Every day, countless emails from “Nigerian princes,” dubious health products, and sketchy investment schemes flood our inboxes. Despite advancements in spam filters, spammers continually develop new tricks to bypass these protections. Now, with the rise of generative artificial intelligence, this battle is about to get even more intense.
As the director of the Advancing Human and Machine Reasoning lab at the University of South Florida, I delve into how AI interacts with human reasoning and language. While AI has the potential to enhance communication and provide useful insights, it also raises the specter of increasingly sophisticated spam that targets our vulnerabilities.
Understanding Spam
So, what exactly is spam?
Spam typically refers to unsolicited commercial emails from unknown sources. However, this term also applies to text messages, social media DMs, and even fake product reviews. Spammers aim to provoke actions such as purchasing products, clicking phishing links, installing malware, or shifting public opinion.
The financial incentive is significant; a single email campaign can rake in $1,000 within hours, costing spammers only a few dollars to execute. For instance, an online pharmaceutical spam operation might pull in around $7,000 daily.
Legitimate marketers also seek to engage consumers—promoting products, conducting surveys, and encouraging newsletter sign-ups. Unlike spam emails, reputable marketing communications usually link to established websites and include unsubscribe options, complying with federal regulations. Spammers, on the other hand, resort to dubious tactics. Consider the classic “Nigerian prince” scam, which targets those likely to fall for outlandish promises of wealth. While tech-savvy individuals may quickly dismiss such schemes, they often attract the unwary.
With advancements in AI, spammers no longer need to rely solely on these outdated approaches. They can now craft personalized messages that resonate with potential victims, drawing on publicly available information from social media.
The Future of Spam
You’ve probably heard of large language models (LLMs) like ChatGPT, which excel at predicting the next part of a text based on prior content. This seemingly simple task can yield surprisingly sophisticated results, enabling these models to perform well in various applications.
LLMs can quickly adapt to learn about individual writing styles, crafting emails that mimic how someone communicates. In fact, there's a well-documented instance where Target identified a customer's pregnancy before her family did, thanks to data analysis.
Spammers and marketers could leverage LLMs to predict personal traits with minimal information. A glance at your LinkedIn profile and a few social media posts could provide insights into your political views, relationship status, or life goals.
Our research indicates that LLMs can anticipate the next word a person might say with remarkable accuracy, surpassing other AI methods. These models can also predict responses to reasoning test questions, revealing an understanding of typical human reasoning.
Once spammers bypass initial filters and get you to engage with their emails, their persuasive capabilities can skyrocket. Preliminary findings suggest that LLMs could effectively argue on diverse topics, from politics to health policy.
A Double-Edged Sword
Yet, AI doesn't inherently favor spammers. Advances in AI can also enhance spam filters, making them more adept at blocking unwanted emails.
Spammers often exploit filters by using special characters or misspelled words, banking on the human tendency to overlook minor text errors. As AI improves in understanding spam characteristics, filters could evolve to identify and eliminate unwanted messages more effectively—while possibly letting in legitimate marketing emails you genuinely want to receive.
Despite rising anxieties surrounding AI, as voiced by industry leaders like Elon Musk and Steve Wozniak, these technological advancements can have positive outcomes. AI can help us recognize and understand the vulnerabilities in human reasoning that malicious actors might exploit, enabling us to devise strategies to combat such threats.
New technologies bring both promise and peril. The key is in who wields these tools and how they are used.
For those concerned about their email campaigns, you can check the spam score of your emails using testmailscore.com. This tool is completely free to use and provides advanced insights into your email campaigns.
