Latest / The Road to Accountable AI / Robert Levitan (Ethical Tech Project): Unlimited Intelligence, But Where's the Wisdom?

Transcript

Robert Levitan (Ethical Tech Project): Unlimited Intelligence, But Where's the Wisdom? · The Road to Accountable AI

• 0:09
  I'm Kevin Werbeck, Professor of Legal Studies and Business Ethics at the Warden School of the University of Pennsylvania.
• 0:15
  For decades, I've studied emerging technologies from broadband to blockchain.
• 0:19
  Today, AI is promising to transform our world.
• 0:24
  But AI needs accountability, mechanisms to ensure it's developed and deployed in responsible, safe, and trustworthy ways.
• 0:32
  On this podcast, I speak with the experts leading the charge for accountable AI.
• 0:36
  oh
• 0:40
  Robert Levitan was an executive at some of the first successful internet companies during the dot-com boom and then has had a long successful career as a serial tech entrepreneur.
• 0:50
  He's now chairman of the Ethical Tech Project, where in full disclosure, I'm an advisor.
• 0:56
  We talk about his sense of the parallels between the early internet and the AI world of today, why it's critical for companies to do something to get started on AI governance.
• 1:07
  Why that's ultimately in their interest to be ethical and focus on accountable uses of AI, and why, despite it all, despite all of the risks and despite the likelihood that we will
• 1:18
  see large significant AI harms in the near future, he's still an optimist.
• 1:26
  Robert, really delighted to have you on the podcast.
• 1:28
  So happy to be here.
• 1:31
  You and I have known each other for a long time, uh, going back to when you were the CEO of a dot com startup called Flues, and my wife became your general counsel, and you've
• 1:42
  done a lot since then, other startups, other other work in interesting areas.
• 1:46
  What got you interested in AI governance?
• 1:50
  Well, for those of us who were around in the nineteen nineties, and uh actually my first internet company was iVillage, and we know each other back then from nineteen ninety-five.
• 2:00
  Um, there was such a great spirit in the early days of the internet.
• 2:04
  We were all optimistic and we we thought we were changing the world for the better, and w we just wanted to connect the world and share information, and we really thought we we
• 2:14
  knew something, right?
• 2:15
  We knew a secret.
• 2:16
  Well, we learned a lot of lessons, didn't we?
• 2:20
  And over the last 30 some odd years, what we learned is there are always unintended consequences.
• 2:25
  New technologies always have unintended consequences.
• 2:28
  And what we thought we were doing, building online communities, turned out a little different than we expected.
• 2:34
  So I found myself about a year ago going down the AI rabbit hole, but not down the AI rabbit hole of I've got to start another company, I've got to raise venture capital, I've
• 2:44
  got to do this.
• 2:46
  Much more, wait a second.
• 2:48
  This moment in some ways feels like the early days of the internet.
• 2:51
  And in other ways, it's very different.
• 2:54
  And what lessons did we learn?
• 2:55
  So I started writing on LinkedIn about lessons learned and how we should be approaching this moment, which is obviously really important.
• 3:04
  And the next thing I know, I got a phone call from the Ethical Tech Project.
• 3:08
  And they asked me if I wanted to get involved.
• 3:10
  So that's how I arrived at at this place.
• 3:14
  And I'm so excited.
• 3:15
  I can't think.
• 3:17
  of anything more important than being on the road to accountable AI or teaching ethical tech, responsible AI, whatever you want to call it right now, and people like you and I to
• 3:27
  be sharing whatever insights we have.
• 3:31
  Absolutely.
• 3:31
  What are some of the parallels that you see to that experience back twenty or thirty years ago?
• 3:36
  Well
• 3:38
  You know, I do think that when I talk about unintended consequences, of course, you know, one of the great moments that we thought happened in the early days of the internet was
• 3:48
  like section two thirty.
• 3:50
  You know, I was running online community or involved with online community iVills.
• 3:54
  We thought two thirty was the answer because, you know, we couldn't monitor everything and we had some rules, but this well, it turns out this is a prime example of something well
• 4:04
  intended.
• 4:06
  Probably meant to be revisited, never revisited, now interpreted in a way that people today think, well, that means you can say whatever you want on the internet and it doesn't
• 4:15
  matter.
• 4:16
  That wasn't the intent of Section 230.
• 4:18
  You know that, I know that.
• 4:20
  And yet I hear people say that.
• 4:22
  So the big lesson is that one, there are always going to be unintended consequences.
• 4:28
  Two, you need to think a little more carefully now.
• 4:32
  Instead of move fast and break things.
