{"id":290962,"date":"2023-05-13T16:25:51","date_gmt":"2023-05-13T20:25:51","guid":{"rendered":"https:\/\/knowtechie.com\/?p=290962"},"modified":"2023-09-20T14:36:30","modified_gmt":"2023-09-20T18:36:30","slug":"ai-generated-spam-may-soon-flood-your-inbox-with-scams","status":"publish","type":"post","link":"https:\/\/knowtechie.com\/ai-generated-spam-may-soon-flood-your-inbox-with-scams\/","title":{"rendered":"AI-generated\u00a0spam may soon flood your inbox with scams"},"content":{"rendered":"

Each day, messages from Nigerian princes, peddlers of wonder drugs, and promoters of can\u2019t-miss investments choke email inboxes. <\/p>\n\n\n\n

Improvements to spam filters only seem to inspire new techniques to break through the protections.<\/p>\n\n\n\n

Now, the arms race between spam blockers and spam senders is about to escalate with the emergence of a new weapon: generative artificial intelligence. <\/p>\n\n\n\n

With recent advances in AI made famous by ChatGPT, spammers could have new tools to evade filters, grab people\u2019s attention and convince them to click, buy or give up personal information.<\/p>\n\n\n\n

As director of the Advancing Human and Machine Reasoning lab at the University of South Florida, I research<\/a> the intersection of artificial intelligence, natural language processing, and human reasoning. <\/p>\n\n\n\n

I have studied how AI can learn people’s individual preferences, beliefs, and personality quirks.<\/p>\n\n\n\n

This can be used to understand better how to interact with people, help them learn, or provide helpful suggestions. <\/p>\n\n\n\n

But this also means you should brace for smarter spam that knows your weak spots \u2013 and can use them against you.<\/p>\n\n\n\n

Spam, spam, spam<\/h2>\n\n\n\n
\"new
Image: Unsplash<\/figcaption><\/figure>\n\n\n\n

So, what is spam? Spam is defined as unsolicited commercial emails sent by an<\/a> unknown entity. The term is sometimes extended to text messages, direct messages on social media, and fake reviews on products<\/a>. <\/p>\n\n\n\n

Spammers want to nudge you toward action: buying something, clicking on phishing links, installing malware, or changing views.<\/p>\n\n\n\n

Spam is profitable. One email blast can make $1,000 in only a few hours<\/a>, costing spammers only a few dollars \u2013 excluding initial setup. An online pharmaceutical spam campaign might generate around $7,000 per day<\/a>.<\/p>\n\n\n\n

Legitimate advertisers also want to nudge you to action \u2013 buying their products, taking their surveys, and signing up for newsletters. <\/p>\n\n\n\n

Still, whereas a marketer email may link to an established company website and contain an unsubscribe option in accordance with federal regulations<\/a>, a spam email may not.<\/p>\n\n\n\n

Spammers also lack access to mailing lists that users signed up for. Instead, spammers utilize counter-intuitive strategies such as the \u201cNigerian prince<\/a>\u201d scam. <\/p>\n\n\n\n

A Nigerian prince claims to need your help to unlock an absurd amount of money, promising to reward you nicely. <\/p>\n\n\n\n

Savvy digital natives immediately dismiss such pleas, but the absurdity of the request may actually select for na\u00efvet\u00e9 or advanced age<\/a>, filtering for those most likely to fall for the scams.<\/p>\n\n\n\n

Advances in AI, however, mean spammers might not have to rely on such hit-or-miss approaches. <\/p>\n\n\n\n

AI could allow them to target individuals and make their messages more persuasive based on easily accessible information, such as social media posts.<\/p>\n\n\n\n

Future of spam<\/h2>\n\n\n\n
\"ChatGPT
Image: Pexels<\/figcaption><\/figure>\n\n\n\n

Chances are you\u2019ve heard about the advances in generative large language models like ChatGPT<\/a>. <\/p>\n\n\n\n

The task these generative LLMs perform is deceptively simple: given a text sequence, predict which token \u2013 think of this as a part of a word \u2013 comes next. <\/p>\n\n\n\n

Then, predict which token comes after that. And so on, over and over.<\/p>\n\n\n\n

Somehow, training on that task alone, when done with enough text on a large enough LLM, seems to be enough to imbue these models with the ability to perform surprisingly well on a lot of other tasks<\/a>.<\/p>\n\n\n\n

Multiple ways to use the technology have already emerged, showcasing the technology\u2019s ability to adapt to, and learn about, individuals quickly. <\/p>\n\n\n\n

For example, LLMs can write full emails in your writing style, given only a few examples of how you write. And there\u2019s the classic example \u2013 now over a decade old \u2013 of Target figuring out a customer was pregnant before her father knew<\/a>.<\/p>\n\n\n\n

Spammers and marketers alike<\/a> would benefit from being able to predict more about individuals with less data. <\/p>\n\n\n\n

Given your LinkedIn page, a few posts, and a profile image or two, LLM-armed spammers might make reasonably accurate guesses about your political leanings, marital status, or life priorities.<\/p>\n\n\n\n

Our research showed that LLMs could be used to predict which word an individual will say next with a degree of accuracy far surpassing other AI approaches<\/a> in a word-generation task called the semantic fluency task<\/a>. <\/p>\n\n\n\n

We also showed that LLMs can take certain types of questions from tests of reasoning abilities and predict how people will respond to that question<\/a>. <\/p>\n\n\n\n

This suggests that LLMs already have some knowledge of what typical human reasoning ability looks like.<\/p>\n\n\n\n

If spammers make it past initial filters and get you to read an email, click a link, or even engage in conversation, their ability to apply customized persuasion increases dramatically<\/a>. <\/p>\n\n\n\n

Here again, LLMs can change the game. Early results suggest that LLMs can be used to argue persuasively on topics ranging from politics<\/a> to public health policy<\/a>.<\/p>\n\n\n\n

Good for the gander<\/h2>\n\n\n\n
\"man
Image: Unsplash<\/figcaption><\/figure>\n\n\n\n

AI, however, doesn\u2019t favor one side or the other. Spam filters should also benefit from AI spam advances, allowing them to erect new barriers to unwanted emails.<\/p>\n\n\n\n

Spammers often try to trick filters with special characters, misspelled words, or hidden text<\/a>, relying on the human propensity to forgive small text anomalies \u2013 for example, \u201cc1\u00eeck h.ere n0w.\u201d <\/p>\n\n\n\n

But as AI better understands spam messages, filters could better identify and block unwanted spam. And maybe even let through wanted spam, such as marketing email you\u2019ve explicitly signed up for. <\/p>\n\n\n\n

Imagine a filter that predicts whether you\u2019d want to read an email before you even read it.<\/p>\n\n\n\n

Despite growing concerns about AI \u2013 as evidenced by Tesla, SpaceX, and Twitter CEO Elon Musk, Apple founder Steve Wozniak and other tech leaders calling for a pause<\/a> in AI development \u2013 a lot of good could come from advances in the technology. <\/p>\n\n\n\n

AI can help us understand<\/a> how weaknesses in human reasoning might be exploited by bad actors and develop ways to counter malevolent activities.<\/p>\n\n\n\n

All new technologies can result in both wonder and danger. The difference lies in who creates and controls the tools and how they are used.<\/p>\n\n\n\n

This article was updated to indicate that it was a teenager\u2019s father<\/em>.<\/p>\n\n\n\n

Have any thoughts on this? Drop us a line below in the comments, or carry the discussion over to our Twitter<\/a> or Facebook<\/a>.<\/em><\/p>\n\n\n\n

Editors’ Recommendations:<\/h4>\n\n\n\n