. It has something to do with interacting with patients. Trying to sort of understanding symptoms, diagnosis, and that sort of thing. That is not going away, because that's going to require a lot more development before a machine could emulate the human interaction component. Those I would say are sort of two classes of jobs that are quite different than say something like truck driving where you describe may potentially be obsolete in something like 10 years from now.
Quinn: There's a really interesting, and to be clear, much of the human conditions in a bad way these days. So hopefully we don't replicate it perfectly. But there's this really interesting story, and I'm going to mingle it. I'll try to find the headline for our show notes. Some guy built for ... So much of an issue with these huge elderly populations in Japan and that's growing in the U.S. once the boomers actually get to elderly and for instance, France. Is how isolated they are and the loneliness issues.
Ignoring that for one second, somebody was trying to deal with just the general day-to-day care that they often need if they're not in a bigger facility that's built for that, because you know a lot of these folks have a hard time or will refuse to leave a house they've been in for 30 years. Have a hard time or will refuse to leave the house they've been in for 30 years. I know it happened with my grandparents.
He built these little robots that drive around their house. It has a screen, and it can check in on them, or they can ask for someone, or they can call a family member or whatever. Very quickly they've realized the number one thing the robot was being used for was for these elderly customers/patients were calling customer service all the time. All the time. It was literally just to have somebody to talk to.
Quinn: I hope I'm interpreting correctly what you were eluding to with radiologist and the medical profession as far as a general practitioner, which is the tools are going to be even more and more useful, and more accurate and more helpful. And save so many lives or find new treatments, or ways of identifying things, or tying different medicines together that we might not of thought that would go together. Like how Viagra was an accident. Penicillin was an accident. Things like this. That human experience is not going away. Maybe it can free that profession to become, or liberate that profession to become a little more centered towards that. Where you can go to a doctor and be like, he's a genius. His bedside manor sucks. At least he saved my life. Well, maybe we can focus on people that just are better at that part of it. And are relieved from at least wondering am I keeping up with the latest medical journals on what this thing might be.
Mohammed: Yeah, that's actually interest. To be honest, I think you're quite right. This is something that I suspect will be true across a variety of fields in fact. I mean you mentioned earlier these sort of generative models. Things that could generate deep fakes like pictures of people what have you. But also then there's some interesting developments where they allow you to take a very rough catch of the landscape and turn it into high revolution image, right. I think there's an episode in Star Trek, some number of years ago, I think had this idea.
Quinn: Let's do this.
Mohammed: Yeah. Well, it's really interesting. As a creative person maybe you don't need to develop your skills to the point of getting every sort of fine line correct. You just need to have the vision, the idea, the creative spark. Then, the machine can take that all the way to a finished product. To a beautiful image or orchestra what have you.
Mohammed: I think this is going to be true across a wide range of fields where we're going to in some ways become more general. We're going to sort of be responsible for the high level decision making, but the machines will be responsible for executing all the way down to the fine grain details. And I think medicine will be in the sense of that.
The other think case is, one other aspect though, which is that I also think this almost sounds silly, but there's a mechanical aspect. Just the fact that you have a human coming into being able to touch. Being able to measure to. These sorts of things actually fairly difficult surprisingly in some ways for machines to do. Robotics is still very much in it's infancy. The mechanical aspect of a human is something that may prove to be surprisingly doable.
Quinn: Sure. I mean look at the Mars Rovers, they're credible feats of engineering and ingenuity, but they drive about 10 feet a day and can barely pick things up. Can't bring anything back. We could walk that far in five minutes and accomplish what all these rovers have done in 25 years.
Mohammed: That's correct. Exactly, yeah.
Quinn: We'd also have to provide oxygen and radiation shielding it and all that shit. Right. There's pros and cons to both. I thought about what you were saying too about again using it more as a tool. I was reading something the other day about how there's always this ... I remember when I first started working. Everyone's like, you have to go to business school. Half the people are going no, you can learn everything you need on the job. The dichotomy there. There's this argument now going forward for training people of the jobs of the future and the problems of the future to be more of a generalist or to be a specialist.
