on being an AI Vegan & ChatGPT

I’m a computer scientist and something I call an “AI vegan”: I don’t feel that it’s ethical for me to consume models such as ChatGPT, which probably means I'm not a lot of fun at parties.

ChatGPT isn’t an “artificial intelligence” as its creators would have you believe: it’s a big ball of numbers that was made, without consent, from the content of the Internet.

ChatGPT itself has no ethical constraint, nor can it “think”, “reason”, or understand “truth” at all. It is a fancy math wibbly wobbly copy machine: put in your words, get random chopped up bits of sentences and words that ChatGPT calculates may match the words that you put in. It is a black box whose contents you cannot examine.

49% of OpenAI, the company that owns ChatGPT, is owned by Microsoft. Microsoft sells their technology to world governments and weapons manufacturers, and systems like ChatGPT are already being used to replace real human decision making in life-or-death sitations.

I'm not giving out brownie points, but OpenAI does discuss their policies around proper applications for ChatGPT.

Their policies claim that ChatGPT is not intended for “ tailored legal, medical/health, or financial advice” unless it's “[reviewed] by a qualified professional” and “[discloses] use of AI assistance and its potential limitations.”

Given the mass amount of data collected by the OpenAI, they might catch cases where GPT is made to impersonate a doctor or lawyer. Maybe they'll stop this from happening. In my opinion? They will not, and they can always sell a “ethics-free license” for the right customer.

Here’s a podcast that explains some of the issues with ChatGPT and similar products:

https://www.techwontsave.us/episode/163_chatgpt_is_not_intelligent_w_emily_m_bender

According to MIT Technology Review, Large language models are wasteful and damage the environment.

AI researchers have written a deep & technical paper on the manifold Limitations and Ethical Considerations of ChatGPT.

This article is a work in progress, and I will update it as I gather my thoughts and evidence. Last edited 2025-01-24

I’m happy to discuss further: I have far more thoughts on the subject and I won't bullshit you. This is not a good technology, and it presents a trap for the unwary. Email me at fred@chocolate.software