This week's episode of the trip Cast is sponsored by the Texas Tribune Membership program as we celebrate Amplify Austin. Hello, and welcome to the Texas Tribune Trip Cast for March fourth, twenty twenty five. I'm Matthew Watkins, editor in Chief of
the Tribune. My two co hosts, Eleanor Klibanoff and James Berragon are nursing their oscar parties hangover, aren't able to make it today, so instead I am joined by Cherry Greenberg, a former state representative and professor of practice at the LBJ School of Government at the UT Austin.
Hello, Sherry, thanks for being.
Here, my pleasure. Hello.
So today we are going to be talking about a subject that you have been studying for several years now, many years even before it became sort of the buzzword in the Texas capital and beyond the idea of artificial intelligence and regulating it at this date level, When the idea of regulating businesses comes up in the legislature, there's usually a pretty strong consensus among Republicans to sort of stay out of the way in the name of growth
and free enterprise. But the conversation this year around regulating AI is turning out to be a little bit more complex. You have powerful conservatives such as Senate Business and Commerce Chairman Charles Schwartner and the Chair of the House Doge Committee, Giovanni Copriglioni, comparing carrying bills that would impose strong guardrails and rules for the technology, but they're finding some opposition from members of their own party. Senator Lewis Cole course
was questioning a bill by Schwartner earlier last week. Tech impresario Joe Lonsdale has really come out against some of the legislation, despite being pretty usually aligned with Republicans in the state. It's a fight that sort of avoids that partisan brack breakdown. But I think there's some fear kind of driving both sides. On one hand, you have lawmakers worried about the dangers of AI on people's privacy, their freedom, and the possibility that really important decisions might be made
by computers not people. On the other side, you have folks who are worried about going too far and restricting AI that would leave Texas behind other states in the technology and maybe even pulling the United States with them behind adversaries like China. So Sherry, You, like I said, have been watching this for a very long time. A lot of the focus so far earlier on this session is on a major bill called TREGA, the Texas Responsible
AI Governance Act. Can we start by just having you explain what's in that bill?
Sure? And for background, which you really alluded to, is what I call the promise in peril of AI. Right, there's great promise with AI with artificial intelligence and healthcare detecting diseases that we wouldn't be able to, but there's peril too. So if we can use healthcare and as an example, big privacy issues, and we also talk about balancing innovation with rules and policies and regulations that provide transparency, accountability and guard against invasion and privacy. And so that's
really what we're dealing with here. And with artificial intelligence, we do find that there are a lot of common concerns across political parties and sometimes you find alliances that you wouldn't elsewhere, and you also alluded to that. So with TREGA, this is a bill that was filed its House Bill seventeen oh nine by Representative Giovanni Capriconi, and there was a lot of discussion about this bill over
the interim. He had a very large stakeholder group that many of us participated in meetings and whatnot, and I think he knew that this is a bill that would have a lot of discussion once it was filed. So TREGA is really a builds based on risk mitigation, and this is what the EU bill really is based on.
And also other states such as Colorado which we can look to, and it's mitigating against risk, which is also what this the Nationaltitude for Standards has really set out in its framework, which is part of the National Science Foundation. So TREGA would create obligations for developers and deployers of any AI system that is a contribution, you know, decisions relating to employment, finance, healthcare. Texas would also be the first state to require high risk AI system distributors to
take reasonable care. The bill would ban AI systems that pose unacceptable risk. Private litigans would have limited right of action. The law would require generative AI to developers to keep detailed records of data sets used to It would exempt many small businesses. So you can see it's very much
a risk based system. It is also a build that very much focuses on what we call algorithmic discrimination, and in the bill, that means any condition in which an artificial intelligence system, when deployed, creates an on lawful discrimination of a protected classification in violation of the laws of this state or federal law. And then it further goes on because you're wondering, what's an artificial intelligence system? Right?
That means the use of machine learning and related technologies that use data to train statistical models for the purpose of enabling computer systems to perform tasks normally associated with human intelligence. And this is really where we get into the AI, the artificial intelligence realm or perception such as computer vision, speech, or natural language processing and content generation.
Okay, so there's a few things that are sort of key phrases in this bill, one of which you already mentioned high risk AI systems. What do we mean by a high risk system in this case.
So a high risk system in this bill, and again this bill is really based on risk mitigation. High risk artificial intelligence system means an artificial intelligence system that is a substantial factor in a consequential decision. The term, though,
does not include an artificial system. If the artificial intelligence system is intended to detect decision making patterns or deviation from prior decision making patterns, and is not intended to replace or influence a previously completed human assessment without sufficient human review. So the following technologies unless the technologies, when deployed, make or a substantial factor in making a consequential decision.