There was an argument to be a generalist would be helpful, because so many of our problems are interconnected and so many ways. We're finding out more and more of that everyday. Why do we have cancer? Turns out our air's been polluted for 40 years. Et cetera, et cetera. That knowing a little less about more things, but knowing that there is a tool that can go deep for you should you want to investigate some of these things or ties things together. It seems like we finally have asset for the first time perhaps.
Mohammed: Yeah, I agree. I think absolutely. I think the generalist ... There's probably will always be room for a small fraction of the population to be sort of ultra specialized. These are probably the people who are super, super smart or extremely. I think for most of us being a generalist will prove to be a more viable path forward. Even things like soft skills, human skills, those sorts of things. It's interesting. I always wonder about the direction of human history. I think we've certainly through the 20th and 19th centuries started placing more and more emphasis on developing very specialized skills. The way even society bifurcated into these sort of specialized vocations. It's sort of interesting thinking about what that was like a thousand years ago, I think that was not the case. I suspect we'll see a reversal where being sort of well rounded human being will prove to be an asset because you're able to navigate so these complex terrains much more effectively. Just rely on machines on trying to take care of the nitty gritty.
Quinn: I mean hopefully right. We're always going to need specialists like you who can actually interpret the data the first time to make it something the generalists can understand and operate under.
Mohammed: Yeah, but actually if I were to take a guess. It's very hard to predict. If I were to take a guess, I suspect that people like me will probably be obsolete before the generalist. Again, it's hard to say. I think over a 50 year timeline say. I think I'm more likely to be obsolete than someone who's more of a generalist.
Quinn: I hope you're never obsolete to be clear.
Brian: This is great news for me. I'm definitely a generalist.
Quinn: That's probably the most gentle way to describe what we do here. Go ahead, Brian.
Brian: Say I'm a young person interested in doing what you do.
Quinn: Who wants to be a specialist.
Brian: Who wants to be a specialist despite what you just said about the obseletion. Why is there hope? Tell us why there's hope and why I should still be excited?
Mohammed: So hope in the sense of why this is still interesting to do, or hope why there's still a viable career path?
Quinn: I mean I think it's of course going to be interesting, because we still have so far to go despite everything you've done. Not knocking your accomplishments. More like because of the tools that are now available. Having so far to go means we still have so much opportunity. I guess if I'm understanding it. Why is it viable as career progression if anything more viable and exciting than it was before?
Quinn: Is that true or false?
Mohammed: Yeah, no. To be sure, we don't know ... I think what's happened in the last 10 years in artificial intelligence is amazing, but we still don't have a clear path forward from where we are today to sort of artificial general intelligence. What I was talking to about earlier in terms of logic, in terms of reasoning. These sorts of skills. They're still very much the preview of humans. We don't have any competition yet. It's not clear that that's going to change in 20-30 years. I mean, that's the primary reason why I would say it's a viable career path. There's no machine like now today that's considered to do what I do, or what people like me do. There's no sort of ... The writing's not the wall in a sense that we can just see that we can scale things up and get to a point where machines can do what we do.
My statement earlier was more just [inaudible 00:36:56] statement. Just saying one or the other I would suspect that the specialist was sort of go the way of the dodo before the generalist. I think if I were inviting somebody today to start college or high school. I would still if they have the aptitude or the interest, I would still very much tell them to go into this space, because I think the option is amazing. There's so much that we still have to do. If we succeed, like I was talking about earlier, I think we could do really ... We could make life a lot better for a lot of people.
Brian: That doesn't sound terrible.
Quinn: Yeah, it's pretty admirable.
Brian: Yeah. How should I be proactively trying to-
Quinn: -Not, not you, Brian.
Brian: Yeah, no, me.
Quinn: Sure, okay, go for it champ.
Brian: Trying to engage these new tools and disciplines to get ahead and to make a dent in something like protein folding.
Mohammed: Well, so there's a few things. I would say more broadly if you think about the health space. What I think is interesting are these new platforms that are being developed, and there are some privacy issues you can be sure. There are these new platforms that basically allow people to provide their data in a way to itemize or what have you. Just build databases that could associate things like genotypes with phenotypes.
For example, in the U.S. we have the ... What is it called? All of us. All of us. It's an initiative. I think it started out in the Obama administration. Basically it's an initiative that's trying to collect data from 100,000 or even a million people eventually to associate the genotypes with phenotypes. That's a great thing you can just go in and sign up for it. You could always whatever measurement's taken that would actually cough all that money for free. You get to have access to all your data as well. That's something that's actually ... It sounds maybe surprising, but we're in dire need of it.