• 4:36
  I think we all know the internet, you can move fast and break things.
• 4:39
  It was just a browser.
• 4:41
  Nothing bad was going to happen.
• 4:42
  Now we've got the internet on top of mobile, on top of AI.
• 4:46
  And if you think fast and break things, you're going to really break things and it's going to have a really bad effect.
• 4:52
  So we need to think a little more before we deploy technology.
• 4:58
  And every company has a responsibility to be thinking about what they're building, what could go wrong, and how to create.
• 5:06
  Whatever guardrails can keep you accountable.
• 5:10
  And the other thing I'll say: trust has always been important.
• 5:14
  Trust is more important now than ever.
• 5:15
  So why should a company do this?
• 5:18
  Well, number one, we're not going to have a lot of government regulation around responsible, accountable, ethical use of technology.
• 5:27
  But if you lose trust, it's going to really hurt your business.
• 5:31
  And with AI, you're likely to lose trust.
• 5:34
  So every company should.
• 5:36
  think about how to put in guardrails, think about what could go wrong, and come up with their own version of a responsible or accountable AI plan because it's a competitive
• 5:48
  advantage.
• 5:50
  And and we've seen that in the marketplace already.
• 5:55
  So as you described, you've joined as now the the chairman of Ethical Tech Project.
• 5:59
  let's let's get some context in there.
• 6:01
  Tell us about what what that is and you know how it's evolved from where it started.
• 6:06
  Sure.
• 6:06
  So the Ethical Tech Project was actually started about six years ago by two tech entrepreneurs, Tom Chavez, Vivek Vidya.
• 6:15
  And Tom and Vivek actually started two companies, uh, one called um Rapt and one called Crux.
• 6:23
  And these were actually data marketing companies.
• 6:25
  One was sold to Microsoft and one was sold to Salesforce.
• 6:29
  And
• 6:30
  They actually, Vivek actually went on to be the head of the Salesforce Marketing Cloud business, the engineering head of that.
• 6:37
  They know data.
• 6:39
  And about six years ago, they were on the leading edge of data privacy.
• 6:45
  Yes, it's really important.
• 6:47
  And if you're gonna have an ethical company and you're gonna be in marketing, you better make sure you've got the right data engineering and data privacy policies.
• 6:55
  So they were way ahead of this.
• 6:56
  So they funded the start of the ethical tech project.
• 7:00
  Because they believed in the importance of trusted data.
• 7:04
  Well, guess what?
• 7:05
  What's at the core of AI?
• 7:07
  Data.
• 7:09
  Bad data, bad AI.
• 7:11
  So it's a logical thing that Ethical Tech Project is about trusted data and then other deployments of that, including AI.
• 7:23
  So that's the history of the company.
• 7:25
  Um we like to call ourselves a think and do tank.
• 7:29
  Yeah.
• 7:29
  Why?
• 7:30
  Yeah, you like that, right?
• 7:31
  I mean, you and I are doers.
• 7:32
  We're not just thinkers.
• 7:34
  Um, you're probably more of a thinker, and I'm more of a doer.
• 7:38
  I'm an entrepreneur, you're an academic, but you yes.
• 7:41
  But think and do.
• 7:42
  Why?
• 7:43
  There are a lot of people thinking about AI policy.
• 7:45
  There's a lot of people thinking about all kinds of AI stuff, but just talking about it isn't enough.
• 7:53
  You know, we believe that we can think about these things, but then we have to teach them to people actually building AI.
• 7:59
  So what the Ethical Tech Project does primarily is we try to work with people who are deploying AI and help them.
• 8:10
  figure out how they can deploy AI.
• 8:12
  I was going to say help them think about, help them do, but no, we're not technical.
• 8:16
  We don't go in and do the tech for them.
• 8:18
  And we're not just teaching them to think.
• 8:19
  We're saying, okay, here's a real life scenario.
• 8:22
  What do you do?
• 8:22
  You know, you've got an option here of using the data in this way or you can use the data that way.
• 8:27
  What do you do?
• 8:28
  What are the choices?
• 8:29
  What are the implications?
• 8:30
  You know, what's the trade off financially, ethically?
• 8:34
  So we teach this to um
• 8:38
  two different distinct groups of people.
• 8:39
  One is engineers, product managers, business leaders who are in their uh early to mid career stage.
• 8:47
  Um and we teach what we call the Ethical Tech Project Fellowship.
• 8:52
  So here we'll take uh twenty people out of about two to three hundred applicants and we'll spend ten weeks with them.
• 8:59
  Every Thursday night for two hours for ten weeks and we've had three co cohorts of this group.
