Don't Be A Victim of AI - Victor Menasce

The Real Estate Espresso Podcast

Don’t Be A Victim of AI

Spotify Google Podcast Apple Podcast Subscribe via RSS Leave a Review

Don’t Be A Victim of AI

Welcome to the Real Estate Espresso Podcast, your morning shot on what’s new in the world of real estate investing. I’m your host, Victor Menasce.

On today’s show we’re talking about how to spot when you’re on the receiving end of content that’s been generated by artificial intelligence.

But first, I’d like to remind you that we have only a few weeks remaining in the calendar year. If you sold an asset this year, maybe you just earned a lot of money, you might have a bunch of tax owing next year. Buying an asset might be the solution. Bonus depreciation is a powerful tool; it doesn’t apply equally to everyone.

We’re hosting a webinar tomorrow on December 4 at 11 a.m. Eastern. We’ll keep it under 30 minutes. The purpose is not to provide you tax advice because we’re not accountants. The goal is to share enough information to stimulate a conversation with your own tax advisors. If you can’t make the 11 a.m. time slot, click on the registration link anyway and we’ll send you the recording. The link will be in the show notes and we’ll see you on Thursday morning.

On today’s show, we’re talking about how to spot when you’re on the receiving end of content that’s been created by artificial intelligence. AI tools have the potential to transform and accelerate virtually every aspect of our modern existence. This also applies to real estate investing. But my observation is that most people want the benefits of using AI as a tool, they just don’t want to be the victim of AI-generated advertising, AI-generated content, AI-generated music, and so on. People crave authenticity.

There was a period of time when both Google Gemini and OpenAI’s ChatGPT offered tools that were designed to spot content that had been generated by AI. These tools generated far too many false positives to be useful and, as far as I know, they’ve been discontinued.

People at Wikipedia face this problem literally with the millions of updates each day. They’ve been struggling now for several years to moderate submissions. Here too, the entire Wikipedia platform is facing the risk of being rendered irrelevant by the pollution associated with AI content, so the people at Wikipedia generated their own handbook designed to help editors spot AI-generated content. And that’s what I want to share with you today.

So while there’s no hard and fast rule, there are certain styles of communication that are frequently present in AI-generated content. These language models are usually trained on data from the internet, in which famous people are generally described with positive, important-sounding language, and as a result the language model tends to omit specific, unusual and nuanced facts. They replace them with more generic positive descriptions which are statistically common.

So for example, the highly specific inventor of the first train-coupling device might have become a revolutionary titan of industry. It’s like shouting louder and louder than a uniquely important individual. It’s like the portrait itself is fading from a sharp photograph into a blurry generic sketch. The subject becomes simultaneously less specific and more exaggerated.

And that statistical regression to the mean, or smoothing over of specific facts into generic statements, could equally apply to many AI-generated topics. It also makes it easier to detect.

So what we often see in AI-generated content is undue emphasis on symbolism, legacy, and importance. These language models often puff up the importance of the subject matter by adding statements about how arbitrary aspects of the topic represent or contribute to a broader topic. There is a distinct and easily identifiable repertoire of ways that it writes these statements.

Here’s an example. The Statistical Institute of Catalonia—that’s a province in Spain—was officially established in 1989, marking a pivotal moment in the evolution of regional statistics in Spain. It’s that phrase, “marking a pivotal moment,” that is an embellishment beyond the significance of what the event really was. It says, the founding of IDESCAT represented a significant shift towards regional statistical independence, enabling Catalonia to develop a statistical system tailored to its unique socio-economic context. This initiative was part of a broader movement across Spain to decentralize administrative functions and enhance regional governance.

It’s that embellishment that is one of the giveaway signs that this was AI-generated. Now, fact-checking usually identifies the gaps, but most people don’t go through the extra effort of fact-checking and are misled. And this is one of the numerous shortcomings of AI-generated content.

I’ll give you a couple more examples. For example, when you’re talking about biology, language models tend to overemphasize connections between the broader ecosystem and the environment, even when those connections are tenuous or they don’t exist. These language models tend to belabour the species’ conservation status and research and preservation efforts, even if the status is unknown and no serious efforts exist.

Here’s an example: it plays a role in the ecosystem and contributes to Hawaii’s rich cultural heritage. Preserving this endemic species is vital not only for ecological diversity but for sustaining the cultural traditions connected with Hawaii’s native flora. There’s just a lot of embellishment here.

Next, you’ll often see emphasis on notability, attribution and media coverage. So for example, you want to watch for words like “independent coverage,” “local,” “regional” or “national coverage.” These language models act as if the best way to prove that a subject is notable is to hit readers over the head with claims of notability, often by listing sources the subject’s been covered in. That may or may not provide additional context as to what the sources have actually said about the subject, and often they inaccurately attribute their own superficial analysis to the source.

Human-written press releases have, of course, also cited news clippings for decades, but language models specifically asked to write a Wikipedia article often echo the exact wording of Wikipedia’s guidelines like “independent coverage.”

So for example, a subject might have spoken on CNN or been featured in Vogue magazine, Wired magazine, Toronto Star, and other media. These references are often superfluous and sometimes even downright inaccurate.

These AI chatbots tend to insert superficial analysis information, often in relation to its significance, recognition, or impact. As you get more and more practiced at spotting these trends, you’ll notice that they’re kind of everywhere.

While many of these words are strong hints of artificial intelligence on their own, they’re an even stronger indicator when the subjects are facts, events or other inanimate objects. For example, a person can highlight or emphasize something, but a fact cannot. Highlighting or underscoring is not something that is actually happening. It can’t be claimed by a disembodied narrator about what something means.

So why am I telling you all this? This is, after all, a real estate investing podcast. We are introducing victor.ai as a tool, and part of our pledge to you, the listener, is that we’re making it available. People want to use AI tools to enhance their productivity, but they don’t want to be the victims of AI-generated content, which is seen as having lower value and impact than something genuinely created through a thoughtful process.

That doesn’t mean AI has no value. Far from it. It means that AI is a tool and, like any tool, it forms part of a process.

As you think about that, have an awesome rest of your day. Go make some great things happen. We’ll talk to you again tomorrow.

Stay connected and discover more about my work in real estate and by visiting and following me on various platforms: