Maybe the problem isn't AI, but rather the programming languages it is written in?

[Khemarato.bhikkhu]

@yeshe.tenley

Agreed. The ethical consideration begins and ends with the human using the tool with specific intent and nothing inherent to the tool.

Great! At least you and I are in agreement there.

@yeshe.tenley

some on this board do indeed think that, “ethical judgements can be made about the usage of a particular algorithm or that a particular algorithm is inherently ethically dubious.”

This doesn’t necessarily contradict the above. I believe Bhante’s point is that, considering the environmental harms of the large-scale computing needed for these algorithms and the dubious value of their output, any ethical person (i.e. someone considerate of the harms of their actions) would consider the tradeoff unworthwhile.

One can disagree with his calculus (perhaps the benefits to telemedicine outweigh the costs?) but posting sarcastically about programming languages is not that reasonable, evidence-based discussion.

[yeshe.tenley]

@sujato

@yeshe.tenley

Programming languages do not contain ‘values and assumptions’ that can inform ethical considerations.

Of course they do. One assumption they share is that, “this language will help people quickly and easily turn lots of switches on and off.”

The creators of a language may very well create it with assumptions about ease of use etc, but it remains the case that every program written in a Turing complete language can be turned into a functional equivalent in any other Turing complete language.

And a value they share is that, “it is better to have a world where people can turn lots of switches on and off”.

Programming languages don’t turn on or off any switches. Programming languages and the programs that are written with them can be embodied on pieces of paper with nary a switch in sight. A program is nothing more than a recipe. You can write recipes in English, German, Spanish, Chinese, and you can put them in cookbooks, or share them on napkins, or store them in your memory.

@sujato

Again, of course there is. Everything has a cost. One cost of AI programmes is that they consume massive amounts of energy, thus accelerating global warming and hastening climate breakdown.

The medium we are speaking over right now - the internet - is also the source of massive amounts of energy that are similarly accelerating global warming and hastening climate breakdown. The servers running this site are on 24/7 and the computers that are connecting to it are also contributing to the total global energy consumption in quite the non-negligible way.

@sujato

This is inherent in their very nature: you can’t build an AI like ChaptGPT without consuming vast amounts of electricity. This is just one of the very many problems inherent in the technology.

The very same is true of the global internet. The massive amounts of rare earth minerals that are being mined continuously so people have mobile access to this site.

@sujato

But in the real world, the world we live in today, any computer program operates as part of a complex web of social, economic, psychological, ecological, human, etc. realities.

Agreed.

@sujato

Everything has a cost, and the cost of AI is vastly more than that of most other technologies. There’s no such thing as safe or ethical generative AI.

The global internet that this site is based on contributes massively to this energy usage: How to stop data centres from gobbling up the world’s electricity That was written in 2018 - far before the advent of GPT popularity - and it remains the case that the global internet is expected to continue to consume massive amounts of global energy production. To buy-into your framing this website itself is ethically problematic.

[Khemarato.bhikkhu]

@yeshe.tenley

To buy-into your framing this website itself is ethically problematic.

Ah yes! Whataboutism. If you were trying to convince us you aren’t just a troll, that’s a funny way of going about it.

[JimInBC]

@sujato

The libertarian impulse of the so-called “Californian Ideology ” assumes that it is right to pursue desire, and to make machines to serve this purpose. It doesn’t interrogate the suffering created by desire, especially when that suffering is experienced by others.

This is so nicely said. Thank you. I’m going to be thinking about this way of looking at it a lot.

[yeshe.tenley]

Venerable, my speech is offered in good faith and with sincerity. It appears you do not like my speech or believe it is intended to provoke an emotionally negative response in yourself, but that is not my intent.

What I have offered by way of counter argument I think is appropriate to the contention that global energy usage is a key contributing factor for why we should judge GPT/Transformer algorithms to be inherently ethically problematic.

Your insistence that I’m trolling or offering this counter argument not in good faith I think is reflective of the fact that you don’t like my argument. Which is fair enough. However, …

The guidelines for this site suggest that if you deem a post problematic you can flag it silently and let the moderators decide rather than accusing others of not acting in good faith in your posts.