“It’s About Us” with guest Bryan Reimer, Research Scientist at MIT

Foreign. Welcome to Building AI Boston. Today's guest is Brian Reimer. Brian is a research scientist at MIT and a global expert on the intersection of AI, human behavior, mobility and public policy. Today we discuss his latest book, how to Make AI Useful, and the important idea that at the end of the day, it's about us. Welcome to the show, Brian.

0:00

Hey, Ann and Cara. Great to be here.

0:30

Kiara and I are huge fans of the book and of course, I'm going to give it a shameless plug right now. For those of you that are looking at it on video, I just. I applaud you. Yeah, we're both fans.

0:32

We both have them. That's awesome.

0:43

Yeah, you gotta sign it next time I see you.

0:44

I can go grab one here too.

0:47

Oh, good.

0:48

You can add a third one.

0:49

The trifecta.

0:52

Yay.

0:53

Congratulations. I know that you've published over 300 publications, you work around the world, you're an in demand guy. And I should say, you and Magnus Lindquist have really developed something special by framing it up as a human issue, and I'd love to break that open with you.

0:55

Yeah. So, Anna, I think the key here is I spent about 20 years or more researching our relationship with vehicles, and the last decade plus focused on assisted and automated driving, where we've been enamored by the embodiment of AI in robots, on our roads. And at the end of the day, the bigger question is how are these systems going to support us? Whether that is in mobility challenges, mobility opportunities. But for a decade or more, I've been watching the technology lead the way, as opposed to the business case, the societal questions and the questions about who and how these systems should serve us. So great. If we leave it all in the hands of corporate America, we get one answer, we leave it all in the hands of social good, we get another. And at the end of the day, we need to find a blended model that works to support us. So at the end of the day, we could talk about the latest breakthrough in AI all afternoon. And I think there's a quote out this morning somewhere that 80% of the advancements in gen AI today are beyond what are really usable to business. So, okay, so are we really pushing the edge to develop algorithms that are needed, or are we really just continuing to fan resources in the allure of Gen AI as opposed to saying, how do we target our investments to ensure the technology doesn't sit on the shelf and get unused and we can begin to reap the benefit of that sooner to enhance society. Yeah, Improve corporate outcomes Very important as well. Can't have a business without bringing in revenue. But at the end of the day, if we're not using this technology for us, what are we doing it for?

1:13

Right? And we're, we're. So when you think about autonomous vehicles being like just a perfect sort of case study for where this, the rubber hits the road, no pun intended or maybe intended, what, what are we missing then in that conversation is that we're, we're too enamored with the technology and we're not thinking about the use case enough or where, where is that disconnect happening?

2:56

So, so I think that, that automated vehicles in some sense are many years ahead of the general AI, chatbot, chat, GTP, etc. In the area of automated vehicles, I think the policy side is the lacking gap. So none of these technical solutions work unless you have an enhancement to human behavior that makes sense. Why are we going to use it a technology that works in the policy infrastructure that supports it? So right now we see a couple key players in the automated driving sphere trying to move this technology faster than the policy really supports that. In essence, we've seen this before. It's textbook case. We're going to control the policy by deploying faster and faster. And unfortunately, that's not something that helps society at large, because what ends up happening is the political forces at some point catch up. So one, we deploy all this technology, then somebody politically says for some reason or not, we can think of a few automated vehicle incidents, one with cruise automation, one with Uber, that say, whoa, slow this down. And at that point, we now need to rebuild human trust with this technology. Because at the end of the day, it's all about the human. It's our ability to trust the technology, that it supports us, it can help us. Trust is built over time and eroded very quickly. So the political forces will win over the long haul. And the question becomes, how do we find that business case for the technology, the consumer who's looking to use it, and the political forces that are willing to support that over the long haul. Now, in the case of automated driving, how safe is safe enough? Is something that we need a societal definition for, not a corporate definition.

