Latest / Thinking Transportation: Engaging Conversations about Transportation Innovations / Will AI Help Us Realize Our Highest Hopes or Our Deepest Fears?

Transcript

Will AI Help Us Realize Our Highest Hopes or Our Deepest Fears? · Thinking Transportation: Engaging Conversations about Transportation Innovations

• 0:14
  Bernie Fette: Hey, everyone. Welcome to Thinking Transportation -- conversations about how we get ourselves and the stuff we need from one place to another. I'm Bernie Fette with the Texas A&M Transportation Institute. Artificial intelligence promises to dramatically alter the way we live -- for the better. But despite its great potential, it's creating a lot of uncertainty and a lot of questions. How, for instance, do we properly consider both the possibilities and the consequences of a technology that's advancing faster than our ability to fully understand it and harness it -- and advancing faster than our ability to regulate its applications? If AI intends to perform tasks currently done by humans, what does that mean for the workforce of the future? Joining us for this episode to help us work through these questions and more is Bob Brydia, a senior research scientist at TTI. Bob, thanks so much for sharing your time today.
• 1:20
  Bob Brydia: Happy to be here, Bernie. Sounds like a great topic.
• 1:23
  Bernie Fette: Yeah, really fast moving , fast growing topic that we're discussing today. Lots being talked about with artificial intelligence, lots being written about it, and I know that you were part of a panel discussion recently for an audience of staff members at the Texas State Capitol . Sounds like that discussion was really well received by that group. So we're really glad that you could come and share some of the same insights with our podcast audience. Starting with a really basic foundation. What is the best way to explain AI for all of us in the non-technical world, please?
• 2:01
  Bob Brydia: No problem. When I think about AI and think about explaining it, I try to envision myself sitting around the dinner table with folks or my mom, and the way that I've described it to people is that really AI is the search for being able to replicate the human brain in electronics. And the human brain really has two sides. It has a creative side and an analytical side. You know, the standard thing that you hear about left brain , right brain people. So AI is really an endeavor to replicate both the analytical capabilities and the creative capabilities of the human mind. In electronic.
• 2:42
  Bernie Fette: We hear all kinds of terms about AI, machine learning, generative AI. Can you sort sort those out for us, please?
• 2:51
  Bob Brydia: Yeah. The truth is that versions of artificial intelligence are basically being able to do really smart things with computers, have been here a long time. I think the first true artificial intelligence inside a computer was in 1952 when a programmer developed a program to play chess that actually learned and wasn't just using rules. So, you know, these terms have evolved over the years. Decision support systems, machine learning. Machine learning is basically looking at things that are happening in the world via cameras or those types of things via sensors and understanding what the environment is saying, what's going on in the cameras, and then learning and reacting based on them. That's one of the premises behind many of the automated vehicles that are on the roadways. They use machine learning from vision systems, generative AI. We've all heard about things like chat, GPT or Bard or any of the other names that are out there. Basically those are using a compilation of knowledge gained from many, many different places. You know, bringing it all in, sorting it all out, and then when you type in a question, it searches the vast amount of information, creates the best answer that it can from that information and presents it back to you.
• 4:19
  Bernie Fette: With artificial intelligence getting so much more attention in recent years, it would be easy for some of us to get the feeling that it's all brand new. You're explaining that's not the case, right? You mentioned the 1950s, an example of a computer that could learn to play chess. Can you give us some other examples of the technology that we've had for some time, but maybe we've forgotten that we had it or just didn't come to know it at the time as artificial intelligence?
• 4:49
  Bob Brydia: Sure. I don't know if it's ever really been termed as artificial intelligence, although it keeps getting smarter, but did you know that the world's first autopilot for airplanes was developed in 1912
• 5:02
  Bernie Fette: ? No, not until now. So tell us more.
• 5:05
  Bob Brydia: I'm not an airline pilot. I don't understand all their technology, but basically used gyroscopes and a bunch of different machines in the cockpit to keep the plane flying level and on the flight path that it was supposed to be. Mm-Hmm. . Similarly, the first landing autopilot system was created and tested successfully in 1937. So planes have been able to land on autopilot for 90 years now. So, you know, when you take a look at those types of things, we don't maybe necessarily consider them as artificial intelligence, but they really are machines taking over what has nominally been a human path . You know , think of your video doorbell from all the different companies that are out there. Most of them have alert to tell you when a package was left by your door, and then they also alert you when the package was removed. Security systems that you can install in your house, not necessarily the full bore ones that come from companies that monitor and everything, but there's systems out there that you can essentially set on your kitchen counter, and if it hears breaking glass at any time, it'll start shrieking with an alarm and alert you that way because it knows the sound of breaking glass, but it doesn't react to your dog barking in the middle of the night because it knows that that's not the sound of breaking glass. So there's all kinds of examples of how intelligence has been built into computers and systems that we use, and really all of that leads up to artificial intelligence. It's all building the base towards the goal of replicating the human brain.
