Latest / Paths of Curiosity in an AI World / Q2 - Power, Law, and Civic Life: When Algorithms Govern Your Life with Maura Grossman, JD, PhD

Transcript

Q2 - Power, Law, and Civic Life: When Algorithms Govern Your Life with Maura Grossman, JD, PhD · Paths of Curiosity in an AI World

• 0:04
  SPEAKER_00: Welcome to Passive Curiosity in an AI world where we explore tomorrow's careers through the lens of AI.
• 0:13
  SPEAKER_00: What happens when the star witness in a trial is a video that looks perfectly real, but never actually happened.
• 0:22
  SPEAKER_00: Dr.
• 0:22
  SPEAKER_00: Mora Grossman is building the tool that might be the last line of defense between a deep fake and a wrongful conviction.
• 0:30
  SPEAKER_00: Let's get curious.
• 0:37
  SPEAKER_00: Dr.
• 0:37
  SPEAKER_00: Mora R.
• 0:38
  SPEAKER_00: Grossman is a research professor in the School of Computer Science at the University of Waterloo and an adjunct professor at Osgood Hall Law School.
• 0:49
  SPEAKER_00: She is a pioneer in evaluating how machine learning can be used in law and healthcare, and she is famously known for serving as a special master in high-profile court cases, where she oversees how technology is used to find evidence.
• 1:05
  SPEAKER_00: In addition to her legal training, she holds a PhD in psychology, giving her a unique perspective on how human judgment and computer algorithms can either work together or fail each other.
• 1:19
  SPEAKER_00: Dr.
• 1:19
  SPEAKER_00: Mura, thank you so much for joining us.
• 1:23
  SPEAKER_01: I'm delighted to be here.
• 1:26
  SPEAKER_00: How would you describe deepfakes and your career to your younger self?
• 1:32
  SPEAKER_01: First, I wouldn't have believed you if you had said you'll be a computer science professor because I hated math, was never very good at math or technical subjects.
• 1:44
  SPEAKER_01: And so that would have been a non-starter in the first place.
• 1:48
  SPEAKER_01: I certainly wouldn't have known what a deepfake was because deepfakes didn't come into being until about 2017.
• 1:56
  SPEAKER_01: My career has been nonlinear.
• 2:00
  SPEAKER_01: I have followed my interests rather than a specific path.
• 2:06
  SPEAKER_01: So I never could predict the next step from the last step because I would fall over some rock.
• 2:14
  SPEAKER_01: I would look under the rock, there would be something very interesting, cute little buggies or something, and then I would do that for a while until it ceased to be fun, and then I'd go looking again.
• 2:26
  SPEAKER_01: So none of this would have been predictable.
• 2:31
  SPEAKER_00: What's one thing about having a dual career in law and computer science that surprised you?
• 2:38
  SPEAKER_00: Either harder or more rewarding than you expected.
• 2:43
  SPEAKER_01: So I actually don't have a degree in computer science.
• 2:46
  SPEAKER_01: I have a degree in PhD in psychology and a JD in law.
• 2:53
  SPEAKER_01: I fell into computer science and everything I know about computer science, I'm pretty much self-taught.
• 3:01
  SPEAKER_01: What I've found is people often want to stay generalists because they're afraid of closing doors for themselves.
• 3:12
  SPEAKER_01: So if I choose this path, that means I'm not choosing that path.
• 3:18
  SPEAKER_01: And I would say in about 2006, 2007, somewhere around there, a mentor I had at the time said specialization is better than being a generalist because there's always a better generalist than you.
• 3:37
  SPEAKER_01: There's somebody who will write better than you, or who speaks better than you, or whatever it is.
• 3:42
  SPEAKER_01: But you can become the top person in the world in a particular niche that you're good at.
• 3:49
  SPEAKER_01: So that surprised me when I moved from being a general litigator into being highly specialized in this area of tech and getting even more specialized over time.
• 4:07
  SPEAKER_00: I think that's so important for students like me to hear.
• 4:11
  SPEAKER_00: As we mentioned, you were a clinical psychologist before you ever went to law school, then spent 17 years at one of the most elite law firms in the world before moving into computer science research and academia.
