An Intelligence Approach to AI

With partisan-laden media takes and AI slop now flooding the internet, it's more difficult than ever to know what to believe. But time-tested frameworks used by intelligence officers can offer some guidance. On this episode of Straight from the Shoulder, special guest Rollie Flynn—former CIA Executive and President of The Arkin Group—shares her tips from the field on how to best differentiate truth from lies. 

Transcript

Julia (00:01)
Today, I'm excited to welcome Rollie Flynn, President of the Arkin Group and former CIA Executive, to our show. We're so happy to have you here, Rollie.

Rollie (00:11)
It's great to be here.

Julia (00:12)
I'm going to put you right in the hot seat on what is a very hot topic of discussion right now, which is artificial intelligence. While AI technology is reconfiguring workflows and processes in so many industries, with our own intelligence backgrounds in mind, today, I thought it would make sense to discuss AI from the Intel perspective. Intelligence officers and analysts know that our assessments are only as good as our sources.

And now it feels like we're being lied to or that we have to question everything that we're reading. With the AI-generated information that's taken over the internet and media these days, how are you approaching what's true and what's false?

Rollie (00:57)
A little like deja vu, because I spent about 30 years as an intelligence officer. I was out in the field running and meeting with human sources, but also did a little bit of analysis along the way. So, whenever you have a large volume of information and you're trying to assess whether it's true, whether it's valid, whether it's relevant to your needs, you want to ask a lot of questions. And the very first question I ask is, what's the source of this information? Where did it come from? Is the source reliable? Is the source's source reliable? Do they have natural access to this information? One of the things we dealt with when I was out in the field was what we called the walk-ins or the volunteers.

Julia (01:42)
Can you tell us a little bit more about what you mean by walk-ins or volunteers?

Rollie (01:47)
I think one way I talk about it is by mentioning a very famous volunteer, Adolf Tolkachev, who tried to volunteer at least five or six times when he lived in Moscow. He's called the billion-dollar spy. And the reason is that he saved our U.S. Navy roughly a billion dollars because of the information he provided. But he tried to volunteer several times, and we ignored him. We ignored him because we thought he was what we call a dangle, a provocation, put up to fool us by the Soviets. Then, at the time, it was the Soviets. And so when you have someone like a walk-in, a volunteer, and we call them walk-ins because they walk into US embassies, but they can also volunteer by bumping into you on the street.

The first thing you want to ask is what's that person's motivation? Are they making up information and trying to sell it to you to make money? This happens all the time; individuals will come up with an enticing story to try to sell it to you. And so, we develop a process whereby we assess the genuineness of the motivations and the veracity of what that individual is telling us.

Julia (03:00)
I'm curious here about the application for those of us today who are searching the internet, which might be sort of the equivalent of having a walk-in or a volunteer when we receive an information search result. Right now on Google, it has this AI summary and has a bunch of little paper clips that link back to the original source. What are we supposed to do? What are some of the techniques that you used in the field to differentiate a good, reliable source from some of the garbage?

Rollie (03:31)
Yeah, and there is a lot of garbage. First of all, you want to click on those paper clips and look at what the original source is that they're quoting. And then read that article or that report and see who their source is. One of the things we want to look for is corroboration. But we also want to be alert to the possibility of what we call circular information. In other words, information that's all sourced to the same place but is being reported multiple times, seemingly to reinforce and verify what you're hearing. But in fact, you're not, because just verifying what has already been reported from the same source.

Julia (04:12)
And I want to pick up on the theme of motivation that you mentioned briefly before, too, Rollie. Because in our media outlets today, there's so much bias, and there's so much motivation on behalf of the people sharing that information or creating that information. So I think there are applicable lessons here for trying to navigate and understand the motivation of the different sources that are coming into your findings.

How do we do that with a written piece?

Rollie (04:39)
Well, you want to look again deeply at the sources and see does the source have a point of view? Are they pushing a point of view? And always ask, what is the agenda of this particular person or organization that is the source for whatever information you're looking at. When I was in the field, one evening, it was Friday night, got an emergency call from one of my sources saying he needed to meet right away. And indeed he did have very important information. It was that he had spent the afternoon with a member of a terrorist organization. And they were planning a hit against a U.S. embassy officer that could happen imminently. 

Fortunately, I asked him a few more questions, including, well, tell me more about this meeting with the source. And it turned out they'd been drinking beer all afternoon, and the source was very drunk when he told him that. Doesn't mean that the information isn't true. It just means, hmm, he might have been exaggerating or embellishing or trying to impress a friend.