• 6:55
  Bernie Fette: Okay. So let's turn to the subject of this podcast now and talk a little about the applications in transportation. Can you do the same thing that you just did for us in the broader sense and give us a sense of the examples of artificial intelligence and transportation that we've had apart from the autopilot in airliners, specifically in cars, some of the things that we've had in recent years?
• 7:21
  Bob Brydia: Sure. I mean, we can go back 20 plus years and talk about systems that utilized sensors in the roadway to understand the congestion level that was being experienced in an effort to pick up crashes. Mm-Hmm. , the whole goal of that is to get a notification of a crash sooner and therefore be able to roll emergency response faster, therefore hopefully saving more lives, getting injuries under control, and ultimately clearing the road back to normal conditions so that secondary accidents don't happen. Many of the early traffic management centers utilize a very extensive set of infrastructure embedded in the roadway and communicated back to the traffic management center, which was then analyzed on a repeated 30-second or one-minute timeframe to understand what the current conditions on the roadway were. Again, we wouldn't have termed it artificial intelligence at that time, but if you look back at it again, what you're really trying to do is to replicate tasks that humans would be able to do or that humans are doing in some of the current applications. You know, we're using what's known as machine vision or what we referred to earlier as machine learning. So imagine you have a camera mounted on a truck that's driving around the roadways and it's analyzing the paint lines . We've all been in a situation where we're driving down a road, it's raining and the lane lines or the shoulder paint lines, they're very hard to see. So now we are developing mechanisms to utilize a camera to drive all of those roads in different conditions and to gauge how good those paint markings are. That means that we now know where the paint markings are wearing down and where we need to repaint and therefore the more efficient usage of resources by spending dollars where they're actually needed. Mm-Hmm.
• 9:35
  Bernie Fette: And I think that cruise control in cars has been around for a few decades, right? Yes . And again, whether or not we would term that artificial intelligence or not, it is another example of teaching a machine to do something that a human would normally be doing. That being, trying to maintain a constant speed on a journey.
• 9:57
  Bob Brydia: Exactly. Blind spot monitoring is very much the same type of thing. You know, we've had mirrors on the side of vehicles for as long as we've had vehicles basically, but it's always been up to the human to turn around, look over to the right, look over to the left, utilize the mirrors, figure out if something is in their path, and now we have systems that are embedded in the mirrors that look along those paths for us, particularly in the areas that we as humans, when we look around and crane our neck, we can't easily see and they tell us whether something is in that path or not, and then they flash a little light that says, Hey, don't move over.
• 10:37
  Bernie Fette: Right. Which kind of brings me to the next thing that I was hoping that I could get you to talk about, which is that example teaching a machine to do something that a human would otherwise be doing could extend to that human's job. So there's both sides to this coin, this artificial intelligence issue, pro and con, good news and bad news, however you might like to phrase it, but there are competing viewpoints about AI. So if I could get you to talk about both please. So let's start with the positive. What bright promise does artificial intelligence offer us now and and uh , years ahead?