• 4:25
  SPEAKER_00: Can you walk us from where you started as a student all the way to where you are now?
• 4:33
  SPEAKER_01: Sure.
• 4:34
  SPEAKER_01: When I was at Brown University, I had no idea what I wanted to be when I grew up.
• 4:40
  SPEAKER_01: I had some talents in art, but I was from a family that didn't encourage me to pursue art as a career as opposed to as a hobby.
• 4:52
  SPEAKER_01: So my parents wanted doctors and lawyers and that kind of thing.
• 4:58
  SPEAKER_01: So I became a clinical psychologist.
• 5:01
  SPEAKER_01: My mother was a clinical psychologist.
• 5:03
  SPEAKER_01: What she did seemed to be very interesting.
• 5:05
  SPEAKER_01: My father was in the insurance business.
• 5:08
  SPEAKER_01: What he did seemed to me very boring.
• 5:10
  SPEAKER_01: And so I became a clinical psychologist.
• 5:14
  SPEAKER_01: And I did that for close to a decade in the United States.
• 5:19
  SPEAKER_01: And it was good to a point.
• 5:24
  SPEAKER_01: My personality, in order to be a really good clinical psychologist, you have to be a really good listener.
• 5:31
  SPEAKER_01: And you have to be somewhat passive because you can't tell people what to do.
• 5:34
  SPEAKER_01: You have to let them tell you what their problems are, and they unfold at their own rate, and you have to have a lot of patience.
• 5:43
  SPEAKER_01: I am neither a great listener nor do I have a lot of patience.
• 5:48
  SPEAKER_01: So that didn't end up to be the best fit for me.
• 5:53
  SPEAKER_01: And I found that the things I like doing the most as a psychologist weren't really psychology things.
• 6:00
  SPEAKER_01: They were more administrative and legal in feel, like hiring people and stuff like that.
• 6:09
  SPEAKER_01: So I did a little research and I decided I would go try law school.
• 6:15
  SPEAKER_01: What's the worst that could happen if it didn't work out?
• 6:18
  SPEAKER_01: I wasted some money and I'd go back to being a psychologist.
• 6:21
  SPEAKER_01: So in my 30s, I went back to law school full-time, lived in the dormitory with all the people who are much younger than me.
• 6:30
  SPEAKER_01: And I thought maybe I would do health law first, given my background in health and mental health.
• 6:37
  SPEAKER_01: But I didn't particularly find that interesting.
• 6:39
  SPEAKER_01: I liked litigation, which is lawsuits, is very much a roller coaster.
• 6:45
  SPEAKER_01: You never know what to predict.
• 6:47
  SPEAKER_01: And so I decided to do that, and I got into a really good firm in New York.
• 6:54
  SPEAKER_01: And I started to do that work, and I found it interesting.
• 7:01
  SPEAKER_01: It was lucrative, but again, it still didn't meet my personality very well because I'm not really a warrior.
• 7:10
  SPEAKER_01: I'm much more of a sort of negotiator, and fairness is really important to me.
• 7:18
  SPEAKER_01: And litigators, you're supposed to beat the crap out of the other side, and you don't really worry about whether it's fair or not fair.
• 7:26
  SPEAKER_01: You're supposed to win for your client.
• 7:28
  SPEAKER_01: And it wasn't a perfect fit either.
• 7:33
  SPEAKER_01: And I started to do some research because we had a problem at the firm where we had relatively small number of lawyers, but we had cases that have millions of emails in them.
• 7:46
  SPEAKER_01: And how do you find the hundred emails that matter for your case?
• 7:50
  SPEAKER_01: So I started to look into computer science and some of the AI methods that were available at the time.
• 7:59
  SPEAKER_01: And I moved into that area, creating a specialty, as it were, in the in technology law, technology assisted review, finding evidence, digital evidence, that kind of stuff.
• 8:11
  SPEAKER_01: And I enjoyed that.
• 8:13
  SPEAKER_01: I did that for a while.