3:18

And I think you bring up an interesting point. It's like all these forces colliding. When I think from the consumer point of view, building trust over time. I mean, most people may not even be aware that there are autonomous features in vehicles now. I mean, I love the feature. If I get too close to the end of a, the back end of a Car. My, my, it signals beep beep, beep, beep beep, Foundation. There you go. And that's why over time I think people are, are blending their trust with, you know, just understanding features they can't live without. That drives the, the point with industry. And you bring up the policy side. I mean this reminds me of history. You know, back in the turn of the century, there was this major problem in New York City with carriages and that was manure. And 10 years into that century. Yeah, the, the driverless car. All right, not the driverless. That's where we are now. But you know, you know what I'm saying, We take the horse out, we, we solve the problem. Are we seeing that today? Are we just seeing crazy policies and hype around what's, what's happening when in fact some of this is going to smooth out?

5:00

I, I think it's going to smooth out over the long time. Look, I think automation wins over long time. But I think we are on a very wandering road and a very bumpy road until we get there. I think right now we see a lack of leadership from the federal side in D.C. so we see a lot of push for state by state efforts and city and state leaders who really don't have the political capital interest. Understanding. It's hard to dissect all that, to really ask the hard questions that they should be asking. If we automate, how do we ensure those in less privileged areas are getting the services they need? If there is a natural disaster, who's responsible? Does it make sense for 50% of the mobility fleet or rideshare fleet to be automated? Because if there's an earthquake in San Francisco, that means there's far fewer Uber drivers and taxi drivers around actually provide services in areas where in this case Waymo would say, no, we're not going there, the risks are too high. So these are just some of the, the questions that you should be asking. But there's also some data needs. When we look at AI utilization in particular, we need to look at transparency understanding. You know, we've all talked about AI 2030 efforts here, responsible AI. Part of that key is transparency, understanding what we're trying to do, understanding how it's going to work and ensuring that it does work over the long haul to meet societal needs. So I think these are big questions that are just not being asked today. You know, to me the real usefulness to AI is less in automating today. Yeah, this elements or automation works, but it's usually in processes that are, that are really well defined and almost simplistic in their use case. When the processes get much more complex, driving a car is a great instance. It's really more about co pilot or driver assistance than autopilot. So we think about when we talked about here for a bunch, automated driving, what most of the world doesn't fully appreciate is, is the degree to humans that are humans are still supervising or supporting the robots.

6:01

Yeah, tell us about that because I think that's fascinating and I don't know how many people know.

8:07

We know it's getting better, but we know there is enormous amount of teleoperation or human support for the robots out there. So when the robot can't figure out whether the traffic lane in front is blocked by a bike or an object that they can move around or a pedestrian, it goes to somebody and says, hey, can you help me? It's not individuals, at least in the case of Waymo, that are physically driving the car from what we understand, but it's asking for in essence, real time labeling. Go right, go left, make a decision. But in essence there is human skill making these great decisions behind the robots. The same is true in a lot of what we really need to see in the advancements of AI. So what most people don't fully appreciate is that when we move over to gen AI in particular, it's an incredible opinion. But when you go to chat GTP or Gemini or, or co pilot and you ask it something, we begin to trust that as it's fact. It's not fact, it's just a really good opinion. Once you begin to recognize that it's an opinion, strong one generated based upon a lot of information but not factual, you can begin to use that and use your brain to begin to augment that in new and unique ways. That means you can use it as collaborative writing tool, you can use it as a research tool. But at the end of the day, much like years ago we used to go to the library, weigh up a whole bunch of people's opinions written on paper together. You need to be weighing what the AI algorithms tell us as well to build that Jetsons like science fiction future we want.

8:12

Are there hopeful advancements? I mean what, how do you untangle these problems? What do you, what do you, what does your day to day research look like to, to really addressing how complex this is.

9:45

So a lot of my day to day research looks at how we relate to assistive driving features and in the context of automation. It's a lot thought leadership and discussion with business leaders around the world on how they are beginning to leverage these tool change in their lives and talking to really advanced users and how they're beginning to augment their experiences. Now lots of folks are beginning to recognize that the AI is more of an assistant than it is a replacement. That means the time savings that we dream of often aren't there. So when we see lots of layoffs, we've talked about around the context of advancing AI, some of it's to save money to invest in data centers and the like. But you know, some of it has been touted as we don't need this workforce because of AI. And I think that many are starting to recognize is that AI is a time sink. It is not being used effectively. And if it is being used effectively, it's an amplifier that enhances the output of my team. So I think we need to really begin to be thinking about how do we get the most out of these tool sets to enhance our capabilities. And I know car you've spent a lot of time in the legal realm looking at how do you leverage these tools to support. But it doesn't replace all the skills of attorneys.