Because in the biomedical community right now everything is very fragmented. Each hospital has it's own data. And for the most part, hospitals don't really share data with one another. So it's very difficult for researchers to be able to actually apply these wonderful developments in machine learning. Two things like health.
If more people would participate. If we're able to on a national level build databases that they can sort of protect privacy, but also expose important patterns. Then I think that can be a win-win.
Quinn: Yeah, you hope so. We had a couple conversations with some brilliant women over the NHS. Talking about how data sharing actually is the default when you're born over there which is interesting. How that is now a really interesting case study compared to here where basically it's opt in here, versus there it's opt out. While they have a much more homogenous population. Obviously a much smaller population. What they're already able to start doing with that data, because it's a much more comprehensive, if not completely comprehensive swath of their population there.
Mohammed: Yeah, yeah. Absolutely. Iceland's probably the best example of this, because they have this remarkable ability pulling data on people. Like you said, and the population's ... But homogeneously you're actually able to infer the relationship within individuals in a population and so on. To be honest, these things do raise important privacy issues. But what we're talking about in the context of the U.S. at least I think it's always much simpler. We're just asking for a lot less. It's not even ... Because you don't have a homogeneous population you're not so much worried about being able to tell who's related to who. That sort of thing. It's much more about these kind of disease associations that are very hard to do all right now, because we just don't have enough data to actually to discover these patterns.
Quinn: Do you think that we are at least making strides toward some serious advances in these places that it is worth some of the privacy trade offs that people are scared of? Is it something where you can say, look, I know it's scary but it could be so worth it in the near future. Are we close to that? Is that a leg to stand on?
Mohammed: I would say two things. I mean, the answer to your question is yes. I mean, I think we definitely ... There's definitely a lot to do, and I think we are at a cost of making some important discoveries based on technologies which have been developing over the last two decades. I mean people have always been saying, treat the cancer is right around the corner and it's sort of never materialized. I do think that the next 10-20 years we will see sort of really exciting and practical progress. I'm not sure that the premise is correct of your question. I don't actually think there are any real privacy trade off. I think we can do this in a way that it's almost completely ... There are different versions of this. I think we can get 90% of the way there without sacrificing almost any of our privacy.
This is part of what I think much of the problem really has less to do with ... I mean I think people are often quite willing to share in an itemized way. The problem has more to do with different institutions, different hospitals, different research institutions sharing information. They tend to not want to do that. That's why I think we need something like a national ... We do have this national effort now. Because that would sort of by construction ensure that data is actually being shared.
The problem is not even so much privacy I think it's really just sort of different groups can be very protective of the data they have. Not because they're worried about patient privacy, but because they want to make [inaudible 00:42:50]. It's like a selfish thing.
Quinn: Interesting. Yeah, that's a much more rational perspective. I feel like from a consumer standpoint, it's less I want to give up my data, and more from Target to Home Depot to Yahoo they see these data breaches, and go well that's fucking happens when I give up my data. I think that is probably the scare point from our folks. Well, shit now if I'm going to start giving up my medical data. What happens when that gets lost? Obviously these companies are now, every time that happens ... hopefully I think these companies are doubling down more and more and more on security and various lock steps for it. I can empathize with it, but also again can see the power of how close we are on some of these things.
Mohammed: Yeah, absolutely. And eventually the tricky thing with that with financial data for example is that cannot be anatomized. Right. I think that's a big difference here is that the point is not to making something like your credit card score for your health. That thing would be a bad thing. It's unnecessary anyway. It's more it's just to collect data in bulk, and to be able to discover patterns. But without being able to even in principle to look back who has what disease or that sort of thing. That I think is unnecessary. That's why I'm saying I think if you do it in a way which really should not have any privacy certain applications at least 99% of the privacy questions or issues could be resolved.
Quinn: Feels like we need more people like you to explain that point to folks, because I think that makes a lot more sense. Which is we don't want your social security number.
Mohammed: No, no. Absolutely right.
Quinn: That's not actually going to help us.