• 8:14
  SPEAKER_01: And one of the things I really liked doing is in the United States, there is something called a special master.
• 8:24
  SPEAKER_01: And that is when the court has a dispute that is very technical in nature.
• 8:30
  SPEAKER_01: They will sometimes call in a neutral third person to help the court decide those technical issues that they don't fully understand.
• 8:42
  SPEAKER_01: And I really, really like doing that work.
• 8:44
  SPEAKER_01: And probably if I had to do my career over again from scratch, knowing what I know now, I would have become a judge because it's similar to being a psychologist, similar to being a lawyer, it combines both.
• 8:56
  SPEAKER_01: But I didn't know that then.
• 8:58
  SPEAKER_01: So I started doing this special masterwork and really liked it.
• 9:02
  SPEAKER_01: My firm was not as eager for me to do that kind of work because you in that role you're choosing winners and losers, and sometimes clients don't like that.
• 9:16
  SPEAKER_01: The firm's clients don't like that.
• 9:18
  SPEAKER_01: So it created some business sort of friction.
• 9:22
  SPEAKER_01: And I also met a Canadian and we decided to get married.
• 9:27
  SPEAKER_01: He hated New York.
• 9:28
  SPEAKER_01: I really liked New York.
• 9:30
  SPEAKER_01: I decided I'll leave my law firm.
• 9:34
  SPEAKER_01: What's the worst that can happen?
• 9:35
  SPEAKER_01: I can come back and work at another law firm.
• 9:38
  SPEAKER_01: And so I came up to the University of Waterloo and where he was teaching, and they gave me a position, but they had no idea what they were going to do with a lawyer in the computer science department.
• 9:49
  SPEAKER_01: And I looked around and I saw that all of the students were just building stuff, but they weren't spending any time thinking about should I be building this?
• 9:59
  SPEAKER_01: Who gets hurt by this thing I'm building?
• 10:02
  SPEAKER_01: Does it discriminate?
• 10:04
  SPEAKER_01: Does it protect privacy?
• 10:05
  SPEAKER_01: So I started to teach courses on law, ethics, and policy to computer science students.
• 10:12
  SPEAKER_01: And then I reached out to a law school in Toronto and said, Do you want your students to be more technically proficient?
• 10:19
  SPEAKER_01: Because I can teach them about technology.
• 10:22
  SPEAKER_01: And so I do that.
• 10:24
  SPEAKER_01: I do some private work.
• 10:25
  SPEAKER_01: I do lots of lots of stuff.
• 10:27
  SPEAKER_01: I do a lot of public speaking.
• 10:29
  SPEAKER_01: I work, I educate judges on technology.
• 10:33
  SPEAKER_01: So I have a very, very varied career at this point doing almost all things that I find fun and interesting.
• 10:44
  SPEAKER_00: Before AI generated fakes, what was the biggest evidence manipulation problem in court?
• 10:50
  SPEAKER_00: And what changed when deep fake technology arrived?
• 10:57
  SPEAKER_01: So before deep fakes, we always had frauds.
• 11:01
  SPEAKER_01: Somebody would uh make a fake will and say, or a fake letter of acceptance or something like that.
• 11:11
  SPEAKER_01: But you needed some talent to do it because you had to get over on somebody, so your signature actually had to look like the signature you were forging, the document had to look real.
• 11:24
  SPEAKER_01: Once we got generative AI, it became very easy to make this fake evidence on a computer for free in under five minutes.
• 11:36
  SPEAKER_01: You could do it, I could do it, anybody could do it.
• 11:39
  SPEAKER_01: So we have now this problem of scale, of scope, of how quickly you can do it.
• 11:48
  SPEAKER_01: I could you have enough of my voice now to make a very good clone of my voice.
• 11:53
  SPEAKER_01: You can go right on to Speechify and do it for a couple of bucks.
• 11:58
  SPEAKER_01: And so the AI generated tools really moved the ability to create realistic looking media, moved it much further along than it was when I was growing up.
• 12:16
  SPEAKER_00: You are building a deep fake detector that courts will actually use.