9:56

Not going to. Exactly. Right. And it can, it can serve as this. And that's how I find, I use it personally, just like with the marketing work that I do like either for the show or for other things, you know, and Claude, Anna and I both have a, a great relationship with Claude where, you know, for writing it's a really good tool. And then I use OpenAI for more structured thinking and things like that. But it's, it's an assistant. Right. And that's what you're saying, to flip it on its head and think about how is it making us, we always say on the show, giving us more time to be human. More time to do the human thing.

11:12

Exactly.

11:45

It's about features and benefits. Yeah, yeah, yeah.

11:46

But so look, I do a lot of co writing with, with Chat gtp little Claude. And why Chat gtp? Because I have so much of a Persona built in Chat gtp. I look for Claude's opinions all the time, but at the end of the day it's the comms team on steroids at midnight, not what it's saying is gospel. I'm not copying and pasting. That's an interesting word choice. That's an interesting structural change. God, I need to get a hundred words out of this thousand word op ed. Hmm. Where do I start? Used to two or three years ago, hand it to a colleague to read and give me some ideas. No, it's not that I look to the colleague to write for me. I'm looking for the colleague on an idea of what extra words they think can be removed. And that's stuff that we can do in the middle of the night now with electronic support. But if used right, you're iterating back and forth with it more than you would ever iterate with a colleague. Hey, how does this read? If it was being read by a technical audience, how does this read by a political audience? You know, you, yes, you could take a document and focus group test it like that, but many of us didn't have the time to or, or, or wouldn't for other reasons. But here we can get a flavor. Not that I agree with it all the time, I don't. It's just a darn good opinion.

11:49

And that's, that's where it gets sticky. Right. Because if you can use it with intention like that, where you're understanding, I'm not taking this literally, whatever, but that's where it can get slippery for some people and maybe when they're using it for more like emotional support or for, you know, those kinds of. Instead of looking at it as, you know, objectively it's. People can start to feel like this is real and this is what I should believe. And that's where it can get really kind of dangerous. Right?

13:02

Yeah. And that's where I think we as humans need to learn the guardrails between the wow, as we put it in how to make AI useful. And the woe is that we need to begin to think about this as an opinion. We need to be able to weigh that opinion and we need to build skills and how to use it. I don't think the, even the educational system is really been built yet. I mean again, new technology takes time to, for institutions to adapt this to really begin to think about where and how do we use this. You know, really grammar Trek on steroids. You know, I, you know, big supporter of Grammarly's efforts. I mean, Jesus, what a tool. You know, clicking on, on changing the meaning and cleaning up my emails with a click of a button. You know, modern day grammar check. But I think we need to think about what has to remain human, what creative processes are really human and where does AI fit Most of the content I think we look at day to day is so AI augmented that we really begin to be blindsided by what is and whatnot. I know, at least in my voice, I cherish the fact that I want my creative voice to shine and I want to use The AI systems as that comms team on steroids and lowering the pressure I put on those around me to edit. But I still go back to colleagues all the time and say, hey, what do you think of this? And get great feedback that is well beyond the reasoning capabilities of AI tools today. Just go there with something much more honed.

13:29

Yeah, no, I love the way you talk about it as a research assistant on steroids. I think this is why we like featuring stories from Boston. I think you all have it right in a certain way. You have this rich ecosystem of human beings, thought leaders and then you all get together and try stuff and bounce ideas off each other. We hope this show is sort of a magnification of that. I know personally that you know, I worked in the legal field and I would not, you know, I would, I would have been so excited to have Cara's tool. Right. Describe a. I, I got it in a second. But that's because I worked in the field. Any tool that helped me serve human beings was so much more important than going to abstracts. And you know, I think I, I think in Boston you have a healthy attitude about sharing information. Again, that's part of why we try to break down the hype on this show. So I, I'm going to ask you a bit of a hype question then. What do you think these hype factors are in transportation? I know that you know again when cars and, and carriages were at odds, people were blowing up the road system. I mean do you feel like there are a lot of, is there a lot of mistrust on the policy side and the industry side that's just a. Afraid of the transportation arms race as it is.