• 11:15
  Bob Brydia: That's a great question, Bernie, and certainly it's very topical and has been on people's minds and in discussion. I wanna make sure that I am understood that my answer is gonna be phrased as an engineer in the transportation arena and not as a computer scientist. So within the transportation world, you know, you're really talking about artificial intelligence. The thing that's in the news the most is automated driving of vehicles such as heavy trucks or something like that. And that is a flash point of discussion in terms of what it's going to do to jobs, in my opinion. And it's certainly shared by other people, but there's also plenty of other people that have a different opinion and we need to recognize all of them. But in my opinion, at least for the midterm future, the next 10 or 20 years, I don't really think much is going to change in terms of job loss or jobs gained. Right now the trucking industry is known to have a shortage of between 80 and a hundred thousand truck drivers. You know, we all drive along the road and see the signs on the back of nearly every tractor trailer that says, "Hey, you need a job? Call." Right ? And yes, automated driving would take some of those jobs out of the seat of the, of the vehicle, but we're not that close to automated driving at scale. Okay? Mm-Hmm , . Now there are plenty of companies that are working very diligently to create a safe and effective and efficient method of automated driving in trucks. And there are companies that are trialing those systems on the roads today, but we're talking 10 or 20 trucks. We're not talking 20,000 or 200,000. So in terms of taking jobs away from people that currently have them, I don't really think that's going to occur. Some of those jobs may change or shift in the types of duties and things that they need to do, but we're also gonna be creating new jobs. We're going to be creating different types of maintenance jobs that are needed on automated trucks to really, you know, look at their systems and ensure that they're safe as opposed to what is now solely a mechanic's job checking the brakes. We're still going to need that, but now we're also going to need computer experts to look at the data, look at the system , look at the technologies that are on the truck, make sure everything is operating kind of like the pilots do as a pre-flight before they take off on an airplane trip. Pilots checking all the systems in the cockpit, those types of things. We don't have those types of jobs currently in the trucking industry because there hasn't been that level of technology there, but those will certainly come in. Additionally, it's very possible that automated trucking could lead actually to an increase in jobs. Again, using artificial intelligence to drive, the vehicle is going to roll out, it's going to roll out on the interstate. So instead of you getting into a truck in California and driving it all the way to Chicago, maybe you have a local driver in California that loads the truck and gets it to the point where it can go onto the freeway and then it sends it off, and then it drives to Chicago by itself, and it's met by a driver there in Chicago who then takes it down into the urban streets and the warehouses and all of those types of things.
• 14:50
  Bernie Fette: Just a different twist on the whole first-mile, last-mile concept that we hear about a lot in the trucking industry. Right?
• 14:57
  Bob Brydia: Right. It will take some time to see how this all plays out. For every article that there's talks about the loss of jobs, there's another article that talks about the jobs that are going to be gained.
• 15:11
  Bernie Fette: So what are some of the bright promise notes that we can think of in terms of what to expect from artificial intelligence in the near term years ahead?
• 15:20
  Bob Brydia: I think one of the easiest ways to visualize what the bright promise is is a show that many of our listeners might be familiar with, and that's Star Trek. When you think about Star Trek, you see very advanced technology, but controlled very easily by humans walking around with tablets and pressing, you know, large buttons on the bridge and , and yet they're flying through space at the speed of light or exceeding the speed of light. That sort of utopian view of what society could be is really where the bright promise of AI is. To bring that down to more practical terms for today's state of where we are with technologies , it's basically about taking away the redundant, boring, mundane tasks of the things that we do in our everyday jobs, and that could be both physical or mental and turning that over to a computer and allowing the person that was doing that to really have a focus then on something that's newer, more challenging, more creative, and being able to expand.
• 16:32
  Bernie Fette: Simply a higher and better use of that person's time, in other words?
• 16:36
  Bob Brydia: Yes, that's a great way of putting it.
• 16:38
  Bernie Fette: So there's a lot of promise there, but there's a lot of doubt. Again, as you said, you can read both sides of this issue depending on where you read and who you listen to. But I was noticing a survey that was published in Forbes Magazine recently about people's attitudes toward ai, and I'll just hit a couple of points to see what your reaction might be . More than 75 percent of the people who were surveyed were concerned that AI would cause misinformation and the loss of human jobs. 77 percent said that AI will be used to deep fake their faces or voices to commit fraud. 70 percent fear that AI will be used by other nations in information warfare against us. Just a handful of notes that paint a pretty gloomy picture.
• 17:31
  Bob Brydia: Yeah, there certainly is a dark side of AI. We've all seen or heard in the news about the deep fakes, you know , actors that were never in a situation, but their image was put into that situation and then it's all over the news that they did something. There certainly is disinformation on the internet. We've seen even, well-known publishing houses have been using stories that are created automatically by AI, by the generative language models that are out there in AI and posted as news stories without ever being fact-checked. Mm-Hmm . and mm-Hmm . because not everything that goes into the database of from the internet is true. You know, some of those stories have had mistruths in them. Mm-Hmm . . So yeah, AI could be a way to perform bad actions to distribute disinformation and it certainly could be used as a weapon against countries or people or anything else, and we're going to have to really take a look at that and to guard against that.