• 12:21
  SPEAKER_00: What's the nightmare scenario you are trying to prevent?
• 12:28
  SPEAKER_01: This is a really tall challenge, and whether we'll actually succeed has a question mark.
• 12:36
  SPEAKER_01: I'll give you a couple scenarios.
• 12:38
  SPEAKER_01: Somebody goes into court, family court, and they say, My husband's abusing me and the kids.
• 12:49
  SPEAKER_01: I have an audio tape of him on my phone, threatening to lock us in the trunk and to drive the car into the lake.
• 12:58
  SPEAKER_01: I want you to give me sole custody of the kids and a protective order that he can't come within a thousand feet of any of us.
• 13:07
  SPEAKER_01: Here, Judge, let me play you the tape.
• 13:11
  SPEAKER_01: Now, how is that judge supposed to have any idea whether that tape is real or not?
• 13:18
  SPEAKER_01: Because the technology today is so good that it's almost impossible to tell if that's real or not.
• 13:27
  SPEAKER_01: So we want to build tools to help courts and people who represent themselves deal with this kind of problem.
• 13:38
  SPEAKER_01: The challenge is as soon as you develop a tool, somebody figures out how to work around it, and then you have to develop a new tool, and then they work around that, and you have to develop a new tool.
• 13:49
  SPEAKER_01: So it's not a develop it once and be done.
• 13:53
  SPEAKER_01: It's how do you develop something that will continue to work?
• 13:57
  SPEAKER_01: And one of the things that we are going to do as part of our project is after we build our tool, we're going to have a hackathon and we're going to invite all kinds of students to come and break it.
• 14:10
  SPEAKER_01: So then we'll fix it and make it a little bit better.
• 14:13
  SPEAKER_01: And we may do this a couple of times to get a tool that's more robust uh uh than uh uh not so brittle.
• 14:25
  SPEAKER_00: Students watching this may see deepfakes on TikTok and other platforms all the time, and the videos are often funny and harmless, but you're focused on civil and criminal trials.
• 14:38
  SPEAKER_00: If someone were to deep fake me or a student listening right now into a crime we did not commit, how would we even begin to prove it was false?
• 14:49
  SPEAKER_01: So there's two ways you could do this.
• 14:55
  SPEAKER_01: The first way is if my video of you, the what we call metadata, all digital media has fingerprints that say when it was created, where it was created, who created it.
• 15:13
  SPEAKER_01: For pictures, there'll be longitude and latitude.
• 15:16
  SPEAKER_01: So I can look at that metadata, and if you were on an airplane at the time and your airplane didn't have Wi-Fi service, something's not right there, right?
• 15:29
  SPEAKER_01: So I can look for corroborating evidence or or evidence that contradicts this video.
• 15:36
  SPEAKER_01: That's one thing I can do.
• 15:38
  SPEAKER_01: I can also look for signs.
• 15:45
  SPEAKER_01: There are subtle signs, so for example, the shadow on your glasses may not be in the right place given where the light is in the room.
• 15:58
  SPEAKER_01: Or like when we all looked at the Princess Kate photo, uh, and we all noticed things like the sleeves didn't match and and her hair wasn't the same on both sides.
• 16:10
  SPEAKER_01: And so we can look for cues in the content that would suggest that these tools, while they're very good, they don't get physics perfectly.
• 16:23
  SPEAKER_01: First, it used to be six fingers.
• 16:25
  SPEAKER_01: Now they get the fingers right, but they'll still screw up other more subtle stuff.
• 16:30
  SPEAKER_01: And then we can also start looking at this metadata and see where does it say this was taken, when does it say this was taken, does it suggest uh that it was manipulated in any way?
• 16:44
  SPEAKER_01: And and normally you need an expert for this.
• 16:46
  SPEAKER_01: You would probably have to hire an expert, many students who were accused during the pandemic of cheating by these AI tools that were used to proctor really had a very hard time proving that they were innocent.
• 17:06
  SPEAKER_00: So, how exactly does your deep fake detector work?
• 17:10
  SPEAKER_00: I know you mentioned some ways you can try to figure out if those videos are fake or not.
