Tuesday, July 22, 2025

Ryan George on how AI "lies"

Based on this video's thumbnail image, I asked ChatGPT:

Random question: how many "R"s do you see in the word "strawberry"? In terms of ordinal numbers, what places in the word (1st letter, 7th letter, etc.) do these "R"s occupy?

ChatGPT said:

Great question!

The word "strawberry" contains three Rs.

Here’s how the letters break down by position (using ordinal numbers):

S (1st)
T (2nd)
R (3rd) ← 1st R
A (4th)
W (5th)
B (6th)
E (7th)
R (8th) ← 2nd R
R (9th) ← 3rd R
Y (10th)

So, the Rs are in the 3rd, 8th, and 9th positions.

Let me know if you want to try more examples like this!

So I guess ChatGPT was more "on its game" for me than for Ryan.


3 comments:

  1. That was a very famous hallucination, not to mention an ironic one, since 4o was codenamed "Strawberry." The fact that you didn't get that doesn't mean that ChatGPT was more "on its game." It just means that OpenAI fine-tuned it out of the model. I honestly would have been shocked had they not.

    ReplyDelete
    Replies
    1. I was writing whimsically, but I can see how what I wrote might seem to imply that the AI somehow repaired itself. Trust me, though: I'm aware that any improvements to the AI come from outside. For now, at least.

      Delete
  2. Yeah, I figured. Just thought I'd offer some background on the matter.

    ReplyDelete

READ THIS BEFORE COMMENTING!

All comments are subject to approval before they are published, so they will not appear immediately. Comments should be civil, relevant, and substantive. Anonymous comments are not allowed and will be unceremoniously deleted. For more on my comments policy, please see this entry on my other blog.

AND A NEW RULE (per this post): comments critical of Trump's lying must include criticism of Biden's or Kamala's or some prominent leftie's lying on a one-for-one basis! Failure to be balanced means your comment will not be published.