The Penta Podcast Channel
The Penta Podcast Channel is home to all of Penta's podcasts, including the Macrocast and What's At Stake: A Penta Podcast. The Macrocast, produced by Penta and Markets Policy Partners, features weekly insight and analysis on the latest macroeconomic trends. What's at Stake: a Penta Podcast features in-house experts and often special guests for analysis on the biggest issues shaping business and public policy.
The Penta Podcast Channel
The future is here: How AI is reshaping business
On this week's episode of What's at Stake, Bryan DeAngelis, Head of Penta's D.C. office, is joined by Alberto Lopez Valenzuela and Andrea Christianson, co-leads of Penta's AI Task Force. Ahead of Senate Majority Leader Chuck Schumer's AI Insight Forum with tech leaders next week, the group takes a look at key considerations around AI governance, including the impact industry can have on responsible adoption and regulation.
As leaders in the field, Valenzuela and Christianson share insights on the practical applications of AI in providing strategic guidance to businesses. Tune in for a glimpse behind the scenes of how Penta is integrating AI into our products and strategic recommendations.
Welcome to another episode of what's At stake Appente podcast. I'm your host, brian D'Angelois, a partner here at Penta, and I'm joined today by two of my colleagues, alberto Lopez Valenzuela and Andrea Christensen, both of whom co-lead our AI task force. Alberto is our senior partner at Penta, sitting in our London offices, and he's the managing partner of Penta Insights. Andrea is a partner sits two doors down for me here in Washington. Both of you, thank you for joining us today on the show.
Speaker 2:Good to be here.
Speaker 1:We're going to jump into the hot topic of AI. We've talked about it a few times on the podcast. This is an area of policy and even capabilities that has really rapidly emerged as a transformative force. It's reshaping industries, economies, even the very fabric of our society. Maybe it's hard to believe it's been less than a year since really the arrival of chat. Gpt OpenAI's tool, which kicked off a lot of this interest in AI, made history as one of the fastest growing consumer apps of all time. I think it was something like a million users in just five days last November. It has really kind of kickstarted a conversation around AI's potential for businesses, some of the questions that are raised from there. Penta has put itself really at the forefront of this movement. We're helping guide business leaders through responsible adoption. We're determining how companies can make these tools work for their stakeholders. We're using a lot of these tools ourselves. There's a lot to talk about and a lot changing that. I'm excited to have the two of you here to help me break it down when we start. It was actually timely.
Speaker 1:As we were preparing for this podcast, Axios came out with a new survey that they did with the Generation Lab and Syracuse University's AI experts. It was focused on regulation and where should this sit in terms of who's going to oversee AI. As we think about this, it started to be very evident that these are questions that will steer how Congress thinks about it, how the White House thinks about it. We've already had folks on the show to talk about how the EU thinks about it. Just this week, we're going to see Senator Schumer kick off and continue some of his listening sessions on regulation. He's got experts coming in from Ewan Musk and Google CEOs, but even the AFL-CIO president and, of course, Sam Altman, open AI CEOs been doing a lot of work in Washington. Andrea, maybe I'll start with you, since we're starting in Washington. What are we expecting, especially as Congress comes back in the fall, in terms of legislative developments or conversations on AI?
Speaker 2:Yeah, thanks for the question. It's really interesting because these listening sessions, these forums that Senator Schumer is kicking off, in coordination with a bipartisan small group of senators as well, is really a different way than Congress has approached other issues in the past, and I think a couple of months ago, senator Schumer made a speech where he outlined his thinking on this and he acknowledged we have to prioritize innovation, but we have to do it in a safe way, and five minutes question and answer in a traditional congressional hearing isn't going to get us there, and so this is just a slightly different approach to, I think, deepening the knowledge of lawmakers of what are these technologies capable of now, what could they be capable of tomorrow, what are some of the limitations, what are some of the challenges and where are we going to control in bringing in opposing voices and really trying to get deep and granular on these issues. So, in terms of legislation actual regulation or bills I don't think we're quite there yet. There are bills right.
