Not sure if this is the best community to post in; please let me know if there’s a more appropriate one. AFAIK [email protected] is meant for news and articles only.
Not sure if this is the best community to post in; please let me know if there’s a more appropriate one. AFAIK [email protected] is meant for news and articles only.
Machine Learning and the training and use of targeted, specialized inferential models is useful. LLMs and generative content models are not.
Let’s not forget about traditional AI, which have served us well for so long that we stopped thinking of them as AI.
What?
As in, I agree with your point. I just want to give a shoutout to the non-ML-based AI.
In the strictest sense of the technical definition: all of what you are describing are algorithmic approaches that are only colloquially referred to as “AI”. Artificial Intelligence is still science fiction. “AI” as it’s being marketed and sold today is categorical snake oil. We are nowhere even close to having a Star Trek ship-wide computer with anything even approaching reliable, reproducible, and safe outputs and capabilities that are fit for purpose - much less anything even remotely akin to a Soong-type Android.
That’s why this joke definition of AI is still the best: “AI is whatever hasn’t been done yet.”
I have forgotten all working definitions of AI that CS professors gave except for this one 🙃
In the strictest sense there is no technical definition because it all depends on what is “intelligence”, which isn’t something we have an easy definition for. A thermostat learning when you want which temperature based on usage stats can absolutely fulfill some definitions of intelligence (perceiving information and adapting behaviour as a result), and is orders of magnitude less complex than neural networks.
What! LLMs are extremely useful. They can already:
-Funnel wealth to the richest people -Create fake money to trade around -Deplete the world of natural resources -Make sure consumers cannot buy computer hardware -Poison the wells of online spaces with garbage content that takes 2s to generate and 2 minutes to read