And in that case, we still took all the precautions just in case it was true.

Julia (05:48)
Think this also touches on another angle of how we assess information and the validity of that information, that is really in the human realm. And it's something that you and Jack Devine have both mentioned to me over the course of our work together. And that is this idea of a gut instinct.

Of this instinct you have as the operator or the analyst to kind of say, this doesn't feel right, this doesn't seem right. And I think we're in a tough spot right now because we want to trust those gut instincts, but it's harder to trust it when you're reading an article. And we have our own bias as a reader and as an operator. So how does gut instinct factor into today's internet searches?

Rollie (06:32)
Well, gut instinct is really important. And it's obviously much easier when you're with a person, because you can read their body language. You can compare how they are today with how you've seen them in the past, if something changed. When you're reading, again, I think you have to go back to the sources. And if you know a topic well, it's a little easier. If you've been following a certain topic for a long time, and you're reading an article or something that comes out of AI about it, you'll be able to spot errors or exaggerations a little more easily. But I think the answer is to find more information, and where you can go back to the original source. If someone's being quoted, look for the original source of that individual actually speaking. Ideally, if you can find a videotape of that person saying whatever is being reported. However, then you have to wonder, is it a deep fake as well? So gut counts for a lot, I think, but gut works better when you're more experienced and you know a topic well.

Julia (07:40)
That leads me into my next question about the quality of information, but also the consistency of information. You've highlighted for us the threat of circular reporting. You know, one thing gets said and it's repeated, and then it becomes truth or it becomes conventional wisdom. And it was never true in the first place. But sometimes when I've been dealing with liars, you know, they forget certain things they've told you, and they tell us a slightly different variation of the story the next time. So I'm wondering if there are any takeaways here for when we're reading articles, if we do notice some discrepancies, if we should maybe be paying more attention to those, or have kind of our spidey sense up for those.

Rollie (08:24)
I think we should. I do think we should. Although human nature being what it is, people who think they're telling the truth sometimes embellish unwittingly or change stories slightly. As time goes by, your memory fades. But absolutely. I agree.

Julia (08:41)
And just to finish up on this idea of conventional wisdom, something that Jack is very cautious of, and I know you are as well. In the intelligence world, our concept of conventional wisdom often leads us to make decisions or to understand a situation without a true deeper attention to what's happening in that one specific occasion. Can you think of any examples of that?

And then, how this phenomenon is playing out in our Google world.

Rollie (09:11)
We often tend to go with what we've already known and with the expectations we already know. I remember when I was coming up and I spent the beginning of my career in Africa, and at the time, fairly often there were coups d'etat, but you never wanted to be the person who cried wolf.

That also makes you move the other way, where you're more cautious about reporting something sensational that might happen. And I think we can point to the various intelligence failures that have happened over the years, whether it be Pearl Harbor or 9-11, where there was some information out there, but again, it was, as was popular to say, the failure to connect the dots and put it all together, which is much easier said than done when you have a huge volume of information and trying to make sense of it. In that case, AI might be a huge assist in that.

Julia (10:08)
Right, I guess it depends on the information again that AI is scraping, because it is conventional wisdom, because it is sort of taking this massive data set and trying to make one takeaway from it, right? It's not usually too nuanced, although it's getting much better.

Rollie (10:16)
And there's thought that people tend to believe that which they hear first, that which someone they know tells them, or that which they hear most often. And the thing you hear most often, you do tend, most people tend to believe it. And you really do have to question.

Julia (10:28)
Hmm.

And I think it's harder than ever because sources that we would call reliable sources or corroborated sources, historically, like certain media outlets that were very esteemed, now we don't trust them the same way. We don't trust our politicians the same way. It's just this shift where a lot of the onus of deciphering truth from fiction falls on the consumer.

Julia (11:04)
And I think it's one of the reasons people come to the Arkin Group to sort of help sift that, but it's also something we grapple with in our personal lives.

Rollie (11:1)
I think that's right. And our weekly 3+1 Report looks at key developments around the world in what I think is a very nonpartisan, analytical way. Analytical, but not boring.

Julia (11:26)
And on that note, it's not boring. It's never boring to speak with you, Rollie. And we thank you so much for joining us today on this show.

Rollie (11:34)
Thanks, Julia. It's great to chat with you.