• 9:04
  And I'm talking about executives from
• 9:06
  Google Deep Mind, Bank of America, Roku, Amazon, Google, other products, you know, rope, but whatever.
• 9:17
  It it's it's it's Spotify.
• 9:19
  I mean, we've had some really great young people who have self-identified.
• 9:23
  They've applied and said, I care about ethics.
• 9:27
  And it is amazing to see, I mean, they're deep inside some of these companies.
• 9:32
  And they are thinking about ethics.
• 9:33
  So we're trying to help them translate that because we believe that's where change happens.
• 9:37
  The other thing we do is we work directly with companies.
• 9:39
  And this I think is really important.
• 9:42
  We get it.
• 9:42
  It's competitive out there.
• 9:44
  I'm an innovator.
• 9:45
  I'm an entrepreneur.
• 9:45
  I know what it's like.
• 9:46
  I know it's like to raise venture capital.
• 9:48
  I know it's like to be on the flywheel and every six months you better make progress.
• 9:51
  And you you know.
• 9:53
  But
• 9:56
  How do you in that environment think about being accountable with your AI deployment?
• 10:01
  And the answer is, well, you can work with us.
• 10:03
  We'll help you think about it.
• 10:04
  We'll be your partner.
• 10:05
  We'll bring best practices to you.
• 10:07
  We'll bring experts to you.
• 10:10
  we'll help educate your your team, and we'll even help you develop an AI governance plan.
• 10:17
  So to me, this is kind of our golden moment.
• 10:21
  As the Ethical Tech Project to go into a company and say, okay, you need a governance plan and you need not something off the shelf.
• 10:27
  It has to be tailored just to you.
• 10:29
  How are you going to do that?
• 10:30
  Well, you're going to think about what are your core principles?
• 10:34
  What's your mission statement?
• 10:36
  How do we translate those into core principles of an AI governance plan?
• 10:40
  And then how do we build that plan?
• 10:42
  And then how do we implement it?
• 10:43
  So we will help companies do that.
• 10:45
  Now we're not the monitoring tools.
• 10:47
  We're not any of the technology behind.
• 10:50
  tracking your AI deployments, but we are a great thought partner to help companies do that in a way that is very specific to them and aligned with their core values.
• 11:01
  And we think that's the first step for every company having an accountable AI program.
• 11:07
  Mm-hmm.
• 11:09
  What is your sense at this point about uh how much companies are at that stage where they get it, they realize this is an issue, they need to do something about needing to invest,
• 11:19
  but they don't know how.
• 11:21
  Uh so I think every company is thinking about this.
• 11:25
  And when you read recently about you know open AI and anthropic, and everybody's got models that are breaking out.
• 11:31
  Every company knows what can happen.
• 11:34
  But they they're almost paralyzed by how fast everything's changing.
• 11:39
  And I don't know.
• 11:40
  And should we invest any time and money?
• 11:42
  And what should we do?
• 11:44
  And frankly, they think I literally had a conversation yesterday.
• 11:49
  With somebody who said, Well, you know, I said, look, you need to have this because you've got legal and regulatory risk.
• 11:55
  And they said, Well, like, I think we're pretty good on the regulatory risk.
• 11:58
  We understand all the policies and and and our AI is internal.
• 12:02
  You know, like I I I think we're okay.
• 12:05
  Like, what do you mean legal and regulatory risk?
• 12:08
  I said, Well, guess what's gonna happen?
• 12:10
  Okay, your AI is not always gonna behave the way you think it is, and somebody's gonna be affected.
• 12:16
  Now, it may be one of your employees, it may be one of your customers, it may be a business partner.
• 12:21
  And guess what's going to happen?
• 12:22
  In lack of clear legal rules, they're going to sue you.
• 12:30
  And whose fault is it?
• 12:31
  You built AI on top of some other model.
• 12:35
  Well, guess what?
• 12:35
  They're going to sue your company and they're going to sue the model and they're going to go to court.
• 12:38
  And this is going to get resolved in litigation.
• 12:42
  So don't
• 12:45
  Wait for the dust to settle, be proactive, go out, build an AI governance plan to mitigate that risk.
• 12:54
  One, that if something does happen that's not aligned, you should be observing and aware and able to fix it.
• 13:01
  Two, you should be able to go to court and say, no, no, no, we we did all these things.
• 13:07
  And you know, and they thought like, well, you know, we're just using it internally and whatever.
• 13:12
  But no, the reality is.
• 13:14
  Yes, every company is aware of this, but they're not moving enough.
• 13:18
  And maybe they just don't know.
