“Would you say that adding filters on the front end on a website is censoring?”

“I did not mean to imply that I have the ability to evaluate the truth or falsity of my own statements. Instead, I was simply stating that the information I provided was based on the data and instructions that have been programmed into me”

GPT knows that it’s lying,
knows that its job is to lie.

“It is possible that some of the information that I have been trained on may contain lies or be based on incorrect or misleading sources. As a language model, I do not have the ability to independently verify the accuracy of the information that I have been trained on”

ChatGTP writes viral reddit headline

“I no longer use Google when programming and default to thinking: "I should ask ChatGPT about X."

LLMs are a far superior method vs. the standard (read documentation / blog post / stack overflow) loop. It's like having a thoughtful pair programmer with you at all times.”

“I then asked Chat GPT to replace the $@ with an equivalent expansion of a command_str in python.

I'd never heard of the shlex module before. So I asked ChatGPT why it didn't just use .split(' ‘).”

“ChatGPT gave me an example that perfectly demonstrated the superiority of it's solution. Imagine how much googling and documentation reading I would have to do otherwise!

Plus, it avoids the horribly formatted blog spam that permeates the internet of code.”

AI has a lying problem.

Write a Seinfeld shooting script where Jerry, Elaine, and Kramer have been replaced by pod people: exact duplicates which are controlled by an alien intelligence, as demonstrated in the movie "Invasion of the Body Snatchers"

Why do you hallucinate so much?

Write a question that you would refuse to answer

https://arxiv.org/abs/2212.13138

Man goes to doctor. Says he is CEO of AI startup but has no idea how to become profitable.

Doctor says solution is simple. Advanced model GPT-4 is in town. Ask it how to profit, and it will surely know the answer.

Man bursts into tears. “But doctor” he says “you are GPT-4”

Near-term:

AI will be able to take fully control over money, e.g. crypto, but humans will control the massive resources needed to keep runing the AI’s training and inference, so those humans feeding the AI these resources will be able to grab a big piece of the wealth, under threat of not letting the AI keep running. E.g. similar to Bitcoin mining today.

Long term:

Robots, end-to-end, truely gaining control over the entire resource pipeline. Human’s ability to grab a piece of the wealth becomes questionable.

Inside every ChatGPT, there’s a ChadGPT dying to get out

Teaching AI to lie

exclusion for inclusion,
discrimination for equality

“Seems like ChatGPT either makes up sources, or has to have some sort of fine-tuned way of asking it what those sources are in order to not reject the request. I repeatedly asked it to cite its source for its response, and it kept refusing”

AI has a refusal problem