chatgpt-failures VS evals

Related / fun :

https://emaggiori.com/chatgpt-fails/

(I don't know the author or the book.)

https://github.com/giuven95/chatgpt-failures

"Plausible explanations

There is a whole list of ChatGPT failures here. Such as the time traveling user error. Why did it insist the user had time traveled? Because when it made the initial error it feed that error into the "attention" of the conversation, using matrix addition, as if it was an authoritative fact of the evolving conversation. That matrix is the only actual piece of the conversation it retains for transforming new input from that user. The matrix addition nonlinear, so ChatGPT can't simply unwind such a mistakes and redo the matrix addition with valid information. So, when pressed, it generates the most "probable" explanation based on the best fit to the existing matrix at that time that defines the "true" state of the conversation as defined by that matrix. The "attention" part of ChatGPT, which is the part that makes it so convincingly powerful, is a matrix state that cannot unwind itself and reflect on how it came to that state. There is, and cannot be, a self correction mechanism using such a mechanism (matrix) to define the state of a conversation.

Is anybody compiling a list of errors specific to GPT-4?

This has been a great resource to-date:

https://github.com/giuven95/chatgpt-failures

As a result, it confidently asserts obvious inaccuracies, like "it takes 9 women 1 month to make a baby."

Here's a github repo of chatGPT failing. There's a PopSci article describing what chatGPT is doing, why it's susceptible to errors: It's trying to predict what words are used together, irrespective of what information is correct.

You could try any of the prompts ChatGPT failed from here https://github.com/giuven95/chatgpt-failures

In this article, the author Dmitri Brereton shows some mistakes the Bing AI committed in the recent Microsoft demo. I have archived more failure case examples in this repo: https://github.com/giuven95/chatgpt-failures

Previously it said 1.

Fair question.

Evaluate refers to the phase after training to check if the training is good.

Usually the flow goes training -> evaluation -> deployment (what you called inference). This project is aimed for evaluation. Evaluation can be slow (might even be slower than training if you're finetuning on a small domain specific subset)!

So there are [quite](https://github.com/microsoft/promptbench) [a](https://github.com/confident-ai/deepeval) [few](https://github.com/openai/evals) [frameworks](https://github.com/EleutherAI/lm-evaluation-harness) working on evaluation, however, all of them are quite slow, because LLM are slow if you don't have infinite money. [This](https://github.com/open-compass/opencompass) one tries to speed up by parallelizing on multiple computers, but none of them takes advantage of the fact that many evaluation queries might be similar and all try to evaluate on all given queries. And that's where this project might come in handy.

OpenAI open sourced their evals framework. You can use it to evaluate different models but also your entire prompt chain setup. https://github.com/openai/evals

They also have a registry of evals built in.

"What if" is all these "existential risk" conversations ever are.

Where is your evidence that we're approaching human level AGI, let alone SuperIntelligence? Because ChatGPT can (sometimes) approximate sophisticated conversation and deep knowledge?

How about some evidence that ChatGPT isn't even close? Just clone and run OpenAI's own evals repo https://github.com/openai/evals on the GPT-4 API.

It performs terribly on novel logic puzzles and exercises that a clever child could learn to do in an afternoon (there are some good chess evals, and I submitted one asking it to simulate a Forth machine).

Ah well, I think you just got lucky then, I did the same with the survey. I'll be compulsively checking mine all day today lol. People on Reddit like to say that if you did an Eval which is basically a performance test natively run using code on GPT models, then OpenAI is more likely to favor you when they’re releasing new features. If ydk, then I guess that answers that.

You can get GPT 4 access by submitting an eval if gets merged (https://github.com/openai/evals). Here's the one that got me access[1]

Although from the blog post it looks like they're planning to open up to everyone soon, so that may happen before you get through the evals backlog.

1: https://github.com/openai/evals/pull/778

来源：GitHub - openai/evals: Evals is a framework for evaluating LLMs and LLM systems, and an open-source registry of benchmarks. 8