And a malware anti virus calculation, cybersecurity database is data storage, firewall, internal domain. The list goes on and on.
So those are things that are not included in the building.
But what we're really talking about here are basically using AI using computers to make kind of important, impactful decisions about customers or people in this state.
That's exactly right, okay, And.
I think the other interesting thing is the idea of using reasonable care. Right, this this idea that if essentially it's up to the developers, the people you know, like open ai or or you know, Twitters developing some some AI systems and things like that, or even the distributors or the employer, the deployers, the people who are you know, utilizing an AI system to evaluate, you know, whatever it is they might do, make decisions about healthcare, health insurance
and things like that. What it's basically saying is that they need to take steps that are done reasonably to protect people from that kind of discrimination in the system.
Is that right?
That is exactly right? And there is a kind of I guess provision you would say for piloting, you know, with a sandbox where there can be some innovation taking place.
Okay, okay, very good. And so I mean what's interesting here. This is the bill that's gotten a lot of attention. It'll be interesting to see whether it advances in its current form, and we can talk about that a little bit more. But what's really struck me about this is there was a lot of conversation, you know, when when chat GPT was really out there, a lot of conversation about the dangers of AI. A lot of those felt like sort of conversations around existential dangers for humanity.
Right, are the right going to take over?
Yes, sentient and they're not cynient.
Exactly exactly, And that's not really what this is protecting about. This is more about privacy concerns. Are AI you know, basically things without a conscience is going to be making you know, life or death or you know, major financial decisions for the lives of people in Texas.
Is that is that right?
Yes, this is not about whether AI is, as we say, sentient, right, It is about looking at those areas that people are most concerned about when they talk about AI, protecting their privacy, having transparency, having accountability, knowing when you know AI is being used. The bill, you know, requires the as we said, an unacceptable risk. It prohibits systems that without express consent, identify emotions, capture biometic intent, you know, those attributes that
were really sensitive about as people in this atmosphere. It also would require generative AI developers to keep detailed records of data sets that they used to train. So the bill, again going back to the unist frame work, really looks at risk and risk management and some of those perils that we talk about. The risk management framework looks at governmap, measure and manage and that is a framework that has been used by NIS, really by the EU, by Colorado
and other states. Now that's not to say that people everyone is happy about it, right, and really with AI, we can go back to nineteen fifty with Alan Turing r. AI has been with us a long time and in our daily lives. But until chat GPT appeared on the scene, people weren't really cognizant about it. Once people became cognizant, then it's oh, what does this mean for my privacy? My data? Is there transparency? Do I know when AI
is being used? However, there is this, as I said, promise in peril or this also this issue of how do we have innovation right without too much risk? As you mentioned, people are concerned if this stifles innovation, where will that mean for us as a state? We know that we're getting ready to have some big investments in this state right in AI. What would that mean for the country you mentioned China, And of course we've heard
a lot about China and AI recently. But on the other hand, people really are concerned about I would say, their privacy and about this algorithmic discrimination in decision making and also other areas. There are other bills, of course, have been filed dealing with using AI for mental health, or using AI and making decisions with insurance and healthcare. So there are other concerns that people have.
Yeah, it seems like the would you say, the broad kind of fee of a lot of these bills is sort of designed from a kind of consumer protection standpoint.
Yes, I would say it is designed from a consumer protection and transparency standpoint. The bills, you know, they may say, well, you can use it for mental health, but there has to be a provider involved. Keeping the human in the loop. I would say, you know, when you're looking at health insurance, you know you have to disclose perhaps that you're using eye and you have to if it's for specialty care,
specialist for instance, has to review it. So consumer protection, transparency and accountability and kind of I would say, keeping that human in the loop.
It's interesting because in some ways this feels like a very familiar kind of conversation between innovation and industry and creation and all those kinds of things, and you know, the consumer protection, the fears about the negative impacts of this technology, and I just I find it a little bit striking that we're seeing, you know, at least a significant segment of the Republican Party sort of going in
the direction of consumer protection this time around. What do you think it is that's different about AI that's setting up this conversation.
Well, certainly, I think that there are a lot of legislators who have followed the path that for instance, social media took and lack of privacy with data and they're feeling that this is a big concern of their constituents. And seeing now with AI, you know, your data, your personal data is being used to train it. For instance, your personal data may be available. You may not know when it's being used. It may you know, is it
making decisions without a human and a loop. And I think that those are issues that you know trands, they're they're beyond party boundaries, and that many constituents across the board are very concerned with these issues and concerned with the I would say, the path for instance, that data and social media has taken with you know, lack of privacy and now seeing AI.