14:59

Yeah, or space race as I've put it a few times better. I think that it's been coined a few times in the last few months. It's not mine but you know, 2026 feels like 2016 all over again. We have forgotten about a few big hurdles. We've forgotten about the Uber incident in Tempede, we've forgotten about cruises issues and the implosion of that organization. But unfortunately in the transportation world we are waiting for the next Sentinel event. It's not an if, it's a when. It's a law of the probabilities given the complexity of the operating environment. So what is that next sentinel event and how is that going to shape the next phase of consumer trust of policy related automation on our roads? And the question is only time. You know, there is no magic bullet here that can solve all the complexities of driving and make it perfectly safe. Besides, a padded vehicle that provides none of the mobility solutions that we all cherish. Maybe a tank that moves at two miles an hour can protect us 99.99% of the time. Not something I'm looking to move around in. So I think we've forgotten so quickly about these major incidents that have eroded trust. And Silicon Valley is so pro tech. How can we not listen to this? It's so cool. But unless we balance the humans trust in this technology and the behavioral side, the market potential of a willing consumer with the capabilities of the technology and the political support for it, we are going to have potholes along the way. How big and how deep? I am not venturing a certainty that today's efforts in automated driving actually proved to be successful business opportunities. I think it is possible they get there, but we could go back to zero before we come back up again on a long term trajectory of automation winning. So for those that continue to put money and billions into this, yeah, there's more than a double digit chance it all blows up and we start over sometime 20 or 30 years ago from now. There's also a great chance that the explosive power of technology in China takes over efforts that we have here in the US because we are so conservative in one sense and focused that we are not regulating this technology as effectively we should. So you know, if we look to, to the regulation of automation, China is now leading the way both in driver assistance. They just announced regulatory efforts in level three and are moving towards highly automated technology. Again, it is not saying they have a perfect regulatory structure, they don't. But they're asking questions and instilling solutions that we in the Western world are still debating and quite frankly rational ones. So I think the same really plays in this gen fad. And I say fad because there'll be a new technology that is going to make gen AI look like deep neural networks of a decade ago. Yeah, but the context of chatbots isn't new. I mean chatbots have been around for years, you know, back to the UNIX days. So what, what's really interesting about today's chat bots is they've finally proven to be somewhat useful. And you know. Yeah, yeah, we all remember they were. Yeah. But you know, we're investing money in companies to produce better and better AI tools that, that aren't necessarily thinking about how do we balance all these factors together. So so much of the capitalistic pressures around our ability to replace workforce as opposed to augment workforce, workforce and I think companies like IBM have made some fairly big taxes of light hiring, younger work, younger skilled labor being attacked recently and saying, no, this AI trend of laying off isn't going to work for our future. So we got to think about where do we invest in the human capital we call, I'd like to call it the wet computer, the brain that sits between the two ears. You know, we have the ability to reason in ways that algorithms don't yet have the capacity, capacity to do. But algorithms can help and support us in unique ways. But this does not lead to the, to the stratospheric valuations that we see in a few companies right now. It says, hey, we got a great business model here, but it's a small fraction of what you're dreaming of.

16:13

And one of the problems with our wet computer, which I love, is that framing is we, we don't know who. We, we don't want to try. We, we're having trust problems, right? So we can't trust. I mean, and this is, this is a bit sweeping, but it's, I think it's pretty accurate. We don't trust the technology companies right now for good reason. There's been a lot of, we've been burned in a lot of different ways, like social media, on and on. You know, we all remember the early promise of the web and, you know, how far things have moved from that, that time. And we, we. There's a real crisis in trust of public institutions, government. We know that is, that goes without saying. So maybe this is where a third space like AI 2030 becomes really interesting. So, Brian, maybe, maybe tell a little bit about that and where you see this kind of ethical AI movement being influential.