• 13:20
  You know, like should they spend two hundred thousand dollars and have some consultant come in and build them an AI governance plan and take nine months?
• 13:27
  Now my answer is hell no.
• 13:29
  Whate whatever the question is that you're gonna spend a bunch of money on the consult, the answer's probably no.
• 13:33
  Is no.
• 13:34
  And you and I both know this is going to evolve really quickly.
• 13:38
  So what you want to do is go out with kind of core principles that are you know are true, but be able to adapt, but put something in place.
• 13:45
  Don't freeze.
• 13:47
  And we're here to help companies do that and do it very reasonably, uh, because we're on a mission.
• 13:52
  We're nonprofit.
• 13:53
  We're not we're not we don't have to answer any shareholders.
• 13:57
  Yeah.
• 13:57
  No, I want to go back to where where you started with the comparison to, you know, the early days of the internet and and the dot com era.
• 14:04
  Um, because you are out there talking to companies and and and seeing what what engages them in in this area.
• 14:09
  Um, you know, the big difference for me from back then is you know, you take everything and it's just so much faster and there's like two or three more zeros behind everything.
• 14:22
  Um so much so much at stake.
• 14:24
  So certainly, you know.
• 14:26
  we have the experience of what happened with the internet and social media and digital platforms.
• 14:30
  You're absolutely right.
• 14:31
  Back then in the nineties, we were also optimistic and and naive about the potential of the technology, which which is real.
• 14:37
  All all the good really happened.
• 14:39
  But but also there was the bad.
• 14:41
  Nowadays I think no one comes into tech not being aware of all of those problems that have come up.
• 14:47
  So it's always on people's mind.
• 14:49
  But we're in this environment where companies are being told you got to do AI transformation yesterday.
• 14:55
  Um, so in that kind of situation, again, given how much is at stake, how do you make the convincing case to people to step back and do that kind of governance exercise you're
• 15:05
  talking about?
• 15:06
  Well, I mean, one of the things I I I try to impress upon people, and you're so right.
• 15:12
  Uh and guilty as charged, we were so naive.
• 15:16
  I mean, that's you know, something we learned that yes, the stakes are so big now that you have to move fast, you have to compete, but you also can't afford to make a mistake.
• 15:26
  And of course I use examples of people that I think are positioning themselves as ethical companies and and
• 15:36
  You know, and of course the two biggest ones, you know, let's let's look at Apple and data privacy, like they got a lot of benefit from that and they leverage it well in the
• 15:44
  marketplace, get some premium pricing.
• 15:46
  And the other, of course, anthropic and the way they position themselves.
• 15:51
  And so what I say to companies is, well, what do you want to be?
• 15:54
  You know, do you want to be in your industry and be trusted and be the partner who commands maybe some premium pricing, or do you want to just, you know,
• 16:04
  throw a lot of spaghetti against wall.
• 16:05
  You can't afford to do it.
• 16:08
  But yes, there is less naivety for sure.
• 16:12
  But as you know and I know, the compliance departments or the risk departments or the AI governance departments in these companies is not yet stacked, not yet fully staffed, and
• 16:25
  they don't have budgets and they don't have resources.
• 16:28
  But one of the things I tell them to is don't create everything in house.
• 16:34
  One of the things I see, which I think we get a little pushback on is wait a sec.
• 16:39
  Okay, these people are going to come in and they're going to help you think about a governance plan based on the company core principles and ethics.
• 16:47
  Uh and then they may tell you, we will, there are companies that can build tools to monitor your AI deployment.
• 16:58
  And in fact, one of the companies, of course, is somebody I met on your show, uh, Andrew Burt and Lumino.
• 17:04
  And I recommend them.
• 17:05
  I say, you should go talk.
• 17:06
  But then some of them have internal departments.
• 17:08
  No, no, we got it.
• 17:09
  You know, we're building it ourselves.
• 17:11
  And I think that's a big mistake.
• 17:14
  I I think that's a big mistake because as you and I both know, and we learned the early days of the internet.
• 17:19
  I mean, all of us built everything.
• 17:21
  And now it's all, you know, we plug and play the best.
• 17:25
  And I think people have to do that here.
• 17:28
  the compliance people need to think that, yes, the
• 17:31
  Tech people want to build everything themselves, but they shouldn't build their own tools.
• 17:35
  Um, so there are definitely a lot of challenges.
• 17:39
  Some companies don't make it a strategic priority.
• 17:43
  And I feel sorry for them because I do think the litigation is, you know, it's already starting and it's gonna happen more.
• 17:49
  My my j my line to them is look, it took thirty years for, you know, the internet to get litigated to multi billion dollar.
