Not long ago I wrote about why I don’t use AI in 2025. I posted it on Lemmy and soon I had started a bit of a internet argument.
I’ve used the same headings here as in my original piece to divide up the criticism.
Energy use
I complained that the energy use of AI was inefficient compared to a conventional search. patientpenguin@feddit.org replied with an interesting article by Andy Masley which is quite persuasive in arguing that the environmental impact of using a LLM is small compared to doing other things.
It is true that there are definitely bigger problems. The International Energy Agency, who certainly sound like they know what they’re talking about, said that only 1.5% of the worlds energy use was for AI last year.
Still, I’m not convinced to change my view on this point - despite it being a small change, using a search engine seems to be still better than using AI. This makes it worth it to me right now.
Some people even suggested using locally run AI models. Flagstaff@programming.dev specifically suggested Jan.ai, which is also open source. These can have lower energy use and you can tell how much energy they are using. This is better but there are two large problems.
- The creation of local AI models still requires a unreasonably large amount of energy.
- Local AI models are reportedly less reliable than cloud models, bring us onto the next point.
Reliability
There was little mention of this point, probably because what I said is widely accepted - AI doesn’t consistently tell the truth. This remains a major downside for me.
Ethical problems
There was naturally lots of argument about this, mostly about copyright. After reading them I am more confused than ever about the legalities.
Whether or not AI created from copyrighted work is legal, however, was not really my argument - even if it is legal then it is bad that nothing is being given back to those who created the works, copyrighted or otherwise that are being made into AI models.
Enjoyment
Saying that I don’t use AI because it is more fun not too is unsurprisingly uncontroversial.
I don’t like it
This is a point that I have changed my mind about. I argued - or rather admitted - that part of the reason that I didn’t use AI was because of the people and culture that I associate with it.
Blue_Morpho@Lemmy.world questioned me on if I hated electric vehicles (EV’s) as well due to their connection to Musk. I don’t because I can separate his support of EV’s and my dislike of him.
I should be able to do this with AI as well - I still wouldn’t use Grok or ChatGPT, just like I wouldn’t buy a Tesla but I can’t write off the whole idea just because of a group of people I don’t like.
Have I changed my mind?
No. I still don’t use AI and won’t in the near future.
I don’t believe that we will be able to make AI accurate enough that I would be happy to use it, however engineering produces unexpected things all the time.
If the reliability somehow improved dramatically without increasing the energy usage then I would be more open to using it.