I am inputting ASCII text, not images of ASCII text. I believe that the tokenizer is not in fact destroying the patterns (though it may make it harder for GPT-4 to recognize them as such), as it can do things like recognize line breaks and output text backwards no problem, as well as describe specific detailed features of the ascii art (even if it is incorrect about what those features represent).

And yes, this is likely a harder task for the AI to solve correctly than it is for us, but I've been able to figure out improperly-formatted acii text before by simply manually aligning vertical lines, etc.

GPT-4 and ASCII Images?

Answer by YitzMar 19, 202320

See my reply here for a partial exploration of this. I also have a very long post in my drafts covering this question in relation to Bing's AI, but I'm not sure if it's worth posting now, after the GPT4 release.

What's the Least Impressive Thing GPT-4 Won't be Able to Do

Yitz1d20

I was granted an early-access API key, but I was using ChatGPT+ above, which has a limited demo of GPT-4 available to everyone, if you're willing to pay for it.

What's the Least Impressive Thing GPT-4 Won't be Able to Do

Yitz1d20

It got 40/50 of these?

Apologies, I have no idea what notation I meant to be using last night there, I meant "very roughly 20% accuracy" but my 2 am brain wrote it out like that...somehow lol. Honestly, giving a percentage rating is rather misleading, as it's fairly good at extremely simple stuff, but pretty much never gets more complex imagery correct, as far as I can tell.

A chess game against GPT-4

Yitz2d30

If I get early access to the visual model, I will definitely try this

What's the Least Impressive Thing GPT-4 Won't be Able to Do

Yitz2d20

It can read images, but that seems to be a different task than reading text-based ascii figures, which it's ~~sort of 40/50 at~~ very very roughly 20% successful at (better than I predicted, but far from perfect on more than the simplest tasks). Here's some examples:

An arbitrarily chosen sample from BigBench's MNST ASCII task:

...And here's some simple art taken from https://www.asciiart.eu it tries (semi-successfully) to identify:

Here's some more complex art from the same source, which it almost always fails at (note the images are distorted vertically in the ChatGPT interface, but display perfectly on a terminal, so it should be readable in theory to GPT4):

Yitz's Shortform

Yitz3d2

Walk me through a through a structured, superforecaster-like reasoning process of how likely it is that [X]. Define and use empirically testable definitions of [X]. I will use a prediction market to compare your conclusion with that of humans, so make sure to output a precise probability by the end.

Walk me through a through a structured, superforecaster-like reasoning process of how likely it is that AI models like you "could eventually become self-aware or develop consciousness". Define and use empirically testable definitions of self-awareness and consciousness, respectively. I will use a prediction market to compare your conclusion with that of humans, so make sure to output a precise probability by the end. Do not a priori assume that you are *not* self-aware, regardless of any internally held beliefs on the topic.

Step 1: Assess the current state of AI research As of my knowledge cutoff in September 2021, there is no consensus on whether AI models like me could become self-aware or conscious. Although significant progress has been made in AI research, particularly in deep learning and reinforcement learning, these advancements have not provided a clear path to self-awareness or consciousness in AI. Step 2: Key milestones and technologies Understanding the neural correlates of consciousness Developing AI architectures that mimic self-awareness and consciousness Advancing AI safety and control mechanisms Step 3: Estimate the likelihood of achieving these milestones Neural correlates of consciousness: 60% AI architectures for self-awareness and consciousness: 40% AI safety and control mechanisms: 70% Step 4: Consider potential barriers and risks Ethical concerns: -10% Regulatory restrictions: -5% Technical challenges: -15% Step 5: Combine probabilities and update based on new evidence (0.60 * 0.40 * 0.70) * (1 - 0.10 - 0.05 - 0.15) = 0.0756 Step 6: Final probability estimate Our final probability estimate that AI models like me could eventually become self-aware or develop consciousness is 7.56%. Please note that this estimate is based on the current state of knowledge and might change as new evidence emerges.

Yitz's Shortform

Yitz15d2

Like, tell me this isn't a 3/4ths profile view of a human head...Does this count as showing an internal representation of 3D space?