20:26

Yeah, car, that's great. A great segue there. AI 2030 is a international effort around responsible AI. It is not saying that we don't have business cases, but it is saying that we need to take those business cases forward in a responsible way for society. And I think that's a really important effort. It really builds very nicely on some work I was involved with a little over a year ago in the waning days of the Biden administration with, within the U.S. department of Transportation, where I sat on the Transforming Transportation Technology Advisory Committee under Secretary Buttigieg, was vice chair of the AI subcommittee. And if you look at the report, it really talks about lots of transparency issues in understanding what these tools are doing. Now, explainable. AI is the fancy word for it, but at the end of the day sometimes, okay, if we don't understand the algorithm, we Just understand what the objectives of the algorithm are and we understand the performance and we understand the guardrails put around it. So it's sometimes okay to have black boxes in places as long as we understand how they're supposed to function. And in safety centric environments there's a societal responsibility piece there. So AI 2030 is really an effort and I know there's lots of components away trying to start a Boston chapter. It trying to get people thinking about the responsible implementation of these tools. Not saying corporate profits aren't important, corporate growth's not important. It's just that we need to take a technology as transformative as electricity and be responsible in its implementation. I like to think about this in this way. It's not the question. The question really isn't what will AI change, it's whether we are going to let AI change us in car. It really comes back to that trust factor. If we erode the trust too far with a colossal mistake, it will take decades until we begin to rebuild those growth potential, the growth elements of trust. You know, let's look at the airlines for a moment. I know there's a travesty the other day in laguardia, a tragic incident, but the airlines had a very poor safety record until the regulators got involved. You know, so at times you need regulation to move things forward cohesively and at other times the deregulation era has allowed some things to push forward from that. So it's this balancing part of a little regulation helps support the societal needs and make sure we're asking the right questions. And in my word, putting the data tracking on there to ensure we're being honest and transparent with ourselves in the market at large, but also allowing industry to innovate forward in ways that are new and unique and ensuring that if they make those capital investments and they make them responsibly, they have a business market protected to do something with. And I think that's where we're going to wrestle with automated driving. I think that's where we're going to wrestle with the future of Gen AI. And what comes next from this fad is that where is that balance and responsibility to ensure we're balancing the public goods needs with, with the societal interest and corporate needs. Now it's not one, it's all and blending that together.

21:18

And I think that a lot of the concern, at least from the consumer side is job loss. And what does this mean to me personal? How is it affecting my bottom line? And yet if you look at, if you look at like the telephone system. I mean, when telephone operators were obsolete and we went into smartphone usage, suddenly a new market opened up for app builders. I mean, there's always going to be some kind of evolution and, and that fact doesn't necessarily sit well with the general public because they don't see themselves in that equation. But I think another thing I applaud you for is just having conversations that let people peek behind the curtain a little bit more.

24:27

So. And Ann, I think you hit on an important thing. The bottom line, and especially for us in an age where the affordability questions are dominating a lot of headlines, we talk about the cost and investment in gen we cost about mobility, fuel cost at the moment, or food costs or through the roof. I think we need to take a deep look at that balancing point. I'm starting a new initiative at MIT focused on the total cost of vehicle ownership. Because at the end of the day, unless we get our hands around the foundations and begin to talk about these giant cracks in what we can afford, what we can fix, what we can afford to fuel and own over the long haul, the answer is going to be much cheaper solutions that are inferior to some of the societal needs that we really need to dot the I's on and cross the t's on.

25:04

Excellent point. Yeah, well, I, I'm still, you know, recommending the book to people, anybody that wants to. I mean, for crying out loud, you've got a hammer on the COVID If that doesn't point to the fact that AI is a tool. I love that you make complex issues basic enough for people to find themselves in the conversation. How. How do people get your book? Is this available everywhere? I want to just make sure people

25:54

grab it real easy to grab on Amazon, either in print and is an ebook or over at Audible, as well as lots of other booksellers from Barnes and Nobles to a few others. But I think we think about AI. We are not thinking about a single tool. We are thinking about a tool set that is going to continue to expand and grow with time. The question for us as a society, how do we harness that tool set most effectively as it exists today? How do we need to amplify it in different ways in the future? And how do we make sure we find a responsible thread between what has to be continued, investment in the funding needed to do that, and using these tools to make our lives better, more productive, more comfortable, Quite frankly, hoping I can spend more time with my kids. Maybe it's an hour or two a week, but at the end of the day, the joys of living and the opportunities that come with life is something we should be looking at. How do we amplify not building tools for the sake of tools?