Speaker 1:For sure yeah.
Speaker 2:Out there right now, but the approach Schumer is taking is really one of deep understanding, so that we do this right.
Speaker 1:I'm going to skip the question of whether we need a federal Department of AI, because I don't think we're there yet, but that was the top choice in this survey. And I guess, maybe to spring a different question on you, are we confident Congress is the right place for this to start and where these discussions should be happening, or is it going to need to be broader than that?
Speaker 2:Well, I think it is broader. Right now, Congress is doing its work, but you also have the executive branch, the Biden administration, you have executive branch agencies being really thoughtful, putting out frameworks, getting voluntary commitments from private businesses, and you have the private businesses and the innovators really being at the forefront of having some of these conversations. So it is sort of happening in a lot of different areas, which is good. Right, it's a marketplace of ideas right now, but I think that Congress's thoughtful approach is smart because, ultimately, for something to be sustained over the long time which AI is here to stay it's going to need something that's sustained over the long term. In terms of a regulatory framework, it's probably best if it comes from Congress.
Speaker 1:Agreed, agreed. We'll need some laws that guide this at some point. How much do you expect industry collaboration to play into this? I mean, they're doing a lot, but it's also Quite a number of big names that have a lot of business, maybe good or bad, in front of Congress? How much are they going to be relied on or engaged with to help set policy going forward?
Speaker 2:Well, looking at kind of the guest list of these forums, I mean right now a lot, because they know a lot more about this product.
Speaker 2:I think that when it comes to questions of bias and how we mitigate concerns about bias, when it comes to questions of competition, which are big questions that Congress has been grappling with for a while, those are more complicated because the reality is it takes an immense amount of data processing to be able to create a large language model like the ones like Bard and ChatGBT.
Speaker 2:You're not going to have a ton of smaller entrants at that level. What you're going to have is a bunch of really great innovators, vc investments, all this kind of stuff as a part of the stack. So what are the apps that are going to be built on these large language models? And so I think we're going to see a lot of robust competition there, and so I think that it's going to be interesting, because lawmakers are going to have to rely on the companies who have created these products to truly understand them, to figure out a path to innovating or regulating them. But I think what we saw from Schumer's comments and his approach thus far is that most American lawmakers, republican or Democrat, very much want American leadership to stay. They want our companies to continue to be innovators and they want to foster that innovation and grow that leadership, and so I think that's good, but obviously they want to make sure that we're doing it in a way that's safe.
Speaker 1:Yes, and in addition to being safe, alberto, I want to bring you in here. American leadership is critical for American policymakers, but it strikes me that AI will no geographic boundaries right, that this will take a lot of collaboration. So I wonder if you could tell us a bit about where we are on AI regulation and the policy landscape in Europe.
Speaker 3:Sure and look, I'll be happy to answer the question, not as a policy expert, which I'm not, but as a keen observer of what I'm actually currently seeing in the regulatory space. Interestingly, as I was listening to a podcast a few days ago and it was the economist that interviewed the founder, the inventor, of the TCPIP protocol, which is basically the Internet protocol, and there was a question around regulation. What was really interesting is that I said look, historically the US approach to regulation has been to wait for something bad to happen and then regulate, and within the EU as always been about predicting or plan what might go wrong and regulate against those risks. So obviously, one approach leads to more innovation and also more risk. Another approach has a focus on mitigating risk from the outset and to a certain extent, it could actually tamper or impact in innovation. If we think about when it comes to the UK, the UK currently runs third for private investment into AI companies, following US and China. Which is saying that he's already mentioned that actually he's got ambitions for the UK to play a key role in AI safety and on their holding an AI safety summit in early November.
Speaker 3:So, in answer your question specifically, the UK and the EU are taking divergent approaches to AI, to AI regulation. The EU approach is a risk based model and basically classifies AI systems and they have basically four categories prohibited, high risk, limited risk and on all other systems. The EU is quite advanced on this, is on the trial phase and potentially could become the first comprehensive regulation on AI and you know they talk about the AI act following the steps of the GDP regulation and that other countries might follow. The UK is taking a very light, as I said, a different approach. They're taking a lighter approach, so they are not proposing legislation, but what they actually proposing is a flexible definition of AI systems and also they recognize that actually AI systems is a generic technology. What that means.
