Stance on use by AI projects

[thegeronimo]

Hello,

You have made clear your position on supporting cryptocurrency projects, specifically related to bitcoin because of the demand they have on energy consumption.

With AI it's clear the demand for energy is massive, causing major power plants that have been shut for decades to reopen and new ones to be built. What is your position on supporting such projects that use websockets?

[aaugustin]

The energy consumption is actually my secondary gripe against cryptocurrencies. I chose to document that one because it's uncontroversial, unlike the primary gripe :-)

My primary gripe is that, after more than 15 years, cryptocurrency still doesn't have any significant use case that's beneficial to humanity. It's absolutely great if you're into:

• enabling financial fraud — tax evasion was the original use case and the same properties made it great for money laundering, financing terrorism, etc.
• moving money from the poor to the rich even faster than capitalism; all the cool tricks from unregulated capitalism in the 1920s work great; insider trading, wash trading, and market manipulation are just the tip of the iceberg;
• trafficking illegal goods — drugs and malware are the most harmless of those; humans and organs are more bothersome; and surely there's some even more ugly stuff I'm not even aware of.
• bonus: ransomware — simply wouldn't exist without cryptocurrency.

This is because the technology behind cryptocurrencies fundamentally solves the problem of transacting with people that you don't trust, without any recourse. (That's also what makes it fundamentally unsuitable to how humans manage money in the real world.) There are very few real-world, legal situations that require this property. (Before you ask, cross-border money transfer could be solved without that property because banks are able to establish trust networks. We are unfortunately stuck with a protocol designed in the days of horse couriers: SWIFT. Blockchain isn't a properly engineered solution to this problem because it lacks some required properties and provides unnecessary properties.)

AI has been around at least since the 60s. The most recent generation, dubbed Gen-AI, became mainstream with ChatGPT two years ago but the underlying technology had been around for a few years already, perhaps five years. Every generation took advantage of hardware progress and increased energy consumption. Gen-AI increased it very badly. I hope that we'll learn to use smaller models when VC stop subsidizing the industry as much as they currently do.

Unlike cryptocurrency, Gen-AI seems headed towards having significant positive impact on the world. There are definitely some risks for how people work, how they learn, etc. If this technology becomes part of the mainstream, it will alter our lives, like the Internet did. The Internet had problematic side-effects but most people agree that it was a positive development. I believe that Gen-AI will land in that category.

I am aware of concerns about the possibility of extinction of humanity through the emergence of AGI. I familiarized myself with the positions of the "doomer" community. I reviewed their reasoning and I see too many gaps to be truly concerned about that scenario. However, I admit that I don't have the knowledge required to make an expert judgment on this.

If humanity has to end because of an unfortunate technology development, I'd rather have it end because we invented AGI than because someone didn't want to pay taxes.

---

This is a longer answer than your question really deserved. Also, it's off topic for this issue tracker. I don't intend to argue my position further. I will lock the issue preemptively.

Let's go back to writing good software now!