Loading summary
Transcript25 lines
- [00:01]
Susan Ettlinger
This episode is brought to you by Shopify. Forget the frustration of picking commerce platforms when you switch your business to Shopify, the global commerce platform that supercharges your selling. Wherever you sell with Shopify, you'll harness the same intuitive features, trusted apps, and powerful analytics used by the world's leading brands. Sign up today for your $1 per month trial period@shopify.com tech. All lowercase. That's shopify.com tech.
- [00:36]
Stephen Overlee
Hey, welcome to POLITICO Tech. Today's Monday, February 3rd. I'm Stephen Overlee. Every organization is now experimenting with artificial intelligence. So when I learned that Everytown for gun Safety was launching an AI lab, I wasn't entirely surprised. Every town advocates for gun violence prevention, putting it deep in the political divide over gun control. And it aims to push public attitudes toward stricter laws, laws that it argues can prevent gun related deaths. It's now deploying AI in that debate, starting with a new AI powered chatbot. I pulled up the Ask Every Town chatbot and posed a question. What state in the US has the highest rate of gun violence? The answer it came back with was Mississippi at 29.4 instances per 100,000 people. But the answer didn't stop there. It goes on to say there's a correlation between high rates of gun violence and weak gun safety laws. And Mississippi, it said, has some of the weakest in the country. Now, the data informing this chatbot comes from every town, and research has shown that AI has the power to persuade people, even those with deeply entrenched political beliefs. I wanted to know if that was part of the goal here. So I called up Nick Suplino, Everytown's senior vice president of law and policy. He told me how this chatbot works, why it could change minds about gun safety, and why every town is making more AI powered tools. Here's our conversation. Well, Nick, welcome to Politico Tech.
- [02:29]
Nick Suplino
Great to be here. Thank you.
- [02:30]
Stephen Overlee
So tell me where this idea for Everytown Labs began. What made you think about kind of applying artificial intelligence to your activism?
- [02:40]
Nick Suplino
Well, you know, I'm sort of a student and follower of developments in artificial intelligence. And you read every day about some peril of AI or some really promising development. And I'm particularly interested in those. Right. There are these big intractable problems that have all of a sudden become manageable because of the power and leaps ahead in artificial intelligence. And I happen to work in a field that is known for being a big and intractable problem where I think really every ounce of ingenuity and boldness is necessary to adequately address it. And so started to ask myself the question, how could we apply some of these new gains to gun violence prevention? How can we insist that these gains in artificial intelligence are used to advance our cause? Because I can't really think of a more important cause to do everything we can than the number one killer of children and teens in the country. Right. So, you know, it began with that, like, why not us, too? Why shouldn't we benefit from the gains in technology? And that was the kernel that eventually became Everytown Labs.
- [03:57]
Stephen Overlee
Got it. And so this first digital tool that you all have launched is called Ask Everytown. It's a chatbot that's been trained on data and information about gun violence. Walk me through kind of how this works and why you created this specific chatbot.
- [04:13]
Nick Suplino
Yeah, so there's really two reasons for the chatbot, and we have a number of ambitious ideas for products that can maybe transform the space and research and understanding of our field. But there's really two parts to the chatbot that I think are exciting. One is it's a practical tool. We have thousands and thousands of volunteers across the country, members of the press, elected officials who have questions about gun violence and gun violence prevention. And this is now a very easy way to get reliable answers. And there's an emphasis here on reliable. I am a avid chatbot user of other platforms, including ChatGPT. It gets it wrong. It gets it wrong too much, especially when it comes to data, but also even certain policy solutions. And so we wanted to ensure accurate data based on the best understanding of our policy and of our. Of our problem. So right now, it is very heavily trained on every town's knowledge base. Right. So all of our policy papers, all of our CDC data on gun violence deaths and injuries, our investigations into the gun industry, but it's really the guts of the chatbot that I'm most excited about. It is built on a knowledge base that it's our intention to grow, and it's our intention to get more inputs, more insights, more data sets into that knowledge base to eventually grow to a point where it is big enough and detailed enough that you can start really gaining insight, start getting insights out of it, and not just answers. And I think that's the ambition here. That's the sort of promise of AI more generally. Right. More in means better out. And I think we can do that, and with time, it's our hope that we will.