Speaker 2:Is that actually?
Speaker 3:what they are proposing is a principle based framework approach for existing regulators. So they are not thinking of creating a new regulator for the UK and fundamentally, they expect existing regulators to apply five principles, which is safety, security and robustness, appropriate transparency, explainability, fairness, accountability of the women and contestability and address. So it's really clear that the UK is trying to avoid a necessary red day and enable innovation, and I think what's the final point I want to make is what's really interesting is that the UK they are really keen to bring together regulators and innovators, For instance, the UK government. It followed the advice of the Patrick Ballance, who was the chief scientific advisor of the UK. He was a chief scientific officer at GSK and his advice was to create a regulatory sandbox to AI that would bring together innovators and regulators. So you can see EU is being quite rigid about it, whereas the UK is being much more working from the principle space and working with existing regulators.
Speaker 1:As you think about that more rigid approach that EU is taking in your mind. How do they, how will they balance that with kind of the international cooperation that I think will be needed? Particularly with the US and China being such for lack of a better term powerhouses in this industry.
Speaker 3:Look, I think the answer is that I can give you is that it remains to be seen. What's very clear is that not everybody is moving at the same pace. So, for instance, in China, in the middle of July, they probably it's an AI regulation which actually has taken effect on the 15th of August. Since then, china has actually approved 70 LLMs 70, and now they've floated the market with these LLMs, which actually it's clearly they're going to be to develop a flurry of applications. So I think China waited a little bit, but actually they've approved, as I said, they've approved these 70 LLMs, and I think that's pretty significant.
Speaker 3:The other thing I would say is that there are many moving pieces that I think we have to be mindful of, of course. Look, the reality is that AI has been around since ARPANET was created. Arpanet was the root precursor to the internet. So I think the magic of Genetic AI has actually captured people's imagination because actually they can do things that actually before they couldn't do. But that actually happens a lot at the consumer level.
Speaker 3:So I think that, from my conversations with corporates, they have been more cautious about how they're going to be implementing Genetic AI and how they actually going to use the existing data sets to develop productivity gains with the LLMs. One thing that, for instance, is really interesting is that I also saw a master class from Stanford University from Professor Andrew NG, who actually is one of the kind of the leading voices on the, which actually he mentioned and this is the point about speed. He actually mentioned that prior to the LLMs, the time from idea to implementation would take really about 6 to 12 months. From working with a supervised model yeah, with an LLM, the idea to implementation actually is reduced to a few weeks on thousands of dollars rather than millions. So you can imagine that actually, when organizations will be to be with corporates, they really understand how they can apply LLMs into solving their business problems. There's actually going to be a huge, huge benefit for those organizations.
Speaker 3:Yeah, but it's a different dynamic to chat GPT and another sort of applications out there which actually can be a big gimmicky. You know it catches people's imagination but actually the real it is in a way reminds me to the internet days you know the internet days. People used to talk about Bricson, morda and internet companies, and then everybody became an internet company.
Speaker 2:Yeah, I think everybody will be using everybody.
Speaker 3:Yeah, there was no distinction about LLNs or not. Either you were in the game or you were completely out of the game.
Speaker 1:Andrea, let me bring you back in here. We were talking a bit about, you know, the US approach. Alberto outlined the EU and the UK, and certainly China is the other big player here. In this space it's a lot for companies to keep up with. What are what are companies asking from us right now on the on the policy front? I mean, I know we're doing I've got a few clients that want to do a lot of monitoring and understanding what's changing but what are you seeing out there in the market in terms of what companies need?