Yeah, I mean among the things in this this trega bill, it bans deep fakes right, and certain use of biomarkers like facial recognition that you know, they don't want people pulling a bunch of pictures from Facebook to be able to identify your face, you know, wherever you are walking across the street. Another thing it bans is social scoring,
which I'm quoting from the MIT Technology Review. The definition of that a practice in which authoritarian governments, specifically China, rank people's trustworthiness and punish them for undesirable behaviors such as stealing or not paying back loans. Right essentially seen as a you know, sort of way to rank citizens or or identify citizens as types of risk and that kind of I mean, I think, you know, another big
part of the Republican Party is freedom right. And there were a little bit about surveillance and just you know, whether it's big companies or the government itself sort of you know, reaching into the personal lives in a way that feels uncomfortable to that's right.
It's that privacy issue. They're very concerned about you not knowing your data is being used, you not knowing that your face is being used, to the biometric markers, your voice being taken. You know, there was a case in Arizona over a year ago, a really horrible situation, and there was legislation in the Arizona legislature where there was a deep fake of a young woman's voice and it was it was her voice, and they called her mother and her younger was there too, and said that she
had been kidnapped asking for ransom. Well, in fact, she had not been kidnapped, and that was not her voice, but it was a replica, let's say, of her voice a deep fake. So there are a lot of concerns I think about criminal elements, about privacy, and as you said, about freedom, my freedom to walk down the street without being ranked as a threat to society, or my freedom to control how my image and likeness and voices being used.
All right.
Today's trip cast is presented by the Texas Tribune Membership program. The Texas Tribune's in depth coverage of Texas politics relies on donations from people like you. Right now, your donation to support this reporting goes twice as far thanks to a fifteen thousand dollars match for Amplify Austin Day. Support independent journalism by making a donation online today. Okay, So I want to ask you a little bit about just the state's ability to succeed in this process.
Right.
We saw a you know, after this bill came out. Joe Lonsdale, who I mentioned sort of in the opening, a venture capitalist, a very influential one both politically and in the tech world, came out very strongly against this trega bill. He says he did kill innovation, kill jobs,
let China win the AI race. He essentially argued that it would create a new AI governing agency in Texas and see the regulation of AI to you know, bureaucrats, right, the dread of bureaucrats tell me a little bit about the industry response.
Sure, sure, and it's not you know, you mentioned his statements, but he's not the only one. There were articles, I think in Forbes there was an opinion piece, and I believe the Dallas Morning News last week.
If you google TREGA, all the top links are industry aligned groups with articles and other things kind of speaking out against this.
Right, you know, there was another article, everything is bigger in Texas. TREGA is no exception. But will it, you know, provide accountability? So and they're saying no, it wouldn't. And so any time that you file a significant bill, Having spent ten years in the legislature myself, I can tell you that you're going it's a starting point, and you're going to expect that there will be changes, that there
will be amendments. And with a bill like this, even though there was a stakeholder group and a lot of input, I think the bill that's filed, I think it's safe to say probably is not going to be the final piece of legislation, and so we take it as a starting point. And certainly there is this concern about innovation.
You know, we're in another industrial revolution, and with every industrial revolution, whether you say we're in the fourth or fifth or whatever, you know, we have seen concerns about the you know, the perils of the new technology, but also the promise. And there are concerns, you know, with AI about work and future work, for instance, will people be out of jobs or will there be new jobs?
And we can look back at history and see this, So I think there there is a real issue here with balancing innovation and protecting people's freedom and privacy and having that transparency and that when you file a bill like this, you know you can expect that there will be some.
Amendments, right right, absolutely, you know.
I Lonsdale, after he put out a little sort of a tweet threadgras this, he came back not too long later said he met with quote relevant leaders in Texas and that they agreed, again I'm quoting him, conservatives must block the creation of AI regulatory agencies.
So I think, you know, you're exactly right.
It sounds like there is there's going to be some resistance to some of these ideas. Right here, I want to talk a little bit about enforcement here. So the bill as it's currently laid out, would place enforcement in the hands of the Attorney generals. That's correct, organizations that violate the rules and that would be subject to fines in the like ten thousand dollars range per instance. But of course we know that a lot of these things are done on a mass scale, so those ten thousand
dollars finds can can really add up. I mean, you may remember the Attorney General's lawsuit against Facebook for use of facial recognition. Again, it was one of those you know, individual finds. Those individual finds ended up adding up to one point four billion dollars worth of settlements. So you could really see, you know, some aggressive actions related to this.