• 18:36
  Bernie Fette: Well, and makes me wonder how do we guard against that? And I'm afraid this is about the point where you're gonna tell me that the answer to my question is over your pay grade . So I apologize in advance for , for asking it, but do you have any thoughts there about how we guard against the nefarious uses or the potential nefarious uses of AI?
• 18:58
  Bob Brydia: Yeah, before we get to that question, which is over my pay grade, I want to go back to some of the other aspects of AI as opposed to the nefarious uses. AI is also known to require a lot of computing power. There's an extensive amount of data that's required for it . Certainly the energy requirements that large scale AI requires. You know , all of those are other concerns related to the dark side and bad actors, invasions of privacy, all of those go along with them . Mm-Hmm. in terms now of going back to the question of how do we guard against that, you know, it's a cycle that plays out almost constantly with technology that the technology advances faster than the ability of people to control it. Mm-hmm. and control may not be the right word, but most technologies that are developed are developed with alternate, maybe kind of silly, but you know, wholesome good thoughts, right? Like this technology could really help people and then there is folks out there that are like, Hey, I could really use that technology to scam grandma out of a thousand dollars by phoning her and deep faking her grandson's voice and say that he's been arrested in England and he needs a thousand dollars to get outta jail. Hmm . Okay. The levels of protection against that are numerous and really they're developing now. They're starting to look at that. Numerous organizations are starting to look at the issues associated with artificial intelligence, trying to understand what best practices are and creating the foundations of experimentation and use to help ensure that it's not used by bad actors to scam grandmas out of a thousand dollars or deep fake celebrities into positions that tarnish their careers and those types of things.
• 20:59
  Bernie Fette: Right, right. Which makes me wonder on the subject of how do we guard against such things, sometimes there are policy-related answers to that question or policy-related actions that can be taken or regulations for AI. What are your thoughts on that front? The whole issue of how do we regulate or to what extent do we regulate development and use of artificial intelligence?
• 21:25
  Bob Brydia: That's a great point, Bernie. I think we're going to see an increasing amount of effort on policy and guidelines and even eventually enforcement of how artificial intelligence is used. Just this morning, the federal government released, I believe it's a policy brief and set of guidelines for how artificial intelligence should be used. I haven't had a chance to look at it yet, but I got a notification on my newsfeed that this was out there. So, and I think that board was only created five months ago. So the fact that they've moved so fast to push out some preliminary guidelines and information about AI as it pertains to the whole country and it's used in the federal government, speaks to the importance of understanding the right way to use it and how it could be used negatively.
• 22:18
  Bernie Fette: It might also speak to the urgency that some people feel.
• 22:21
  Bob Brydia: Yes. Agreed.
• 22:24
  Bernie Fette: Are there any emerging questions that need to be addressed in your view, either through laws or agency regulations? And I'm not asking you to produce the answers or endorse a particular policy position because that's not your job, nor is it the job of the agency that we work for, but maybe share your thoughts on what questions we need to be addressing right now, whether for the near term or further down the road.
• 22:49
  Bob Brydia: Okay. I will caveat that by saying that these are my opinion only, and like you say , I'm not responsible for producing those types of responses, but um, like everyone, I have an opinion about things and sure. Uh , happy to share that. I think AI as the current state that AI is in, even though we've had it for decades and have called it different things, I think the state of AI that we are in now, which is such a dramatic increase in capabilities and potential, is perhaps in its infancy of, you know, this stage of technology maturation. So part of the issue that needs to be determined is where really does artificial intelligence fit in and what path can it do? The first level I think is just understanding the uses. We've done a creative use of artificial intelligence by utilizing machine vision from a camera and looking at the grade arm of a railroad grade crossing. When the grade arm is up, of course the crossing is open to use, but when the machine vision looks at the grade crossing and sees the grade arm coming down and that it's closed, we now have the ability to send an alert to emergency services that utilize that route as their primary access point to get to any of the calls that they're called out on and tell them that that crossing is blocked and that they need to take a different route. Mm-Hmm . . At the same time, we can also get information about the train in terms of it heading speed , the number of cars that are on the train and all that can be used to determine how long we expect the crossing to be closed. So you put all of that information together and that's a very novel use of technology that could be implemented in numerous places to help advance emergency services and their understanding of the route and the congestion that they face on the route. Again, part of the answer of what do we need to be concerned about with AI is continuing to understand the proper application of where, when, and how it can and should be used. Mm-Hmm. , in addition then in the background, we have to be aware of the different types of data that it uses, the different types of data that it collects. We don't want to get into privacy concerns. Privacy concerns are significant. We all live in a very, very connected world with smart phones and devices all over our house that are connected to the internet at any given time. All of that produces an incredible amount of data that people can get to and analyze and use for their own purposes. How many times have we visited a website that was selling something where we made a purchase and then the next 30 websites that we go to for the next two weeks all have an ad plastered in the middle of the story that we're reading? That is the exact product that we bought on that website. You know, many people term that an invasion of privacy because it's tracking our purchasing habits. So that's one example of the use of the data that is collected trying to get you to buy more things. We would need to be careful that AI doesn't do that same type of effort to a larger scale.