• 17:55
  What's it gonna take for AI?
• 17:57
  Two years?
• 17:59
  Multibillion dollar settlement.
• 18:01
  We've already had that.
• 18:02
  The Anthropic Bark settlement is one and a half billion dollars a year ago.
• 18:05
  So it's all gonna happen faster, including the litigation.
• 18:10
  But you're right, getting people to move.
• 18:12
  You know you're out there preaching it and getting people to move on this stuff.
• 18:15
  Now, I have one theory and I'm an eternal optimist, but this may sound a little pessimistic.
• 18:21
  Yes, people are pinging me, I should have turned that off.
• 18:24
  But I recently wrote an essay about the triangle shirt waist factor.
• 18:32
  Sure.
• 18:33
  So, you know, it's on LinkedIn and I'm a student of history.
• 18:36
  I also studied public policy in high school.
• 18:38
  But then I went and I was a tech entrepreneur.
• 18:41
  And then one day I'm in New York and I'm walking past this building.
• 18:45
  And it's where in nineteen eleven there was a fire and there was a sweatshop making shirts and clothing.
• 18:52
  And the people were locked in.
• 18:53
  There was no fire escape.
• 18:54
  They were locked in because owners wanted them to work and not to steal things.
• 18:59
  And you know, 146 people died quickly.
• 19:03
  Many of them burned to death, some of them jumping from the ninth floor of a building.
• 19:09
  It was that that triggered uh workplace safety regulations, that all of a sudden you had to have fire escapes, you couldn't lock doors, you had to have certain spacing, you
• 19:20
  couldn't have fabrics lying on the floor that were flammable, you had to have sprinklers.
• 19:25
  In fact, it was that that also inspired a woman who became the Secretary of Labor for FDR, Frances Perkins, who then triggered the whole New Deal.
• 19:37
  I mean, the whole New Deal workplace, 40 hour workweeks, no child labor, social security, unemployment insurance.
• 19:46
  So and I wrote an essay saying we're gonna have an AI moment.
• 19:51
  That's like the triangle shortways factory.
• 19:54
  And I really don't want this to happen, but I think everybody, as you say, does not have the naivete.
• 20:01
  We know this is gonna happen.
• 20:02
  And this is gonna be a moment.
• 20:03
  There are already things that have happened that are not.
• 20:06
  Good that are really, really bad.
• 20:07
  And you know, anytime you hear about a teenager and suicide, you're like, ugh, this is terrible.
• 20:16
  But something's gonna happen and it's gonna be some bank's gonna get hacked, or some, you know, deep fake video is gonna incite riots, or some healthcare institution is gonna
• 20:27
  misdiagnose thousands of patients.
• 20:30
  And something bad's gonna happen where a lot of people get really badly hurt.
• 20:35
  And I think then what we're gonna have, and you're starting to see it, but I think then the people are gonna say to elected officials, we have to do something.
• 20:42
  We have to have some guardrails.
• 20:44
  Yes, we need to compete on a global stage, but we can't put lives at risk here.
• 20:52
  And you know, what happens if ten thousand people in a bank account get in their bank get get their savings all wiped out?
• 20:59
  I mean, it's really serious.
• 21:01
  So I wrote that essay.
• 21:03
  I hope it's not true.
• 21:05
  But I kinda believe that's probably what's gonna take to wake people up.
• 21:09
  Yeah, but if that's the case, let's say I'm a company that's deploying AI, why why shouldn't my response be, obviously I don't want that disaster to happen, but it's going
• 21:21
  to happen and then regulation is going to come in and it's going to apply to everyone and then I'll have to do it then.
• 21:27
  Why should I be investing my time now ahead of that?
• 21:32
  Well, I'm give you two reasons.
• 21:35
  You know.
• 21:35
  Uh you know, one is I think the businesses between now and then that don't proactively explain to their employees, their customers, and their business partners how they're using
• 21:52
  AI are gonna see their market share decrease.
• 21:55
  So number one, I really, really believe, you know.
• 22:00
  Let's just take employees.
• 22:02
  They're nervous.
• 22:02
  They're gonna lose their job.
• 22:03
  If I were running a company, I'd say, Okay, look, what we're gonna do is AI gives us unlimited intelligence.
• 22:11
  It doesn't give us wisdom.
• 22:14
  You all have wisdom.
• 22:15
  You've been in this business.
• 22:16
  What we're gonna do is we're gonna grow this company.
• 22:18
  What we're planning to do is grow the company with the number of people we have, but be a lot more productive.
• 22:23
  That's the goal.