Mohammed: See, it gets a bit complicated right, because for example if you were to send your data to consumer genetics company that's a different calculation. There are in fact interests in maintaining that association. It's not like I'm sort of beating at a horse. This is why sort of national effort is a bit different. The intention for the get go is not to monetize this. It's sad to say we want to collect this data for research purposes, so we are going to do it in a way which is maybe not particularly useful for commercial applications, and so we'll lose some edge. From the scientific side, we can do it in a way that save guys their privacy concerns while at the same time being almost good as having the full and atomized data there.
Quinn: Sure. So in your profession so far, besides deep minds showing up and blowing up your spot. What are the biggest obstacles you do run into in your job on a day-to-day and a long term basis. We want people to understand it's not all roses over there with your protein folding.
Mohammed: No, no. Absolutely, not. That's actually part of what made the deep mind entry sort of interesting in a way. Because I think it highlighted structural inefficiencies and structural problems in the way academic research is done. To answer your question, I mean so one common problem that I think almost all academic researchers complain about is access to computer sources, where actually relative to companies we have a lot more limited access to compute. To be able to sort of carry out these very large computations. This may sound like a very sort of technical issue. It's become really sort of the bloodline of computational research.
Frankly computer science has been having to deal with this for the last 10 years, because there's been this almost exodus of academic researchers from universities to companies precisely because of the resources that they're able to amass or to leverage in companies. I think we're beginning to see that in biology as well. That's something that we run into everyday, and it's something that will require I guess something more of a national effort at the level of the Department of Energy or the National Institute of Health to say, okay, this is a serious issue and we need to be able to provide these resources so that academic research can sort of compete at the highest levels with industry.
That's one thing, the other thing I would say is just sort of software infrastructure. The way sort of academic incentives are structured. People are incentivized to create novel pieces of software, or novel research ideas. They're not incentivized to create robust software that actually works and that people want to use. There's this very strange thing where academic software is sort of synonymous with being sort of very fragile, and practically useless and hard to use. That hole fulfilled back. There is no reason for it to be this way. Again, sort of going back to deep mind where they have. They understand this issue very well. They've built sort of card way of software engineers that are able to go in and engineer things that are very robust, and that can scale it. A lot of the researchers, they actually operate at a very high level. While academic researchers on the other hand are sort of left in the dust, because we just don't have access to this human resource.
These are issues that I think the scientific enterprise as a whole and the funding agencies in particular can pay a lot more attention to sort of restructure the way science is done so that academia is a healthier place.
Quinn: Well, and that's what happened with Uber and Carnegie Melon 7-8 years ago right. Because they just completely decimated that entire department by hiring everyone for their self driving car unit.
Mohammed: Yeah, absolutely. It's not just ... I mean that's probably the most extreme example, but frankly, almost all the top elite universities, I mean they've really suffered. Especially in CS departments, computer science departments exactly for this reason.
Quinn: Yeah. It's an issue. Things to fix. Brian, go ahead.
Brian: Let's get into some action steps here. We always want to end this with things that our listeners can do to support your mission. Actual actions they can take. We like to say that they can use their voice, their vote, and their dollar. Let's get into that. Let's start with their voice. What are big, actionable, and specific questions that we should all be asking of our representatives in an effort to support you.
Mohammed: I mean, you ask my answer will be very general. I mean, I think basic science is important, and we ought to fund it more. There's a slight distinction there where I think there is arguably a reasonable amount of funding. There can always be more for clinical or for medical research. The kind of basic questions tend to be sort of left behind. That I think is very short sited, because ultimately the long term breaks things like crisper, we talked about that earlier. Have come from people asking very fundamental questions that don't necessarily at least initially have any connection to sort of practical matters. I would say, tell people, tell your representatives to support investment in basic science. To actually increase funding for basic science and science funding in general, because it's one of the few things where I think that it's an investment. The intent on investment is almost guaranteed. It's one of those things where of all the things you can spend your money on, I think research is one of those silly safe bets.
Quinn: We've gotten that answer a lot, which is just tell them to support basic science funding. Because it's almost like you just never know what you're going to get out of it. There's no wrong that can be done there.