• 17:15
  SPEAKER_00: We've mentioned how these technologies just keep getting better and better.
• 17:20
  SPEAKER_00: So we haven't built it yet.
• 17:22
  SPEAKER_01: We are in very early stages.
• 17:25
  SPEAKER_01: So we're just getting permission from the university for the legal research team to interview judges and other court personnel to find out what they would want in a tool.
• 17:41
  SPEAKER_01: How transparent does it have to be?
• 17:43
  SPEAKER_01: How easy or hard to work?
• 17:46
  SPEAKER_01: Is there a certain percent they have in their head of accuracy that they want?
• 17:51
  SPEAKER_01: So we're getting we filed for permission to do these kind of user interviews.
• 17:58
  SPEAKER_01: And the technical team is working on creating a data set.
• 18:03
  SPEAKER_01: In order to build such a tool, you need a really good data set that has both real and fake stuff.
• 18:12
  SPEAKER_01: And the fake has to be really good fake stuff because if you could already tell it was fake without the tool, what's the point of the tool?
• 18:20
  SPEAKER_01: And believe it or not, finding those kinds of data sets is really, really, really hard.
• 18:28
  SPEAKER_01: They're that are public.
• 18:30
  SPEAKER_01: Of course, if you wanted to give me your email, that would be fine, and then I'd have to build a set of fake email, but nobody wants to give me their email to use real email or something.
• 18:40
  SPEAKER_01: So you have to work really hard to find data sets.
• 18:44
  SPEAKER_01: We are using some celebrity images and things like that.
• 18:49
  SPEAKER_01: And once we build our set, we are going to train this system on examples of this is real and this is fake, and we will teach it to learn the difference between real and fake.
• 19:03
  SPEAKER_01: That's usually called supervised machine learning.
• 19:06
  SPEAKER_01: And the more you train it, the better it'll get, and then you try it out, and then we're gonna have people try to break it and trick it, and you improve the tool, and that's the process.
• 19:18
  SPEAKER_01: But we're just at the very, very beginnings of design work, and what's a little bit different here is we don't just want our computer scientists to design this without talking to us.
• 19:31
  SPEAKER_01: We want the lawyers and the computer scientists working together because it's really important for lawyers that we know how this thing works, that it be transparent.
• 19:42
  SPEAKER_01: So that's a challenge, is we have to build something that will tell us this is how I know this is fake.
• 19:48
  SPEAKER_01: Not just say this is fake, but this is how I know this is fake, because otherwise it won't work in the justice system.
• 19:56
  SPEAKER_00: You train judges worldwide on AI evidence.
• 19:59
  SPEAKER_00: What's the question judges ask you most often?
• 20:02
  SPEAKER_00: And what might this tell us about how prepared or unprepared the legal system is for this?
• 20:10
  SPEAKER_01: One of the questions I get the most right now is less about the deepfakes, but the judges want to know how they can use generative AI themselves safely.
• 20:23
  SPEAKER_01: And can they use it safely?
• 20:26
  SPEAKER_01: Do they have to worry about hallucinations?
• 20:28
  SPEAKER_01: What can the tools do?
• 20:30
  SPEAKER_01: What can't they do?
• 20:32
  SPEAKER_01: So I get lots of questions on that.
• 20:35
  SPEAKER_01: I do get some questions on how do we approach this AI evidence?
• 20:42
  SPEAKER_01: What are the questions we should be asking?
• 20:45
  SPEAKER_01: Do you have the original copy of the photo?
• 20:48
  SPEAKER_01: Do you have the camera on which the photo was taken?
• 20:53
  SPEAKER_01: We would like, would you give be willing to give that to an expert to take a look at that?
• 20:58
  SPEAKER_01: And so teaching them the kinds of questions they need to be asking to get smart about this stuff.
• 21:05
  SPEAKER_01: So I think our federal courts are offering training.
• 21:12
  SPEAKER_01: In the United States, our federal courts are offering training.
• 21:15
  SPEAKER_01: Our state courts are a little further behind.
• 21:18
  SPEAKER_01: And in Canada, we're just starting with educational programs, both in the federal and the provincial courts.