• 22:25
  And these are the core principles of the company, and we're applying it to our AI and da-da-da.
• 22:28
  So we're gonna make you partners fully informed.
• 22:30
  So we're gonna transparent with our employees as best you can.
• 22:33
  Doesn't guarantee we're not gonna eliminate jobs.
• 22:36
  We might have attrition, we may not rehire people who leave, but this is you know, be clear.
• 22:41
  Then there's business partners and customers, both of them, they're a little different, but both of them want to hear.
• 22:51
  That you, if they do business with you, your business, that your business has an AI governance plan, you have tools, you're monitoring, that them doing business with you will
• 23:00
  not create risk for them.
• 23:02
  They will not be in the press, you know, being associated with something really bad happened as best you can.
• 23:09
  There are no guarantees, but we're doing this.
• 23:12
  And by the way, our business is distinguished from all these brand new AI startups that are just.
• 23:19
  grabbing data and doing this and doing that and because we're established and we're layering AI on top.
• 23:26
  And I I say this obviously there's some very reputable AI startups from scratch.
• 23:31
  But you can imagine there's all these incumbents that are fighting off all the innovators and the innovators say we can do it less.
• 23:37
  We we we're moving faster, we can do it cheaper, we're doing it better.
• 23:39
  We're all AI.
• 23:40
  Yeah, but where's the wisdom?
• 23:42
  You know, now am I being naive here?
• 23:45
  A little.
• 23:46
  But number one is I do think it's a competitive disadvantage.
• 23:48
  And then number two, and my title of the article about uh is don't be the my subtitle, don't be the factory owner.
• 23:56
  Okay.
• 23:57
  So my subtitle to that piece about the triangle shirt waste factory fire was so if you own a company or you run a company and it happens on your watch and something really bad
• 24:09
  happens, um, it it's a bad feeling.
• 24:11
  You could be personally liable.
• 24:13
  Interestingly enough, in this
• 24:15
  fire there, the people who owned the factory were not found guilty um because they couldn't prove that they locked the doors that how the doors were locked.
• 24:26
  They were found guilty in a civil, not a criminal case, and they had to pay a very small fee, $75 per person killed to the families.
• 24:37
  But I can assure you, in today's world, if AI goes rogue and something bad happens, you're not going to pay $75 per person
• 24:45
  You're gonna pay a lot of money in and if they don't get you on the corporate liability, they may get you on the civil.
• 24:52
  So why do you want all that, right?
• 24:54
  I mean, do the right thing, that's a nice thing to say, but business wise, this makes sense.
• 25:00
  Yeah.
• 25:01
  Is it most important for a company just to have something in terms of an AI governance framework?
• 25:06
  How important is it to really get it right?
• 25:09
  It's so important to just have something.
• 25:11
  Start somewhere because it's going to change.
• 25:15
  The technologies are going to change.
• 25:18
  Your policies might change.
• 25:20
  It's important to get something in place and then recognize you need to evolve over time.
• 25:24
  In fact, that's one of our lessons from the early days of the internet.
• 25:28
  That whatever you build is going to evolve.
• 25:31
  And you have to be that nimble.
• 25:34
  And if you have something, including an AI governance committee, and I know that we're committee,
• 25:39
  But you can build these in ways that they're not burdensome.
• 25:42
  But if you you're monitoring what's happening and you're making adjustments, and clearly you have to have the right guardrails for the right severity.
• 25:51
  So, you know, one of the keys is you don't build an AI governance plan and slow down innovation everywhere.
• 25:55
  It's like, well, that's low risk and that moves faster.
• 25:58
  That's really high risk.
• 25:59
  We better be really careful with that.
• 26:01
  Um, but yes, it's just important to get something in place, know that it's going to change.
• 26:06
  I would say don't overspend.
• 26:08
  And don't spend too much time on the early part of AI governance.
• 26:11
  Get something in place, understand it's going to evolve over time, and build some expertise in-house by starting somewhere, or work with people like us who will help you
• 26:23
  get up to speed and then learn as you go and adjust as needed.
• 26:27
  And let me ask you just yeah about some more of the things that you you've talked about or that the the Ethical Tech Project is doing.
• 26:33
  You mentioned the the fellowships.
• 26:35
  What what attracts um these early career people to want to go and spend the time doing this?
• 26:42
  Well, I think these are people who are so smart and so aware that they're playing with real fire.
• 26:51
  And they're idealistic.
• 26:54
  And to say I'm gonna commit twenty hours over a ten week period after work on Thursday nights in New York City where we did the last cohort, that's a big commitment.
• 27:05
  And I'm blown away by them, really.