Mohammed: Yeah, it might seem a little bit intuitive. Historically, it's not been the case that basic science has had to ... and you make funding. Then you can find what we see now. But there's been this shift, and particularly in the last decade I would say especially in life sciences. Where practically everything we do now has to cure cancer. That's really short sided. I'm not entirely sure whether the force of that play that led to this situation. And partly I was thinking it was sort of justify the taxpayer saying, I'm going to cure disease. I'm going to discover some new form of bacteria. The point is to try to communicate this truth that progress does seem to come from asking basic questions. The more we know about how the world works, the more we're able to make progress in very applied areas of research.
Quinn: Yeah, I think that makes a lot of sense.
Brian: Yeah. And I guess as we move on to vote here, we can kind of ... There's a similarity of vote for people who are supporting this, and ... Are there specific scientific perspectives that are missing from our elected representatives?
Mohammed: Yeah, I was going to say, I think when it comes to vote. I would say, vote. Vote for people who are scientifically aware, progressive and so on. The obvious things that need not be said. I would also say actually, if you have a scientist running in your district. Give it a serious look. I think scientists as a community bring a certain perspective included to policy making that may be quite fresh. That that's quite different than somebody who's spent their entire career in politics or in business. I think as a whole, scientists are sort of very much underrepresented in Congress.
Mohammed: It's not just about them being pro science, that's one aspect of it. I also just about them having a different sort of more evidence based perspective on things. I think they're less likely to be driven emotional currents. More likely to be thoughtful about how to approach a bigger problem. I think that's a good thing to have in general.
Brian: I mean, I would like to scream that from the rooftops. It seems just like duh, right. It's insane that the number of scientists that are in office. It's crazy low. I forgot what the stat is. It's just mind blowing.
Quinn: It's great that people are in their jobs doing science, but at the same time we had a conversation last fall with a gentlemen, Sean Casten, who's now representative Sean Casten.
Brian: That's right, Illinois.
Quinn: From Illinois who ran for office and part of his platform was like, look, I've built clean energy. That was my job. My perspective is inherently, extremely valuable in the conversations we're going to be having in the next 5-10 years. Because I've actually done this, and I can speak it, and I can understand it. I don't have to have science 101 come to my office to tell me how a windmill works. It shouldn't be all those people, but the more of it ... Look at Representative Underwood, the young black nurse. Who made it and we're trying to fix healthcare and that perspective is certainly super helpful. Someone who's in it and doing it everyday.
Mohammed: Yeah. I think especially to a very polarized environment. The problem is I just think we've become too emotionally caught up in the moment. We're not able, I mean this sounds a bit cliché, but it's true. I mean I think we're just not really talking to one another across the ideological divide. It's certainly about ... I mean that's one issue. Right. It's just that it's the matter of being able to take a step back and just sort of rationally think about something and not get caught up in the tribal kind of warfare.
Mohammed: That all countries tend to send it into. That alone I think just by itself would again with wild thing to sort of have in Congress. I should say, I think after 2016 I mean my sense is that within my circle I feel like there's a lot more scientists that are sort of getting interested in this, so that's great. To see more people getting engaged in policy, and thinking okay, maybe we can compete in this way instead of just doing research. I think there is a positive sort of development in that direction, but I think it needs to go much, much further.
Brian: Yeah, it's a nice start and one of the good things that came out of that election is just how more fired up people were especially scientists.
Quinn: We work with a great group called 314 Action and 314 Pie, they work explicitly to support scientists and things like that getting into office. Seeing the huge Darth that exists there.
Brian: Mohammed, what about money? What can our listeners do with their dollars? Anything specific outside of the box where we could send some cash?
Mohammed: That's an interesting question. I actually ... Well, so there are organizations. I don't know whether they're taking donations. I should call them up, because I think they're doing very good work.
Mohammed: One of which is the Zuckerberg Initiative. So if you're not familiar with them, they're sort of this project research institute, apart from the [inaudible 00:55:34] organization founded by Mark Zuckerberg and his wife, Priscilla Chan. One of the interesting things they've zeroed in on is this question of being able to provide software, software development, software expertise as a general good to the scientific community no strings attached. I think they understand the value of having that sort of resource be publicly available.
I don't know if they take donations. I think we need more organizations like them that support science while being aware of the structural inefficiencies that it has instead of trying to actually be sort of intelligent about how to do these specific inefficiencies.
