Evan Prodromou
A major difference between LLMs and cryptocurrencies is this:
For cryptocurrencies to be valuable to *me*, I need them to be valuable to *you*, too.
If you don't believe in crypto, the value of my crypto goes down.
This isn't the case for LLMs. I need enough people to be interested in LLMs that ChatGPT stays available, but other than that, your disinterest in it is only a minor nuisance.
In a market, I benefit from your refusal to use it. AI-enhanced me is competing with plain ol' you.
clacke
@evan The dangers of "AI", really the dangers of accelerated automation of capitalism with increasing complexity and diminishing human control and oversight, are neither theoretical nor future tense.
They are not reflected in the majority's daily usage, but they've been here for decades and are growing exponentially.
It's not about sentient machine overlords, it's about concentrated power to human overlords abstracted away from the harm they're causing.
Evan Prodromou
Probably the biggest difference between cloning and AI is that human cloning was still theoretical in 1996, without practical use.
In 2023, 100 million people per month are using ChatGPT.
I don't think that many people are going to be willing to give up a useful tool for theoretical dangers that aren't reflected in their daily usage.
Evan Prodromou
Do I know how we should use cloning? No, I don't.
Do I think we used this 30-year pause to carefully think over, as a society, how to use cloning technology ethically?
Absolutely not. We're no more ready for it than we were in 1996.
Evan Prodromou
The closest analog to the current situation with LLMs is human cloning.
https://en.wikipedia.org/wiki/Human_cloning
In 1996, with the successful cloning of "Dolly" the sheep, there was a flurry of concern about the ethical dangers of human cloning.
Many countries and international organizations moved to outlaw or "pause" cloning of humans until we as a society figured out how to use the technology ethically.
Almost 30 years later, human cloning is still not a mainstream technology.
Evan Prodromou
Probably the biggest threat to me and my use of AI is that you squawk enough to get the technology restricted to only "authorized" people -- governments, big companies.
I think the horse may be out of the barn on that one.
Evan Prodromou
@timbray @Jonathanglick this is true!
About 2-3% of global emissions comes from computer technology, and it would be a big help to get them under control.
Video games and streaming video are by far the biggest emitters, but search engines, ecommerce and social networks come in pretty close, too.
I'd love to see more public providers using carbon offsets to balance out their emissions, and carbon taxes may help more.
It's an important topic to bring up!
Tim Bray
On another dimension, crypto & LLMs are similar in that they are both really expensive, requiring large investments in infrastructure, and with a high carbon cost.
(Not what you’re talking about, I know.)
Jonathan Glick
@evan Right. I think, though, that it’s maybe more than LLMs ‘needing enough people to be interested.’
Somewhat like crypto, for AI to be widely adopted and integrated, it needs popular support to prevent heavy regulation *despite belief of serious potential risks and a lack of immediately evident social benefits.*
Evan Prodromou
@Jonathanglick it's not *strictly* true. OpenAI uses the conversations people have with ChatGPT to train the next generation model. So more people having better conversations with ChatGPT is somewhat better for me. But not entirely necessary, in the same way as crypto.
Jonathan Glick
@evan and I really liked your broader point — that crypto is necessarily a ‘multi-user game,’ whereas AI is not.
GNU social JP is a social network, courtesy of GNU social JP管理人. It runs on GNU social, version 2.0.2-dev, available under the GNU Affero General Public License.
All GNU social JP content and data are available under the Creative Commons Attribution 3.0 license.