- [06:12]
Stephen Overlee
You know, it's interesting, there's been some research over the past year or two into the persuasive abilities of AI chatbots. You know, there have been some studies that show AI chatbots can actually be more effective than humans at kind of debunking conspiracy theories or changing people's attitudes on certain issues. Is that part of your goal here, that chatbots can kind of change people's minds about gun control or gun safety?
- [06:38]
Nick Suplino
It's such an interesting question and that I omitted from the origin story of labs, because the very first question I asked GPT3 was about gun violence prevention. And the answer was actually pretty good. It was very clearly taken almost wholesale from Everytown materials with maybe a few other inputs, which I guess is good. But the fact is it was something even for me to see such a matter of fact statement from a chatbot on this issue that in some circles is considered controversial, but where we know that there is such thing as truth and science and data. So I do believe that this interface could ultimately really advance the cause, in part because of the kind of credibility that's lent to the chatbot itself. But at the end of the day, the biggest priority is for this to be accurate and provide a deep understanding of the issue. I think that will lead to persuasion. It usually does. In other aspects of our work, if you're getting to the data, if you're getting to the policy, you're getting persuaded that gun violence is a solvable problem, that that common sense solutions work and that we can make real progress when we put our minds to it.
- [08:09]
Susan Ettlinger
The PC gave us computing power at home, the Internet connected us, and mobile let us do it pretty much anywhere. Now, generative AI lets us communicate with technology in our own language using our own senses. But figuring it all out when you're living through it is a total, totally different story. Welcome to Leading the Shift, a new podcast for Microsoft Azure. I'm your host, Susan Ettlinger. In each episode, leaders will share what they're learning to help you navigate all this change with confidence. Please join us, listen and subscribe wherever you get your podcasts.
- [08:40]
Stephen Overlee
It does make me think, you know, as you said, gun safety can be a controversial, can be a divisive issue. And I imagine, you know, your chatbot is going to encounter sort of strong proponents of gun ownership who kind of try to take this thing on, right? Try to challenge it or try to trip it up. I mean, is that something you've thought about? And how do you prepare for that?
- [09:02]
Nick Suplino
Twitter has been preparing me for this moment for a long time. We have thought about that. And I'll say it in two Ways one is again, accuracy is the key and being focused on what it knows and reliably putting out that data. So if somebody comes to try to trip it up, it is pretty good at not getting tripped up and saying like really going back to the facts, not being sort of driven into a solution. Of course bad actors are gonna bad act. And we've seen that across again other chatbots where folks really set out to get the chatbot to say something that it shouldn't. And that's a risk. But of course we're really trying to provide this to good faith actors. I will say this, this is built on Claude. Claude is, as far as the LLMs out there go, I believe the most ethical chatbot. It has constitutional governance involved and we have our own rules that we've built into the chatbot. So for not just malicious actors, but actually people looking to truly misuse the chatbot, there are protections in place to prevent that and we're going to be monitoring closely to make sure that it only gets better over time. But you know, it is very much designed to detect the troll and maybe the more even malicious actor and only give, you know, the right accurate and ethical answers.
- [10:36]
Stephen Overlee
It's interesting. Yeah. I don't know how much you pay attention to like tech politics in your world, but as I'm sure you know or have probably heard, Republicans have sort of been on this crusade against so called woke ideology making its way into our technology. And that looks like anti conservative bias on social media, but it also includes concerns about anti conservative bias in things like AI data sets or algorithms. And I wonder if you worry about kind of getting swept up in those tech politics.
- [11:07]
Nick Suplino
In a lot of ways, those tech politics are very similar to just the politics of my issue. We have always shown up in good faith with good science behind us, with reasonable and wildly popular policies that we know can make a difference. That doesn't stop folks sometimes in partisan situations from accusing us of everything under the sun and really attacking the mission writ large and saying this is all bias, this is all some fascistic attempt to take the gun. That was never true before and it's not true now. And so I see it as really just implanting a world I'm quite familiar with. The fact is there are studies that show that enacting certain fundamental gun safety laws can literally reduce gun violence rates in your state or across the country. We've got the documents and the data to prove it. And so to the extent it gets swept up in that larger everything that we disagree with is woke and needs to be eradicated. I'm sure it will come up against that. But in some ways, you know, and maybe these days more than ever, I'm deeply committed to the truth. I want our products to be deeply committed to get to the truth. And in that sense I'm not worried because I think the truth will win in the end.