Quinn: I think one of my friends runs PR for them. They do some really interesting stuff over there.
Quinn: Awesome. Awesome. Awesome. We can't thank you enough for your time here today, Mohammed. We just have a few more questions for you. If you can think of anybody else, and you can send this to us later. We always love getting recommendations from our current guests about other ground breaking, world changing, folks out there that we would love to talk to. That our folks, our listeners would love to hear from. If you have any ideas on that we would love to have them.
Mohammed: There are so many interesting people. Can I send you a list later?
Brian: Yeah. Please.
Quinn: Yeah, absolutely. Yeah, for sure.
Brian: That would be wonderful. I appreciate it.
Quinn: Absolutely, you can't put Brian on that list. We've already talked to Brian.
Brian: Yeah, I'm always on this thing.
Quinn: It doesn't count.
Brian: Are you ready for a lightning round, Mohammed?
Quinn: It's not a lightning round, but we called it that because I haven't changed it in 72 episodes. Mohammed, when was the first time in your life when you realized you had the power of change or the power to do something meaningful?
Mohammed: I think 15. 15 years old. Do you want to know the context?
Quinn: Hell yeah.
Brian: Yes, please.
Quinn: Hit us.
Mohammed: This is going to sound strange. I was trying to get my green card at the time. I was young. I would say my parents were trying to get a green card, because we had just immigrated to this country.
Quinn: Where are you from? I'm sorry, Mohammed.
Mohammed: Yeah, yeah, sure. I'm from [inaudible 00:57:37] originally.
Mohammed: At the time, the setup was you actually have to lineup in the middle of the night starting at 10:00 PM in front of the immigration office to wait to get in for your interview. So it's this very bizarre thing where you have to line up at 10:00 PM and interviews are usually at like 10:00 AM.
Mohammed: So you're spending like 12 hours in the middle of the ...
Mohammed: Yeah, it's really stupid. Anyway, if I was there, and I think I feel for it. We decided to go this new by McDonald's. There were a bunch of people there. People from sort of low means. It just struck me at that moment that there is so much that I could do to fix these problems that are around. I mean, I saw at the time all these homeless people at McDonald's. I contrasted it with my situation, and my of course my situation was infinitely better. It was just the fact that I was there in the middle of the night seeing these people kind of opened up this perspective that I hadn't seen before. Almost this world that I didn't know existed. That made me feel like I have to dedicate my life to something that is worthwhile. That I should try to do something with my life.
Quinn: That's pretty awesome.
Brian: Yeah, that's pretty wonderful.
Quinn: Well, I think it's insane that that's the process for the interview. I'm glad that you had that moment, because clearly a lot of folks are going to benefit from it.
Quinn: Mohammed, along those lines. Who is someone in your life that has positively impacted your work in the past six months?
Mohammed: There's a fellow here at where I work. He's a professor, his name is Peter Stoger. He's just really been over the last six months even longer, I would say he's just been a really wonderful mentor. He's been somebody who's sort of ... I think I believed in the kind of science I was doing, even though it was very risky. And sort of helped support me in this enterprise. Even though when he didn't need to. It was sort of very much a bet on ... Because the work I do is quite different than the work he does. So it's very much kind of a bet on the person as opposed to anything else. I would say without that support, I wouldn't be where I am. That's who I would say.
Quinn: That's pretty rad.
Brian: Great answer. Mohammed, what do you do when you feel overwhelmed?
Quinn: Which everybody talks about self care these days. When things get to be too much do you go on a run, play video game, just eat a bunch of ice cream. What's your ball game?
Brian: He doesn't need suggestions.
Mohammed: I think I like to ... playing video games would be a good idea. I don't actually do it nearly as much as I ought to. I was a gamer, but I can't call myself that anymore. Probably today it's more like reading, or just listening to a book on audible. Just reading a book. Just think about something else. So strangely, I find that when I'm engaged with someone else's problem very seriously that it helps me relax. I love solving problems, but my own problems they stress me out. So when I'm somehow focused on somebody else's I get to do what I like without having to have that stress. That's [inaudible 01:00:54] reading the books. Two things I guess, reading books and solving someone else's problem.