Speaker 2:Yeah, I'd say for a lot of our clients right now they're really interested in what does the regulatory landscape look like, what are the big things that are upcoming, what are the big changes, what are the things that are going to affect their business and how they approach things? What is the collaboration between the US and the EU and other countries look like? Also, very much like, who are the big players here? Right, who are the stakeholders that they should be talking to, that they should be engaging, and, frankly, what are their own internal policies look like and what is their perspective and point of view? Some companies have very strong points of view already. Others are still developing them, and so I'd say that there is a lot of monitoring. And then, from the strategic advice standpoint, we're saying hey, here are the people that you should be thinking about engaging with and here are some of the ways that you could put your point of view and your perspective in front of decision makers, as they are developing their own perspectives on this. And so that was one of the reasons that we launched this Penta AI task force a couple of months ago that Alberto and I lead together and frankly not to be like overly marketing here why Penta is really uniquely qualified to be helping clients in this area, because we've been working in AI for over a decade.
Speaker 2:Alberto has built systems that have won awards, and so we're already processing like massive amounts of data to help companies better understand what their reputations are with which stakeholder groups, and at a really granular level.
Speaker 2:That can help them make better business decisions, and we're able to use LLMs in new ways to help us parse that data in really exciting ways.
Speaker 2:You know, one example has been a vetting for a client who wanted to understand if somebody was safe to do a partnership with, but they had a ton of YouTube videos and watching all of those videos would have taken hours, and we were able to use an LLM to create a system that enabled us to flag potential videos that should get a human watch, and so it was a process that took us several hours instead of several weeks, and so we're able to kind of answer questions we haven't been able to answer before faster than we would have been able to do it in the past, and then pair kind of our strategic expertise both in technology which we have a very strong bench of tech folks here at PENTA but also just how do you engage in a conversation like this, and then also with a lot of other communicators?
Speaker 2:How do we need to be thinking about the reputational and brand risks to your organization as it relates to AI, and so? We've had other podcasts on this, we've had papers on this, but that's a lot of the work that we're doing right now with our clients.
Speaker 1:Yeah, and the big theme that's jumping out of me there is speed. Right Back in the days of the tech fights, you had an industry that matured first and then Congress caught up with it and, to Alberto's point, maybe waited until something broke. Where everything's happening now, it's white space. You've got to be engaged in this. You don't have the time to get necessarily all your policies and duck in a row before Schumer's already have a meeting. So knowing where to engage, how to engage, is a lot of what I'm seeing our clients drive us towards too. Alberto Andrea, I mentioned a little bit about the task force and what we're doing, and mentioned some of the work you've been doing for years. Tell us a little bit more about how PENTA is already using AI as an embedded function and a lot of our intelligence tools and capabilities. Sure.
Speaker 3:So look, I mean, as Andrea says, ai is not new to us. In 2010, we pioneered the use of AI technology and supervised learning ie labeling technology to develop stakeholders sentiment, which would actually use to help our clients understand how they were being conceived by the stakeholders and also providing them a comprehensive graph of the issues that presented reputation of race to the companies and opportunities. So, right now, what we actually do? I'll answer your question first on what we are doing and then how we're helping our clients, but right now, what we actually do is we are reaching the data set that we developed over 12 years, and we developed that data set with supervised learning and we are reaching the data set with AI. What we already see is that actually, the improvements in terms of accuracy, autobie identification and entity sentiment are significant, and so significant that actually it's really giving us now the opportunity to think about other applications that we can build on a dataset that doesn't require so much supervision or human supervision.
Speaker 3:So what we have in the in the pipeline right now, it falls into two categories. One is reacting to the present, reacting to what's happening, and that's what crisis detection, crisis monitoring, the whole mapping, misinformation detection and even synthetic data detection, which synthetic data actually leads sometimes to misinformation. And then the other category we're focusing on is really helping our clients sort of shape in the future, and that's more about helping clients around message testing, scenario modeling and even forecasting repetition and impact with of different scenarios. So that's kind of at the high level. But what's very clear for us is that actually the fact that we have that existing existing dataset it actually give us an advantage over other organizations that they don't have that dataset, because actually when we develop our solutions we always develop them with the corporate first function in mind. Do we have a lot of accumulated learning that we can apply with any of Hans' dataset?