• 26:36
  Bernie Fette: What I think I hear you saying at the risk of oversimplifying, but just to help me understand , uh, help others understand, it sounds like some of the biggest questions that we need to be asking right now have to do with what are the appropriate uses for this technology, because there might be some things that AI can do, but perhaps we as society don't want it to do just because it can do it. Am I characterizing your response fairly there?
• 27:04
  Bob Brydia: Yeah, and I think that's a great summation, much more concise and better than I originally stated, so thank you.
• 27:10
  Bernie Fette: Okay. Tell me what the future holds. This is your chance to channel your inner Gene Roddenberry or your inner Stanley Kubrick .
• 27:18
  Bob Brydia: I think the future with respect to AI is that eventually it's going to be pervasive. When you look at technology phases throughout history, and I'm not a historian, but we've all seen cycles where a technology is introduced into society and then 20 or 30 years down the road, it's everywhere. I mean, cell phones , latest and greatest example of of that happening. When the first cell phone was introduced, it was about the size of a shoe box . It had to plug into your car and now there a couple of ounces fit into your pocket and have hundreds of times more computing power than the first lunar lander mission from the U.S. I think that cycle is going to continue to be repeated. AI is going to go into our normal lives and going to be accepted. We already have refrigerators that know what contents are inside the refrigerators, know when their expiration dates are, wow , and we'll send you a grocery list of what you need to replace. We've got all these systems on vehicles that are essentially versions of AI, analyzing the environment and helping you drive safer and avoid crashes. We've got these systems that are coming into houses, smart light switches and light bulbs that turn on automatically when you walk into the room and everything. Eventually, AI, I think, is going to infuse itself into every aspect of our lives that we really know it because the technology is going to continue to advance and going to continue to integrate into our normal life, and that's all the more important why we have to have effective understanding of best practices and the guidelines to minimize the nefarious use, that certainly could result in that.
• 29:21
  Bernie Fette: This is not your first time as a guest on this show, and so you know, I've asked this question before, so you're free to repeat the same answer you've used in the past if you wish to or you can come up with a new one. What is it that motivates you to show up for work every day ?
• 29:38
  Bob Brydia: What motivates me is the challenge and the ability to be a part of the future and to help transform today's transportation to meet the needs of tomorrow. Those needs are continuing to change. They're continuing to morph with respect to the technologies that are developed and in addition to having additional needs. We also have greater tools now, so the challenge is continual of trying to utilize the technologies effectively and efficiently and not haphazardly to address the need of today to move into tomorrow's transportation world. That's the excitement of it, and I love my job and I love what I do.
• 30:27
  Bernie Fette: We have been visiting with Bob Brydia, a senior research scientist at TTI. Bob, thanks so much for helping us understand this rapidly changing and rapidly growing field. We appreciate it very much.
• 30:42
  Bob Brydia: Thanks, Bernie. Appreciate being here and had a great time.
• 30:47
  Bernie Fette: Artificial intelligence seeks to replicate both the analytical and creative capabilities of the human brain. There's mounting evidence to show that it's already finding its way into our daily existence, influencing in a positive way the myriad ways in which we work, shop, move about and live. The lofty promise of AI is here, as are the doubts. Depending on who you ask, AI represents either the manifestation of society's most enthusiastic goals or its deepest concerns. Maybe it's a little of both. Thanks for listening. Please take just a minute to give us a review, subscribe and share this episode, and please join us next time for another conversation about getting ourselves and the stuff we need from point A to point B. Thinking Transportation is a production of the Texas A&M Transportation Institute, a member of the Texas A&M University system. The show is edited and produced by Chris Pourteau. I'm your writer and host, Bernie Fette. Thanks again for joining us. We'll see you next time .