• 21:28
  SPEAKER_00: What's the biggest misconception students might have about working at the intersection of law and technology?
• 21:37
  SPEAKER_01: I thought you'd have to be really, really good at math, but I've been able to do what I do without all that strong a math background.
• 21:48
  SPEAKER_01: So even if math scares you, I wouldn't run off and say this isn't a field I I couldn't go into because I think, yeah, you've got to learn something about statistics.
• 22:00
  SPEAKER_01: Statistics, but a lot of it is common sense, a lot of it is just reading, like puzzles and things like that.
• 22:08
  SPEAKER_01: So you don't have to be a math maven to be able to do legal tech.
• 22:12
  SPEAKER_00: Why is explainability so critical when we're talking about someone's freedom being at stake?
• 22:21
  SPEAKER_01: So black box technology is technology that we don't know how it works.
• 22:28
  SPEAKER_01: So right now we have things called deep learning algorithms that you can show it a picture and it can tell you what it's a picture of, but I couldn't explain to you how it gets to the answer.
• 22:41
  SPEAKER_01: No clue.
• 22:42
  SPEAKER_01: We know that there are many layers of algorithms, and these algorithms are sharing information in some way, but that's about all we know.
• 22:49
  SPEAKER_01: Glassbox attempts to build systems where we that are transparent enough to understand the mechanisms and how they're working.
• 23:00
  SPEAKER_01: Sometimes you have to train trade effectiveness for transparency.
• 23:07
  SPEAKER_01: So in order to have get more transparency, I'm gonna have a less sophisticated algorithm, a less sophisticated AI.
• 23:18
  SPEAKER_01: Now, sometimes, like maybe for medicine, if it's diagnosing brain tumors, maybe I want it to be really, really, really, really accurate.
• 23:30
  SPEAKER_01: And I I'm a little less concerned about how it works.
• 23:33
  SPEAKER_01: Because we all take medicines, we don't exactly know how they work in our body, but somebody's tested them and told us this is a safe medicine.
• 23:41
  SPEAKER_01: In law, the values are a little bit different.
• 23:45
  SPEAKER_01: We don't try or aspire not to take away somebody's freedom or life using a tool that nobody can explain how it got its answer.
• 23:59
  SPEAKER_01: Explainability in the justice system, transparency, is more important.
• 24:05
  SPEAKER_01: You don't care how Spotify chose your next song, or if Netflix screws up on a movie, it you watch the first five minutes, you don't like it, it's not a big deal.
• 24:19
  SPEAKER_01: Whole different ballgame when we're in the legal system and you can be found guilty or liable or something like that.
• 24:27
  SPEAKER_01: Then understanding how these tools work becomes very, very important.
• 24:34
  SPEAKER_00: Looking 10 to 20 years in the future, will juries even be able to trust video evidence anymore?
• 24:42
  SPEAKER_01: I don't even think we're talking 10 years.
• 24:44
  SPEAKER_01: I think we're talking in the next year or two or three that you will see videos purportedly of the war in Iran or wherever the latest war is, and you will not know whether they're real or not.
• 25:05
  SPEAKER_01: So I I think we're already moving into that in that world, and it's a real problem because jurors make their decisions based on seeing and hearing evidence.
• 25:20
  SPEAKER_01: And if they can't use their senses anymore to do that, then how are they supposed to make these decisions?
• 25:29
  unknown: It's gonna change that quickly.
• 25:31
  SPEAKER_01: Yes, we it we're already almost there.
• 25:35
  SPEAKER_01: If I showed you when I give talks, I do a test called real mora, fake mora, and I show my audiences ten pictures, five real, five fake, and nobody does better than chance most of the time.
• 25:50
  SPEAKER_01: 50-50.
• 25:53
  SPEAKER_00: And in that future courtroom, what will humans be doing that the AI cannot?
• 25:58
  SPEAKER_00: And what skills do you think will still matter?
• 26:02
  SPEAKER_01: The first question is: will we still have a courtroom with judges and juries?