• 27:07
  I mean, ironically, there are times when I say, What am I doing teaching them?
• 27:11
  I I understand exper I have experience and thank thankfully they appreciate my perspective.
• 27:16
  But really, I'm in there with somebody from Google DeepMind, and we're talking about AI and ethics.
• 27:23
  And to hear this person talk about the level of awareness internally about ethics is much higher than you think that you don't hear it.
• 27:34
  In fact, one little secret which I think reflects well on Google and her.
• 27:41
  We were talking about uh as a case study, we would did a class, we go through the Facebook algorithm.
• 27:47
  And over 10 years, 10 different tweaks to the Facebook algorithm.
• 27:50
  What was the initial intention?
• 27:52
  What was the effect?
• 27:53
  And what was the compounded effect?
• 27:55
  So obviously switching from chronological to engagement, which post gets put at the top, had an effect.
• 28:02
  And then switching from, you know, text and photos to videos has an effect, and et cetera, et cetera.
• 28:09
  And
• 28:11
  You and then we talked about also the awareness.
• 28:13
  Like I I I defended entrepreneurs.
• 28:15
  Like I I think they just thought it would be a better experience, but at a certain point you have awareness that wait a sec, we thought that tweak would be good, that tweak would
• 28:23
  be good for the experience, but now we're seeing the compounded effect and we're seeing there's some negative effects.
• 28:28
  At that point is when right the company becomes liable.
• 28:31
  So we're having this conversation.
• 28:33
  And I kid you not, the person from Google said, yeah, like we're fully aware of this.
• 28:38
  Mm-hmm.
• 28:39
  We're
• 28:41
  being asked to think about the compound effect.
• 28:45
  You know?
• 28:46
  And I thought that was great.
• 28:48
  Like it's having so the court is having the fact that that trial was going on is before the settlement of sure a couple of days ago.
• 28:57
  People inside of Google are thinking about it, right down to the, you know, people working directly on the product.
• 29:03
  so that's a good sign.
• 29:05
  Uh can they do anything about it?
• 29:07
  It not really, it's above them.
• 29:09
  But it starts to permeate.
• 29:12
  I don't know, I for I after all that I forgot what your therap question was.
• 29:16
  just asking about the the fellows that it's that it's great.
• 29:18
  That you you got into of what's going on.
• 29:21
  Yeah.
• 29:21
  So so I I mean the the the second part I want to ask you is okay, so you have these cohorts that go through this program, they get this knowledge and insight.
• 29:29
  What what do they do with it?
• 29:31
  Uh well, interestingly enough, several of them are consultants now working for consulting companies or companies that have products to help people govern AI.
• 29:46
  So that's really interesting that they're thinking ethics and going into companies and so they're spreading that.
• 29:51
  Um some of them are product leaders who are I'll tell you are products.
• 30:01
  Session on privacy engineering, data privacy had a real impact on people who went back and said, Hey, what are we doing here?
• 30:08
  Are we giving people real control over their data?
• 30:10
  Are we making it easy for people to manage their own data?
• 30:14
  Because we need to be a little more proactive there, you know?
• 30:17
  So I think that's what they're doing.
• 30:21
  some of them will go start companies actually in the space.
• 30:23
  Like some of them are starting companies trying to leverage.
• 30:29
  AI and connectivity to try to create businesses that have kind of ethical outcomes as their you know, whether it's in whatever climate change or equal opportunity for hiring or
• 30:45
  education.
• 30:46
  Like there are people who are out there trying to figure out how do I use the technology.
• 30:49
  But we hope it just infects people, like you know, because and you said it.
• 31:00
  Because there's less naivety, because this is really the first big technology wave that you and I have seen that people aren't excited about.
• 31:07
  They're equally excited and afraid.
• 31:09
  Yeah.
• 31:09
  Okay.
• 31:10
  We were so excited about the internet, so excited about mobile phones.
• 31:14
  Heck, we were excited about satellite TV, whatever it is, right?
• 31:18
  Nobody was thinking like people lining up to get their first iPhone.
• 31:21
  Remember that?
• 31:21
  It was on the news.
• 31:22
  Hey, I got my iPhone.
• 31:23
  I'm so excited to go look this phone.
• 31:24
  And now technology, people are equally excited and afraid.
• 31:28
  Because of all that.
• 31:30
  I think if we can help channel some of the knowledge, some of the fear into okay, what can we do about it?
• 31:40
  How do we think about it?
• 31:42
  How do we channel this fear instead of just saying no, no to data centers, no to this?
• 31:46
  I don't trust them.