But I mean, one of the things I wonder about is I read this bill is Texas state government isn't exactly known for being a tech know, logically advanced, and how it goes about things. I mean, you know, the computer systems, all those types of things.
Well, there is another bill that calls for training of certain employees on cybersecurity and artificial intelligence.
Well there you go.
I mean, what does Texas have the capacity to regulate this. Do you think if even if a bill like this were to.
Be passed, you know, the Attorney General's Office, as you said, would be the enforcement mechanism, not the Department of Information Resources, does it have the capacity? I think that depends on the who are the employees and what training do they have? And as I said, I think that is incredibly important having of course the systems in place, but also the training of state employees.
Okay, and so the other thing here to be thinking of is just again that debate. I mean, this is going on in a lot of different ways. I think one conversation is do you do this in a piecemeal approach? Right with individual you know, here's an AI healthcare bill, as an AI education bill and all those types of things. There are a lot of different measures about that. But even when you get into the individual cases, it can it can get a little bit dicey when you go
into the details. There was a conversation in the Senate Business and Commerce Committee last week Charles Schwartner finding a bill that would ban health insurers from using AI to deny treatment or downcoading patient's treatment. He argued pretty passionately against the idea I'm quoting here that insurers are quote deploying AI that is focused on their bottom line in my opinion, over you know, patient well being basically.
But los culd course.
Another very conservative senator came right back at it and said, you know, this bill could cost billions of dollars to the state, right that we have you know, Medicaid programs and things like that that might be able to use AI to root out fraud or identify in efficiencies and things like that. And I understand that there's fears, but there's also real opportunities here for.
It to get back to the promise and the peril again, right, And this has been a national issue. You've probably seen some headlines nationally about insurers using AI and were they relying solely on an algorithm to deny coverage and insurance, particularly for specialty care. And there have been several bills filed, as I mentioned, dealing with healthcare, with mental health, and
with health insurance. And I think the question will be do you have a bill that denies the use of AI completely or do you have a bill that says you can use AI. But as I said, you have to have a human in the loop, a human has to review it. And you know there are other bills that have been filed that say, well, you can use AI. However, if it's specialty care, the specialist, for instance, has to review it, so you keep that human in the loop.
And a lot of what we're seeing around the country and around the world is that discussion of whether it's in domestic violence cases in Spain where you had a situation where a particularly police department was relying solely on AI predictive analysis to say whether an abuse or would abuse again, and there was not a human in the loop, and a woman ended up dying. And so a lot of the conversation now has turned to using it but ensuring that a human makes the final decision and is reviewing.
So I think it's this discussion of where are you on this continuum. Do you not use it at all, do you use it and you solely rely on the AI, or are you somewhere in the middle in this continuum where you use it. But you must have humans and professionals who are in the loop and who are reviewing and making the final decision.
How important is it that the legislature figure out some answers to that question this session. I mean one of the challenges of the Texas legislature is they only meet every two years. That's right, at the pace that technology is growing and changing. Is this sort of their last chance to sort of put down a marker on this, do you think? Well?
I think that's why you have seen so many bills filed this session. As you said, technology is changing, as we are doing, as we are talking right now, right, and with the biennial legislature, you do have limited opportunities. And again with chat GPT, with what we've seen with health insurance and others. Yes, I think that there are many members across the board who feel that it is important this session to get a handle on this and pass some legislation.
Okay, I mean my other question is is the state level the right level to be doing this?
Right?
I mean, a lot of this employment will I mean, this AI deployment will be happening outside the state but affecting Texans, but it also affects you know, Californians and right Vermontors and everyone or else? I mean, should this not be a federal level discussion as opposed to a state level well?
Or should it be a worldwide I mean technology does not know boundaries, right. And this is a really good question because as we know, you know, federal law superseded state laws, and state laws supersede local laws. So the Austin City of Austin has been looking at policies and looking closely at what's happening with TREGA. They want to, you know, align they have common issues and concerns, for instance. But what you tend to see is when you don't
have action at the federal level. So we have had no congressional action and we you know, the Biden executive Order of course no longer exists. And so when you don't have action at the federal level, and legislators have constituents, and if those constituents have concerns, what you see is there's this vacuum, and you will see the states, not just Texas, step in. And if constituents at the local level have concerns, right, and the states are not stepping in,
you will see individual cities step in. And then you will see industry say, wait a minute, we have this patchwork of different cities within the state or different states, right. I think that what happens in a Texas is very important whether it's textbooks or AI right regulation because we have a very large population here in this state, right, and so that's important whether it's Texas or California or Florida,
New York. And you do see this kind of alignment of states who have already passed or are considering around this NIST model, the one that the e used use, the one that Colorado used around mitigating risk.
