Tobold's Blog
Saturday, May 18, 2024
AI comment spam

My previous post about the economic gameplay of Crusader Kings 3 compared to Victoria 3 received two interesting comments:
Victoria 3 is a masterpiece in grand strategy, brilliantly capturing the economic dynamism of the industrial revolution. The exponential GDP growth and the strategic reinvestment truly reflect the historical wealth accumulation of that era. On the other hand, Crusader Kings 3 offers a fascinating look into a slower, more turbulent period, where economic growth is modest and more reflective of historical realities. The contrast between these two games highlights Paradox's commitment to historical accuracy and diverse gameplay experiences.
Great analysis! I appreciate your deep dive into the economic systems of both Victoria 3 and Crusader Kings 3. Your points about the impact of compound interest and the differences in historical economic growth are spot on. Your strategy to play a non-expansionist count with a "lucky inheritance" sounds like a creative way to enjoy the game on your terms. Thanks for sharing your insights and strategy!
Both then revealed themselves to be actually spam comments, by adding links to some dodgy site. Now my site has received thousands of spam comments over the last two decades. But these two were way more sophisticated than anything I have previously seen. Both paragraphs make it appear as if the commenter has read my post. But I am pretty certain that it wasn't a human being, because that would have been way too much work for a spam comment. Instead, I am pretty certain that both these paragraphs have been written by AI, some typical large language model. This is exactly the level of AI we all have free access to now. The ability to summarize a text in a paragraph is what software like ChatGPT is great at.

Although I have spam filters on my comments (and I do apologize when sometimes legit comments get stuck in that filter and I only see that days later), these sophisticated AI spam comments were undetected by the the filter. To me, the attempt to link to some commercial site was a clear indication that this was in fact spam, but the spam filters apparently are only apt to catch the previous generation of spam, which was a lot more primitive.

AI is already being used to make phishing e-mails look more genuine. And now it apparently has reached comment spam on sites like Blogspot and LinkedIn. O brave new world!

And thus the great enshittification of the internet continues apace. I am convinced that an increasing number of clickbait web pages are also written by AI. They are particularly insidious because they initially read like real article and then you realise that you have read three paragraphs and it hasn't said anything of substance. My Google feed is full of crap like this.
I rarely look in my spam filter. Maybe once or twice a year. Occasionally there's a week when a bunch of stuff from the same source manages to get through, usually the same comment on multiple posts.

Obviously, they always have a link attached. It seems to me that a setting on the blog itself that simply blocked all hyperlinks in comments would be more of a deterrent. I ought to see if Blogger has one.

Just like with viruses and spam, our future will consist of the neverending arms race between malicious AI and counter-AI to detect and deal with it. The day is coming where "mere humans" won't be able to tell the difference.
It's unsettling to see how AI-generated spam comments are becoming increasingly sophisticated, as evidenced by your recent experience. The fact that these comments appeared genuinely relevant to your post is a testament to the capabilities of current AI technology, particularly language models like ChatGPT.

Your observation about the potential implications of AI-generated spam extending beyond emails to platforms like Blogspot and LinkedIn is indeed concerning. It underscores the need for more advanced spam filters and cybersecurity measures to counteract these evolving tactics effectively.

As you rightly point out, the ability of AI to summarize text and mimic human communication patterns poses a significant challenge for traditional spam detection systems. It's a reminder of the ongoing arms race between AI-driven deception and our efforts to safeguard online spaces from such malicious activities.

Thank you for sharing your insights and raising awareness of this issue. It's crucial for us as users and platform administrators to remain vigilant and adapt our defenses accordingly in this brave new world of AI-driven spam.
Wait, just what are you doing Dave?
Post a Comment

<< Home
Newer›  ‹Older

  Powered by Blogger   Free Page Rank Tool