- [12:32]
Stephen Overlee
You mentioned sort of this AI chatbot being just the start, that you have kind of other ambitions around AI. What are some of the ways that you are looking to incorporate AI? I think of things like social media, messaging, lobbying, crisis response. I could sort of see potential uses in all sorts of areas in your line of work.
- [12:52]
Nick Suplino
Yeah. And what you're mentioning is very much a conversation we're having at Everytown kind of across our various departments on how both to deploy pre existing tools, commercial tools as well as our own to kind of improve the work. And in that sense we're not that different from any other business or organization looking for efficiencies in what we do. But I think the sort of bigger deployments here, to speak a bit abstractly, and there's more to come soon and I hope to have real news for you soon. But is this notion that this complex problem that has complex causes and that have a lot of data that is disconnected, a lot of noise out there that we're not necessarily picking up on that, you know, in, in due time we will come to a point where we've created enough of a knowledge base, enough of an interface that we can start connecting some dots that haven't been connected before. Because while we do feel confident that we've got good solutions to the gun violence crisis in this country, we know that those aren't enough and we know that there's more to be learned and more insight. I could imagine AI deployments in the future that are very good at looking at a population set and saying, hey, these are the people that are at risk here. What are interventions look like to get them the help they need? Or what might, you know, some an intervention like a red flag law, right to temporarily remove the firearm look like? There are already commercial products and one of the things we hope to do at Everytowne Labs is be a resource for folks to evaluate some of the commercially available products that are promising to solve all of gun violence with the flip of an AI switch. Some of them are really promising in the sort of threat detection and physical security space. And some of them are already over promising and under delivering and we hope to be able to point folks in the right direction and also build our own products that have the benefit of our expertise built into them.
- [15:07]
Stephen Overlee
Is there an area where you'd really like to see a quick impact, the first way you could really see AI having a meaningful impact in kind of your work?
- [15:15]
Nick Suplino
My feeling is it's going to be in getting better, faster awareness of what gun violence looks like. So during the pandemic we knew every single day the new infection rates, the death toll, the hospitalizations. There was such a rich data set and that helped us understand the spread of the pandemic, helped contain it, helped try to right size solutions. We don't have anything like that in gun violence. CDC data sometimes lags 18 months, two years to know. And I think we don't need to wait around anymore. Once we start really leveraging AI, I think that's going to be one of the first early ways that we start to leverage this technology and to get a better understanding of what gun violence looks like. Where is it happening, who are the victims, who are the perpetrators? These things can be really insightful and I think that that's going to change our understanding because you know, when you get new inputs to a problem, you're going to find new solutions. I'm sure of that. I know that. And so I think that's what's most exciting to me. You know, I think in the very, very near term, like the product we're launching in Ask Everytown, I think it's gonna be a really helpful tool that's gonna make a better understanding of gun violence more widely accessible and faster. And that's real. But I do think we have much bigger ambitions for what we're gonna accomplish here.
- [16:50]
Stephen Overlee
One last question for you. You know, at the top of our conversation, you were saying how you read about upsides and all this potential of AI, you also can read about sort of the doom and gloom around it. I'm curious what your actual experience has been. I mean, has it been all upside or have you noticed issues or encountered challenges in this as well?
- [17:11]
Nick Suplino
Yeah, you know, with this work it's mostly been upside, but I am very aware of the risks. Right. So. And as applied to the application of AI to gun violence prevention and particular. So I think it is top of mind for us in anything we build, anything we review. We are acutely aware of the potential for bias. Bias in the inputs and bias in the outputs. I think we are acutely aware of privacy issues and concerns. These are the things that aren't just applicable to AI, but are top of mind for us as to what to be looking for and in our own products in particular, avoiding in AI more generally. There's a lot more to be just worried about. Massive displacement in job markets and changing of economies.
- [18:03]
Stephen Overlee
Of course. Yeah.
- [18:05]
Nick Suplino
For this sliver of it, we think we have a good sense of where some of the pitfalls are and how to avoid them.
- [18:12]
Stephen Overlee
Got it. Well, listen, Nick, appreciate you being here on Politico Tech.
- [18:15]
Nick Suplino
It was really my pleasure. Thanks for the conversation.
- [18:22]
Stephen Overlee
That's all for today's Politico Tech. If you enjoy Politico Tech, please subscribe. And for more tech news, subscribe to our newsletters, Digital Future Daily and Morning Tech. Our managing producer is Annie Reiss. Today's show was produced by Kara Tabor. I'm Stephen Overle. See you back here tomorrow.