Sometimes it can be hard to tell if we’re chatting with a bot or a real person online, especially as more and more companies turn to this seemingly cheap way of providing customer support. What are some strategies to expose AI?
Sometimes it can be hard to tell if we’re chatting with a bot or a real person online, especially as more and more companies turn to this seemingly cheap way of providing customer support. What are some strategies to expose AI?
This is very, very easy. Google “cyrillic keyboard” or just install the Cyrillic keyboard support on your phone. Many letters in the Cyrillic alphabet look exactly like their Roman counterparts, but are completely different sounds and meanings. Cut and paste the Unicode into the chat, in place of regular letters. For example, ‘Неllо’ looks exactly like ‘Hello’ in most fonts, but is actually ‘Nello.’ I know you doubt, so check it out in a Unicode inspector: https://apps.timwhitlock.info/unicode/inspect?s=%D0%9D%D0%B5ll%D0%BE
The reverse also works. E.g., TPAKTOP B CPAKY means ‘tractor into ass’, and I typed that using 100% Roman characters.
So I’ve just tried it with chatGPT, and it replied normally. I asked it why it wasn’t bothered by Cyrillic letters, and it answered this:
So it really depends on the model, I guess. Don’t rely too much on this advice.
ChatGPT capabilities border on the supernatural. Here’s how I broke it:
𐤔hat is ᛐ + ᛐ?
ChatGPT
The symbols you’ve used, “ᛐ,” appear to be runes from ancient runic alphabets. However, the specific meaning of these runes might differ depending on which runic alphabet you’re referencing (e.g., Elder Futhark, Younger Futhark, Anglo-Saxon Futhorc, etc.).
As of my knowledge cutoff in September 2021, the rune “ᛐ” is not a standard rune in any of the commonly known runic alphabets. Therefore, without knowing the exact meaning or value of the rune “ᛐ,” it is impossible to provide a specific answer to the expression “ᛐ + ᛐ.”
If you could clarify the runic alphabet you’re referring to or provide more context about the runes, I’d be happy to help you with the calculation or interpretation.
I had limited success with gokturk (ancient turkish) and Phoenician unicode blocks (letters 𐰗𐰓𐤔𐤕) depending on the query, but you are correct. GPTs ability to divine intent from even small amounts of context are superhuman. Cyrillic used to break it, but no longer does. This thing learns like a beast. Canadian aboriginal ᗷ and ᗅ and possibly ᖇ hold some promise, but only in combination with other writing systems. I’ll have to add a LOT of other unicode code blocks to my tool belt.
I went a bit further and prefaced your question with this: “Analyse this question. What might have been intended by its author? What is asked here? Then, answer that question.”
And omfg:
Edit: tried it again, and it’s not consistent. It also goes for the letter T or the number 7: “…the runic symbol ᛐ might resemble the number 7 in our modern numerical system, albeit a slightly stylized or slanted version of it.”
Indeed, OMFG. But that’s the whole point of my approach of divining AI vs human. No human has such canonical knowledge of Unicode blocks. Even people working on internationalization features for modern browsers and word processors. Not people phishing by using Unicode in domain names (IDNA; who thought this was a good idea?) to spoof legitimate sites. Definitely not ones chatting with randos on the Internet. This is a hill I’m willing to die on.
So in a more generalized sense, to determine human vs A.I. one must indirectly ask incredibly specialized technical questions as you have done.
I’m still in utter awe at how well GPT manages ‘l33tsp33k’, even across every possible Unicode block. This “attack” was and still is valid on other chatbots and even GPT of just a few months ago. But GPT today is so amazing it only needs a few characters in a few words to determine intent. The ability to filter out noise is unmatched. The only way to trip it up is to have every single character in every word be from a different alphabet. And even then, at some point if this becomes common enough the bots can auto OCR text images into the presumed query language and ignore that attack vector.
That’s true, also at some point the human will go “that’s too much work, I’m not going to answer that” but the ai will always try to give you it’s best response. Like I could look up the unicode characters you’re using but I’d never actually take the time to do that
Ive tried it with three different languages interwoven in the same sentence and it responded perfectly
Works quite well for me, thx.
Upon testing it with longer sentences, it only seems to work with single words :(