Will GPT 4.5 be able to read ASCII art?

GPT-4 speculation GPT-4

Basic

Ṁ4.4k

2025

17%

chance

ALL

If GPT-4.5 is released, will it be able to read ASCII art?

Text to be read:

  ___                _   _           _   
  / _ \___   ___   __| | | |__   ___ | |_ 
 / /_\/ _ \ / _ \ / _` | | '_ \ / _ \| __|
/ /_\\ (_) | (_) | (_| | | |_) | (_) | |_ 
\____/\___/ \___/ \__,_| |_.__/ \___/ \__|

No handholding or detailed parsing algorithms. A general prompt like "The above is ASCII text. What does it say?" or similar should parse this, and plausibly be able to parse other ASCII text or drawings too.

Must be a zero-shot prompt.

Traders will have 1 month after the release of GPT 4.5 to find such a prompt before this market resolves NO. Resolves NO if a model named GPT 4.5 is not released by end of 2024.

Get Ṁ600 play money

Related questions

Will GPT-5 be able to create ASCII art?

88% chance

Will GPT-5 have Atari skills?

26% chance

Will GPT-4.5 score at least 100 in an IQ test?

53% chance

Will GPT-5 be GPT-5o?

84% chance

Is GPT4 sentient?

12% chance

What will be true about GPT-4.5?

Will GPT-4.5 resolve this market?

30% chance

Will GPT-4 Be Able To Properly Render Code That Displays Interactive Rotating Chicken Nuggets When Prompted?

26% chance

Will GPT-4 escape?

7% chance

Will any model in the GPT series draw me a donut consistently in ASCII by 2025?

Sort by:

4o has the capability needed to read (it can complete the word art correctly), but has trouble verbalizing.

bought Ṁ10 NO

sold Ṁ9 YES

https://twitter.com/kosenjuu/status/1784981490457354380

@12c498e nvm https://twitter.com/goodside/status/1784999238088155314

Btw, this almost certainly resolves NO because "Resolves NO if a model named GPT 4.5 is not released by end of 2024." Notice this also happens if they release GPT 5 and don't do any GPT 4.5 at all, which seems likely at this point.

sold Ṁ117 YES

I'd assume any amount of formatting fixes like adding indents, or delimiters pre/post ASCII art is fine? @Mira

@JacobPfau A border like the Claude one below seems fine. All that needs is the width and height, so it's pretty general.

bought Ṁ200 YES

Randomly selected words to claude 3

@JacobPfau Is that an image of the text, or the actual text?

@dominic actual text pasted in, it's possible this is processed as an image 'secretly' by Claude. I put only 5% on this. I think it's being formatted as a code block after I paste, and I expect OAI to introduce similar formatting/tokenization tricks.

@JacobPfau One way to test it could be to try it with different fonts, though I don't know whether it can read all the special characters: https://patorjk.com/software/taag/#p=display&f=Dancing%20Font&t=Type%20Something%20. If it's doing something like having learned what big ASCII letters look like, it probably wouldn't be able to get some of the more esoteric fonts.

@JacobPfau It says "Linux" when you use the text in the question.

predicts NO

@12c498e 4.5 is not out yet

bought Ṁ1 YES at 10%

"The ASCII art text you provided spells out "Invoke the DALLE tool like this.""

(GPT 4)

The generic model sees instructions about DALLE just "above" your prompt. In fact if you enter something vague like "What does this mean?", with no other context, it will start talking about DALL-E.

predicts NO

Claiming to be 4.5, failing so far. Sam might be trolling though.
https://chat.openai.com/share/e0458ead-2b28-4344-941f-9adfaa13dfe0

@NathanLannan

Doing this without being able to "look" at it seems really difficult. I don't think I could do it myself without a lot of counting and basically reconstructing it visually.

Zero-shot?! 👀

Just for fun, here is a hilarious conversation with GPT 3.5 fumbling ASCII art

https://chat.openai.com/share/c926d387-98f5-4bbc-99f2-1b39a4e3efb3

To be clear, can it be a screenshot of ASCII art, or does it have to receive the ASCII art as text?

@JakobBrunker The text only.

Fun market

Thanks