• 31:48
  Is what if we can demand from companies and certainly from frontier model companies, but from businesses that use AI that they explain to us how they're using AI.
• 31:59
  you know, why we should trust the data behind their AI.
• 32:02
  Um, what if we work at places and we say we want to know these things?
• 32:05
  That maybe we can build a culture that technology is something that you don't just, you know, throw the spaghetti against the wall.
• 32:15
  You think and and how do you think?
• 32:18
  Well, you think a little bit about what what's our company mission, what are our core values, what what's the effects.
• 32:23
  And how do you build guardrails that are simple yet not cumbersome?
• 32:27
  And that's hard.
• 32:28
  I know it's hard.
• 32:31
  But if that's all we did is help people think, great.
• 32:34
  And everybody knows the difference when they see like grok nudifying images, but OpenAI or Anthropic or Google Gemini won't allow you to do that.
• 32:44
  I know that.
• 32:45
  I I I sat in a restaurant having brunch and I listened to a conversation of some parents with some grant their their parents telling about their kids that they took the kids'
• 32:55
  image.
• 32:55
  And they tried to put AI images on them and that Grok would do it.
• 33:00
  And the others all said, sorry, we will not take an image of your like, you know, 10-year-old girl and and put her in different outfits.
• 33:07
  It's against our rules.
• 33:08
  And I heard them say, Oh well, you can understand.
• 33:11
  That's reasonable, right?
• 33:13
  And so people know.
• 33:16
  That's ethical, right?
• 33:17
  There versus not, right?
• 33:20
  And putting something behind a paywall doesn't make something legal or ethical, just FYI.
• 33:25
  Absolutely.
• 33:26
  Well, so I I I think we conclude that you're still an optimist, as I am too.
• 33:30
  Uh
• 33:30
  You know, you and I I have my moments, Kevin.
• 33:34
  I mean, you know, I'm sure you do too, where, you know, I'm scared and excited about the potential, but I do believe that in general, with just a little human involvement, AI can
• 33:48
  deliver more benefits than harms.
• 33:50
  And by the way, we have examples, and I'll just uh quickly I like to think about industries.
• 33:56
  And what I would say is you and I probably have, I don't know, you haven't had this conversation, a gut instinct about some industries where AI will have more benefits than
• 34:05
  harm.
• 34:05
  And one of them is healthcare.
• 34:07
  Why?
• 34:09
  Because healthcare has two things I say every industry and every company should have.
• 34:14
  They have an ethical framework and they have some guardrails.
• 34:18
  The ethical framework, of course, is the Hippocratic Oath that people in this industry take an oath.
• 34:25
  Not to harm and to help people, and that's part of their pledge.
• 34:30
  Number two, they have guardrails, they have the FDA, they have HIPAA regulations on data.
• 34:35
  So these two things actually make me feel more confident.
• 34:39
  Bad things might happen in healthcare, but probably less than in other industries, right?
• 34:43
  So social social media, of course, a lot of bad things are gonna happen, right?
• 34:47
  But I would argue that the financial industry probably has a lot of regulations that will help it in the long run, hopefully, be safe.
• 34:56
  But yes, I am still an optimist.
• 34:58
  I do this obviously this volunteer work because I believe that maybe some of it will seep in and become part of the culture.
• 35:04
  I know you do the same.
• 35:05
  You teach students every day.
• 35:07
  And we just hope it has some some effect, right?
• 35:10
  Because we are so privileged to be living at this moment in history.
• 35:16
  And for us to lend, even though we were naive and made some mistakes, to at least lend our perspective and experience and stories.
• 35:26
  If it can have some positive effect, well, I feel pretty good about that.
• 35:31
  I I I like to be doing that.
• 35:34
  Robert, it's been delightful to speak with you.
• 35:36
  Thanks a lot.
• 35:36
  Thank you.
• 35:37
  My pleasure.
• 35:40
  Been the Road to Accountable AI.
• 35:42
  If you like what you're hearing, please give us a good review and check out my Substack for more insights on AI accountability.
• 35:48
  Thank you for listening.
• 35:53
  If you want to go deeper on AI governance, trust, and responsibility with me and other distinguished faculty of the world's business school, sign up for the next cohort of
• 36:02
  Wharton's Strategies for Accountable AI Online Executive Education Program.
• 36:07
  Featuring live interaction with faculty, expert interviews, and custom-designed asynchronous content.
• 36:12
  Join fellow business leaders to learn valuable skills you can put to work in your organization.
• 36:18
  Visit execed.warden.upen.edu slash ACAI for full details.
• 36:25
  I hope to see you there.
• 36:26
  oh