I hate generative AI for many many reasons, but one of my major pet peeves is the fact that it muddies the water for what other non-generative AI does (and how many of them are in fact actually helpful and used in much more ethical ways)
Comments
Log in with your Bluesky account to leave a comment
And also confusing different types and forms of AI just makes it more difficult to actually name and fix the issues that are present in different technologies
Can you give me some examples of non-Gen AI that have helped you? I find Claude quite helpful for my learning difficulties but I am a bit concerned about the environmental impact and whether or not what I'm doing is ethical.
And even some of the current crap may well be important as it'll be a step on the road to something useful.
But trying to sell most of the current stuff as a viable product is like trying to sell a prototype propeller engine as an F16 Fighter Jet
Very true. The misrepresentation of Gen AI as a General Artificial Intelligence (AGI) on its way to human reasoning drives me up the wall. It's a Narrow Artificial Intelligence (ANI) that does a great job at giving you the statistically most likely set of words in response to a prompt. 1/
Understand or apply reasoning to sources? No. That's literally not what it's designed to do. (And that's why marketing it as a search engine is so infuriating.)
And even Machine Learning (ML) like comparison models for skin cancers, tanks, facial recognition has had subtle but deal breaking 3/
Problems pop up. Identifying rulers as a cause of skin cancers, soviet tanks come out at night and, according to the model, all Black people look the same and do crime. (All real examples)
Amazon used AI to screen candidates and if you weren't a tennis playing white man called Dave? You were
Comments
But trying to sell most of the current stuff as a viable product is like trying to sell a prototype propeller engine as an F16 Fighter Jet
I think the use of "AI" as a general term for these Large Language Models (LLMs) and picture slurrification devices is deliberate marketing strategy.
It's "intelligent" in that it is good at giving you the most statistically likely responses. Can it evaluate 2/
And even Machine Learning (ML) like comparison models for skin cancers, tanks, facial recognition has had subtle but deal breaking 3/
Amazon used AI to screen candidates and if you weren't a tennis playing white man called Dave? You were
Yes, gen AI being sold as A(G)I is giving ML in general a bad name.
ML in general has had definite, serious, dealbreaking problems.
It's very hard to convince people who have been unintentionally "cold read" by an LLM that it's not really smart.
https://softwarecrisis.dev/letters/llmentalist/