• 26:10
  SPEAKER_01: Because maybe we just submit the dispute to an algorithm that can look at all the cases from the last 50 years and make a prediction.
• 26:20
  SPEAKER_01: And we don't even need to have the judge or jury, it just decides one way or the other.
• 26:26
  SPEAKER_01: And there is experiments with AI judges and things like that in low value cases.
• 26:35
  SPEAKER_01: So I bought a pair of earrings online a year or two ago, and maybe I spent$12 or$13.
• 26:45
  SPEAKER_01: And one of them showed up, it was very bent, the post was bent, couldn't get it through my ear.
• 26:51
  SPEAKER_01: And so I asked the seller, uh, I really love the earrings.
• 26:57
  SPEAKER_01: Can you send me, just replace the one that's broken, the left one?
• 27:02
  SPEAKER_01: And he basically said, there's a language barrier here.
• 27:07
  SPEAKER_01: If you weren't such a moron, you'd just fix it.
• 27:10
  SPEAKER_01: And so I took it to my husband, and he looked at it and he said, No, this isn't really fixable because if I try to bend it, it's gonna snap right off.
• 27:19
  SPEAKER_01: So I'm having this argument with this vendor, and he says, send me the earrings back to China.
• 27:27
  SPEAKER_01: So I go to find out how much that'll cost, and that's$30 for a pair of earrings that cost$12.
• 27:35
  SPEAKER_01: So I decided to do the online dispute resolution system.
• 27:40
  SPEAKER_01: So I type into the computer my case and I take photos of the earrings and everything, and I argue my case to the AI, and I won.
• 27:51
  SPEAKER_01: But I won$2.38 because the algorithm decided I still had one good earring, even though it wasn't matched anymore.
• 28:02
  SPEAKER_01: I still had one good earring, so I got some value.
• 28:05
  SPEAKER_01: And he did spend money on postage and stuff, so they thought the broken earring was probably worth about$2.38.
• 28:15
  SPEAKER_01: Now, you may say, that's silly, and that's not fair, and that's not how a real judge would have decided.
• 28:21
  SPEAKER_01: But let's take the same example and say you're in a dispute with your landlord.
• 28:29
  SPEAKER_01: He's not giving you your$800 deposit back.
• 28:36
  SPEAKER_01: If you take a day off from work, you may get fired or you may lose more than that because you don't really have a non-sick day to take off.
• 28:47
  SPEAKER_01: Maybe you don't speak English well, maybe you're an immigrant and you're afraid if you go to court, ice may scoop you up because they've been going to courts and scooping people up.
• 29:00
  SPEAKER_01: Maybe you can't afford childcare or the transportation to get to the court.
• 29:05
  SPEAKER_01: So maybe going online and getting$238 of your$800 back is still a better solution than doing nothing.
• 29:16
  SPEAKER_01: And that's what some people have argued is we're going to get to the point where going to court is so expensive and takes so long that some or many decisions can be made with algorithms.
• 29:30
  SPEAKER_01: But then you and I might say, do we want child custody decisions decided that way?
• 29:36
  SPEAKER_01: Do we want murder cases decided that way?
• 29:38
  SPEAKER_01: No, maybe there's still some kind of cases that there is a value to having a human because a human can use their empathy, a can look at the whole circumstances and understand the difference between somebody stealing to feed their children versus stealing to support a drug habit versus stealing just because they can.
• 30:02
  SPEAKER_01: And maybe those are subtleties we care about.
• 30:05
  SPEAKER_01: So there still are human values that where we would still want the courts, at least in part, to be run by humans.
• 30:15
  SPEAKER_00: What would you tell a student who's worried that the career they're excited about today, especially at the intersection of law and technology, won't be the same in less than a decade?
• 30:28
  SPEAKER_01: I'm teaching law school this semester, and my students ask me, Am I gonna get a job?
• 30:36
  SPEAKER_01: Is it gonna look the same?
• 30:38
  SPEAKER_01: Am I gonna have a career five years from now with all this generative AI?