Quinn: I love that. My wife and I are both writers and she always talks about how her favorite thing to do is to work on other people's shit, because it means she doesn't have to do her shit.
Brian: Love it. Speaking of books. You mentioned books. If you could Amazon Prime one book to Donald Trump, what book would it be?
Quinn: Anything. We've had everything from clean books to the Constitution.
Mohammed: I don't know if he'll benefit from ... I would say maybe Better Angels of our Nature. Just because I think it's such a rational take on things. A rationalist take on things. I don't know if it'll make a dent.
Quinn: Why do you feel that way?
Brian: What do you mean?
Mohammed: I think it's a book that a lot of people should read, because even if not all the facts are right. I think it's a good way of thinking. It's a good way to sort of train your mind. I think that's a good thing to have. That's by Steve Pinker, by the way.
Quinn: Yeah, that's a good one. Hey, last question. Anything else you would like to say. Any way you want to use this podcast to speak truth to power last things to our listeners?
Mohammed: Yeah, look I think the future can be very bright. I don't want to go out on a limb and say it will be bright. I don't know that that's true. I do think we are living in sort of very transformative times. I think the challenges are great and sometimes it feels overwhelming. I do think that if we do this right, we stand together quite a lot. Like I started out in this discussion, I feel like we really can ... We have this preface where we can redefine what it means to be humans, and who we are, and what our future is as a species. That's incredibly exciting. It's an opportunity that I think few other humans have had. It's not our existence. It's optimistic of the opportunity at least. I hope that we ...
Speaking of Better Angels. I would hope that our Better Angels will come to the forward and that. We will see our way through this time.
Quinn: I love it. Hopefully it's a message we can print on billboards everywhere.
Brian: Yeah, Mohammed, where can our listeners follow you online?
Mohammed: Probably Twitter is the best place to follow me online. I have a blog post as well. That tends to be a bit more technical, and I don't update that frequently. My twitter on moalquraishi is probably the best place to follow me.
Brian: Awesome. Very good.
Quinn: Awesome. We will put that in the show notes.
Brian: Sorry, one more question. Are you able to make a deep fake for me of Quinn telling me that he loves me and I'm the best?
Quinn: Nope. We're not doing that right now, Brian.
Brian: Got it okay.
Quinn: Nope. That's not what he's here for.
Brian: No problem.
Quinn: Great, great, great.
Brian: Just thought I'd ask.
Quinn: Mohammed ... Yep just sneak it in. Mohammed, thank you so much obviously for being here today with all your travel fun. I'm sure you're like not even unpacked. [crosstalk 01:03:52]
Mohammed: I am very jet lagged.
Quinn: God, It's the worst.
Mohammed: I'm half asleep right now.
Brian: I'm half asleep every day.
Quinn: Brian doesn't get up before 10. Thank you so much for your time, and obviously for all your doing every day and the perspective you bring to it. I think and I hope that we can all learn from it again. We've got some serious shit going on. But as you eluded to, man if we can get past some of it. There's a pretty magical future out there, and hopefully it's a lot more equitable and just for more folks out there as well.
Mohammed: I think so and I hope so. Thank you very much for having me. I enjoyed this conversation as well.
Quinn: Thank you so much, and hopefully we will talk to you again soon once you fix and solve the whole thing. No big deal.
Mohammed: Sounds good.
Quinn: Thanks to our incredible guest today. And thanks to all of you for tuning in. We hope this episode has made commute, or awesome work out, or dish washing, or fucking dog walking late at night that much more pleasant. As a reminder, please subscribe to our free email newsletter at importantnotimportant.com. It is all the news most vital to our survival as a species.
Brian: And you can follow us all over the internet. You can find us on Twitter at importantnotimp. Just it's so weird. Also on Facebook and Instagram at importantnotimportant, Pinterest and Tumblr, the same thing. Check us out. Follow us. Share us. Like us. You know the deal. Please subscribe to our show wherever you listen to things like this. If you're really fucking awesome, rate us on apple podcast. Keep the lights on, thanks.
Brian: And you can find the show notes from today right in your little podcast player and at our website importantnotimportant.com.
Quinn: Thanks to the very awesome Tim Blaine for our jamming music, to all of you for listening, and finally most importantly to our moms for making us. Have a great day.
Brian: Thanks guys.