Which is closest to your view?

[bartholin@fops.cloud]

@ZachWeinersmith A1

[degroof@mastodon.social]

@ZachWeinersmith Probably A2, but mainly due to the difference between actual AI research and the generative predictive engines currently being marketed as AI.

[tatra@masto.adeserv.uk]

@ZachWeinersmith Depending on our definitions of "AI" and "work", my answer varies. If we're talking about the current crop of Gen AI tools, and by "work" we mean something like "reliably delivers useful results when used in the intended way" - A1, easy.

[ela@infosec.exchange]

@ZachWeinersmith A3.

[mikanux@eldritch.cafe]

@ZachWeinersmith A2

[jpkolsen@social.data.coop]

@ZachWeinersmith A2

[foolishowl@social.coop]

@ZachWeinersmith A2.

[orb2069@mastodon.online]

@ZachWeinersmith

A1 - Cherry-picked results don't mean "working", even if it's the chatbot framework doing the cherry picking.

[kolamagpie@this.mouse.rocks]

@ZachWeinersmith A2

[erinaceus@chaos.social]

@ZachWeinersmith A2

Assuming by „AI“, you mean generative models such as LLMs and image generators.

Imho, the technology _is_ impressive but very overhyped. The harm to society and environment is greater than its usefulness, by far.

[csstrowbridge@mastodon.social]

@ZachWeinersmith

Between A1 and A2.

First of all, AI doesn't exist and it was recently mathematically proven that what we have now, machine learning / LLMs, can NEVER grow into full AIs.

What we have now is good a pattern recognition. You can feed it all of the drugs used to fight a certain disease, their effectiveness, their side effects, etc. and it can spit out other drugs that would be useful targets for study.

What we have would fit into A2. What we are being sold is a scam.

[rexo@st.transbus.social]

@ZachWeinersmith@mastodon.social A1, maybe A2 but I'm somewhat skeptical

[fonant@social.vivaldi.net]

@ZachWeinersmith A1

The bubble can't burst too soon.

[none@tkz.one]

I'll say A2

(It seems I'm not the only one)

@ZachWeinersmith

[c0dec0dec0de@hachyderm.io]

@ZachWeinersmith AI is a vastly overloaded term, but I assume that we’re focusing on LLMs. For LLMs, I’m in the A2 camp, but hold A1leanings.
For classifiers and other classical machine learning methods, C3 except that they’re not new. One also still has to be aware of input biases and “empiricism washing” or whatever where people pretend because you’ve used a mathematical tool that the results are morally clean and correct.

[kichae@wanderingadventure.party]

Zach Weinersmith
A1.5. At least it we're specifically talking about generative ML models. I don't think it truly works in any domain, but it gets close enough that it looks functional in small chunks (and may be good enough in the moment). Kind of like how the Earth is experientially flat at the human scale, and can be assumed t( be so for most daily tasks.

[cassandra@ottawa.place]

@ZachWeinersmith Currently A1 but willing to nudge toward A2 if the evidence shows, e.g., fewer people are now dying of cancer.

[cupidissimo@fnordon.de]

@ZachWeinersmith I suppose C2, I mean in theory A2 but in practice, to me the relative wastefulness of AI (and by that I mean LLM) does not actually make it a good fit for a lot of tasts it is currently being used for.

[mistersql@mastodon.social]

@ZachWeinersmith c2/c3 because in the context of setting all the fossil fuels on fire to make food and heat our houses, we have to burn fossil fuels to solve the problems created by being unwilling to stop burning fossil fuels for food and space heating. Everything we do is setting ancient forests on fire, just some of it is socially acceptable.

[hist_myth@mastodon.top]

@ZachWeinersmith A2