• 30:44
  SPEAKER_01: And to be really honest with you, it's the first time that I can't really give you a good answer to that question because none of us know where this is going and how good it's gonna get and what jobs it's gonna take over, and whether there'll be new jobs.
• 31:04
  SPEAKER_01: I can guarantee you that things are not gonna look the same in five years or ten years than they look now.
• 31:12
  SPEAKER_01: So what I would tell somebody to do is instead of sitting down and trying to predict the future, which is very hard, or to figure out what makes money now or what's likely to make money or something like that, to figure out what am I really good at, what am I really passionate about, and what will add value somewhere, because if you add value, you'll always get paid if you add value.
• 31:47
  SPEAKER_01: And I would pursue that, but remain flexible like I have in my career.
• 31:54
  SPEAKER_01: If something else comes along that looks really interesting, don't be afraid to switch, switch directions.
• 32:02
  SPEAKER_01: I think that will in my generation, people decided what they wanted to be, and they went and did that, and they did that for their whole career.
• 32:12
  SPEAKER_01: Somebody like me who changed careers a couple of times, that's very, very unusual for somebody my age.
• 32:18
  SPEAKER_01: For somebody your age, that's gonna be very typical.
• 32:21
  SPEAKER_01: People are not gonna choose one job.
• 32:24
  SPEAKER_01: They're gonna have to be flexible and they're gonna have to roll with the punches.
• 32:28
  SPEAKER_01: But I wouldn't try to pull out a crystal ball.
• 32:31
  SPEAKER_01: I would try to figure out what am I really good at, what am I really interested in, and what do I think can add value to the world and do that and stay open, stay open and flexible to things, other things that come along.
• 32:45
  SPEAKER_00: So you're saying figure out what you're good at and then stay adaptable throughout your entire life to figure out where you can add value and where you can get paid.
• 32:55
  SPEAKER_01: Right, it never makes sense to work in a field that where you're not good at it.
• 33:01
  SPEAKER_01: Because it's always easier to become better at something you're good at than to shore up things that you're not good at.
• 33:09
  SPEAKER_01: And it doesn't make you feel good to do things you're not good at.
• 33:12
  SPEAKER_01: So you should do things that you're good at.
• 33:14
  SPEAKER_01: You have to be really excited and passionate and love doing them.
• 33:18
  SPEAKER_01: I gather you like doing these the podcasts.
• 33:21
  SPEAKER_01: That's good.
• 33:22
  SPEAKER_01: You do better work when you like what you're doing.
• 33:25
  SPEAKER_01: And but it has to also add value.
• 33:28
  SPEAKER_01: I mean, if you're the best hockey player, but they're the only teams they have is a baseball team, that's not gonna do very much good.
• 33:36
  SPEAKER_01: So you also have to take into account where you can add value or where there's a need for what you can do.
• 33:45
  SPEAKER_01: I think learning something about AI and not just hiding and hoping it's going to go away, because they don't think it's gonna go away, but also not giving yourself over to it so that you don't think anymore.
• 34:02
  SPEAKER_01: I see lots of my students using the AI to answer the questions and they stop learning themselves.
• 34:07
  SPEAKER_01: That you can't do, but you gotta learn enough so that you know how it can help you, where you can use it, but you still need to be able to, there will always be a place for people who are analytical, can solve problems, and who are flexible.
• 34:23
  SPEAKER_01: Most people are not flexible, most people are rigid and have a really hard time dealing with change.
• 34:30
  SPEAKER_01: People who can deal with change are gonna do very well here because we're gonna see a lot of change in the next decade or so.
• 34:40
  SPEAKER_00: Dr.
• 34:41
  SPEAKER_00: Grossman, thank you so much for joining us today on Paths of Curiosity.
• 34:45
  SPEAKER_00: This conversation really shows just how unpredictable this career is at the moment.
• 34:53
  SPEAKER_01: It isn't, and that's scary for me as an educator because normally I do I'm trying to guide my students and I don't know how to guide them terribly well right now.
• 35:05
  SPEAKER_00: This has been paths of curiosity in an AI world, exploring careers through the lens of AI.
• 35:12
  SPEAKER_00: Until then, stay curious.