Speaker 1:Let me ask either one of you. I read a report this morning from McKinsey on how the most innovative companies are using AI and it struck me as something we've talked about a lot and similar to what we're doing and bear with me here. But the five actions they looked at is having that ability, knowing how to ask the right questions, knowing how to spot wrong answers fast, as you were saying, alberto continually building that proprietary data, those datasets, making sure the organizations can learn quickly and then creating that no human touch to workflows, which there'll always be humans, I think, in our part of the work, but there is the advantage of how do you remove some of the human element to get at a faster speed that generative AI offers and then bring it back in at the right time. How do you too, as the chairs of Art Task Force, kind of think about the way they're talking, about how companies innovate? And then, how is Penta actually staying ahead of the curve on a lot of this technology?
Speaker 3:Again, this was a point that was reinforced on the Stanford class that I saw, which was there's a lot of emphasis on the technology aspect, but also the sort of the domain expertise the sector domain expertise or the subject matter expertise.
Speaker 3:It's actually equally important Because actually when you combine technology with sector or with domain expertise, then you can actually apply the technology in a way that actually solves real problems in that particular domain. So I think that aspect is super important and we're very lucky at Penta that actually we have hundreds of people that they are experts in different domains, in a wide range of domains, and we also have experts on AI that have been doing this for a long time. So we are very lucky to combine that in the business. I also think that actually what's really important here is it's a culture of innovation, To actually have it's not only about the experts on the tech, but it's also a culture of innovation, a curiosity of to really ask the questions to our clients, to actually understand what is it that they want to achieve and actually ask those different questions. That by putting together the tech experts and the domain experts, we can actually do it in a more efficient way and hopefully we more accurate insight.
Speaker 2:Well, and I'll just add, I think what's really interesting is our clients help keep Penta innovative right, because they have these problems that they don't know how to solve and we're like, well, let's figure out how to solve them.
Speaker 1:Yeah, good point.
Speaker 2:And so we have to start asking questions. So, for instance, we pioneered something called text analysis, that we have an amazing media monitoring system, and when you think about how we've been able to use AI to increase the speed of that media monitoring from what was a human driven effort a couple of years ago, but we still have human oversight to make sure that it's still relevant and you're getting what you need. But if you're trying to understand how effective are surrogates, for instance, in all of the media that we're trying to go have them go and talk on behalf of us or talk about an issue, how can you know that? You would have to read every single news article, which isn't feasible, but what is is to okay. Well, what are all the names of your surrogates?
Speaker 2:We're gonna go find out how many times they were mentioned, in what kind of publications. Were they top tier, were they regional? Were they trade publications? What was it? And so we've been able to constantly be innovating, because our clients keep asking us to help them solve problems, and so that, I think, is really exciting, because it's forcing us to do all of these things that McKinsey lays out as being what are the most innovative companies and it's really wanting to solve problems for people, and so it's just really, really fun, I think, to be in that situation.
Speaker 1:Yeah, it's a great opportunity. I mean, the tools Alberto's talking about, the mindset you're talking about clients are. They're living in a rapidly changing world and being in front of those problems and helping them solve them is really what they're looking for in firms and agencies. So I'm gonna end this show with if you're out there and you've got a problem, you know to call Andrea and Alberto now, because they are on top of this with our AI technology and our problem solving capabilities.
Speaker 1:But both of you this was as with all AI topics this was a lot to jam in 30 minutes, but I appreciate you guys breaking it down for us and we'll certainly have a lot more opportunities, both on the show and Andrea mentioned some of the written content we do on our website around the work we're doing, the changes we're seeing coming all through the prism of AI. So thank you both so much for joining us today. We appreciate it.
Speaker 3:Thank you, thanks, brian.
Speaker 1:And to all our listeners remember to like and subscribe wherever you get your podcasts. You can follow us on Twitter now ex at Penta Group. At that's at PentaGRP, I'm your host, brian DeAngelis. As always. Thanks for listening to what's At State. We'll see you guys next time. Good to go to the choir.