Okay, okay, and.
What about just there's other aspects to this too, which is trying to attract you know, while you're thinking about how you regulate it, you're also trying to attract companies and developers into this state. I mean, Joe Lonsdale is an example of that. You came from California two crisis. There has been a lot of activity recently. You saw Apple make an announcement of a development data center in Houston.
Facebook bringing their content moderation staff here. I mean yes, you know, particularly after Trump one the election, there is data centers exactly. There seems to be a lot of interest in bringing folks here. But that creates its own challenges, not just in terms of how you regulate it, but also how do you build.
Up the infrastructure to support it.
Let's talk a little bit about the power grid and how that ruts into this.
Yes, certainly we've heard a lot about the power grid in Texas over the past few years. We seem to everyone or be having some very very cold weather at times, and we know that we have our own grid. Now there's discussion about some connectivity with another grid, but we'll put that aside for the moment. Texas has its own
power grid. And it's not just power, it's water. So these data centers, whether they're beings for AI or crypto right, use an enormous amount of power and an enormous amount of water and they're noisy, right, And so there are concerns about the power grid and will there be the capacity and should there be requirements for some of these data centers to have their own source of power you know many nuclear or other sources that people are talking about.
And what does it mean also for water, we know that this is a major session for water and last session was too a lot of funding, but this is a really really major session for water and water supply and distribution. And how does that fit in with the data centers because they also use a lot of water?
Absolutely, absolutely, and then there's one more aspect that I want to talk about, which is how are we preparing you know, our students are our future workforce for an AI world. Is that a topic that is that is being discussed in the capital as well. Oh?
Absolutely, And it's a topic that I have spent a lot of time on the past years and continue to. You know, I think that learning begins, you know, the day you're born, and it ends the day you leave this earth, and that it's very important that we be cognizant starting in elementary school all the way through people's PhDs, that they have the appropriate curriculum related to AI. You know, is coding going away, Well, it's coding at a different level.
I think that it's very important that we focus on our students and the residents being continuous learners, having critical thinking skills, having people's skills, being able to write, and continuous learning, and that we look at jobs where they may need new trading right as we say, upskilling or reskilling, and that this be part of our curriculum in an appropriate manner throughout education and beyond, whether we're partnering with
community colleges or industry. And there are many that need within their own organizations to have this continuous training, whether you're a large city or a major nonprofit, or a big company or even a smaller company.
So I mean, I hear when people talk about sort of AI's impact on work in the you know, near future, they talk a lot about how, you know, AI can be a very very speedy and useful intern or entry level employee.
Right right, everybody's adding an assistant exactly.
You can you talk about coding it? Can it can write code when given direction, Yes, if you're if you're at a law firm, you know, it can maybe do some of that paralegal work. That's that early kind of work in sort of reviewing documents, legal system exactly what it's not quite there yet on is making big strategic decisions and everything like that. So how how does the education system adjust for that world?
One of the ways that the education system needs to adjust for that world is really focusing on the critical thinking skills of students. Right, So, you may be a financial analyst, but it's no longer there are no more Excel spreadsheets, right, you have AI, and you need those higher link level critical thinking skills to do the analysis. You may be a coder, but you're coding at a different level because AI is doing the we'll call it
fundamental coding. Right. You know, you may be a paralegal, but you are now needing to think at that higher level. And so that is really what is so important, those analytical skills, those critical thinking skills, being nimble. Also, there are various situations where AI does not and I don't
think ever will have the capability. And that's for instance, if empathy is required, right, if it's a situation with a doctor or a lawyer, or a teacher or an educator where you have to have those people skills, the empathy leadership within organizations.
Okay, and so I guess here's my last question. How confident do you feel in the state's ability to get this right this session? How this feels like a complicated issue. It feels like there's a lot of nuance, a lot of expertise needed. Is the Texas legislature equipped to end in a good place in early June?
Texas legislature has dealt, is dealing and will deal with in the future a lot of very complicated issues. And I think that the process is set up, you know, it's set up frankly, to kill bills, not to pass them. Say, visualize the Summer Olympics with those people who are running the hurdles one after another. So the system is set up really to vet bills, to get input, to have amendments, to have discussion, to get to a place where people feel comfortable.
All right, well, we will watch and see what happens. Indeed, thank you so much for being here. Thank you to Rob and Chris. A reminder the Texas Tribuunion membership program as participating in Amplify Austin this week. Please check out our website if you would like to donate and contribute. Eleanor and James will be back next week and we'll see you then