26:18

Yeah, well, incredible wrap up. And you know, Brian, I want to encourage you to come back. I think that in this environment that both of you are in, in Boston particularly, I think you're able to digest the practical and the, and the, the upside of AI, which is that it's doubling every four months on the scary side of that, but I really enjoy the human conversations. So you're part of the babe tribe now and consider yourselves an alum with us. I know the two of you have many exciting. Anything you guys want to shout out from Boston at least?

27:22

Yeah, if you don't mind.

27:52

Oh, please. Yeah. That's why we're here.

27:54

So thank you. So April 10th is our inaugural event for the Boston chapter of AI 2030. So just find it on the website. It's ai2030.org right, Brian? Is that right? I gotta make sure I got that.

27:56

Yes, yes.

28:11

So come hang out with us. And it's at Babson's, um, downtown location and there'll be some awesome speakers there talking about AI and responsible AI and you can get to meet some really cool people. So come, come hang out with us.

28:12

Sounds super exciting.

28:25

Yeah. Well, you'll be there, right, Brian?

28:26

Oh, yeah, I hope to be.

28:28

Anything else you want to shout out?

28:32

I, I think that in my side, I'll shout out, as I mentioned earlier for this new initiative around total cost of ownership, then build, building. And I'm doing it out of a passion. If we don't begin to balance the costs of these technologies, the societal needs, the regulatory needs together at least, and we may be able to answer some of these questions, but if we don't begin to talk about them, the trajectory is very clear. Technologies that we can't afford to leverage, mobility options that are difficult to afford. We need to begin to think about where does it make the most sense to put our investments and much more copiously in the future. Let's look, let's thank Silicon Valley for an enormous amount of investment into the, the AI tools we have today. But if we had to start over again today, I'm not sure they have the money to invest. So we better be really smart on what we're doing forward from here.

28:34

Yeah, I love that, I love that it's down to the practical and, and that consideration strikes a chord with me as well. I think a better informed public is always going to fuel this conversation so you and audience land like and subscribe. If you want to hear more, please buy Brian's book. And. And we should say that Magnus is the one that is narrating this on the audible side. R is a big fan. Thanks, Magnus.

29:24

Where? Yeah, you did a great job.

29:48

You did a great job.

29:50

I'm sure he'll listen to this. Yeah, right now.

29:51

Oh, Magnus. Honestly, it's voices like this that matter in this, this time. I. I've never been more excited to be a human being in this time.

29:55

Oh, I am so excited to be a human being. I mean, it is just such a moment to see how some of these tools are going to come together. But I think that we need to remember that it's not how they come together next year. It's how they come together over the next decade. It's going to take this. This is a long and winding path until we see how these technologies really influence how we live and move. It's not tomorrow. You know, I watch my parents use AI tools today. I watch my kids use them in different ways. We all learn from each other. No one has the right answer. I can't wait to start seeing some of the educational opportunities on how to optimize your AI tool. Come up in the next not so sure, not so short distance that are worth watching. It's common, there's no question about that. But I think is really taking this tack and this is what Magnus and I try to get across in our book, that it's, you know, making this useful is about centering around us. It's about the evolution of things to support us.

30:03

And that is the bottom line. It's about us. This has been a fun episode. Really appreciate you bringing it to the practical side. Transportation, of course, is something everybody can relate to. And please come back and share more. Brian.

31:04

Happy to join you guys at some point in the distance. Thanks for having me.

31:18

Thank you for joining us on Building AI Boston. Stay tuned for more enlightening episodes that put you at the forefront of the conversations shaping our future.

31:21