Apr 19, 2023

AI Hype: Dissecting the Claims

This article looks at the hype surrounding AI and examines the claims made by a particular article. Read on to learn more about the implications of this technology and the risks associated with it.

AI ETHICS

@emilymbender@dair-community.social on Mastodon

Professor, Linguistics, UW // Faculty Director, Professional MS Program in Computational Linguistics (CLMS) // she/her // @emilymbender@dair-community.social

Member of Linguistics

Alright folks, ready for this morning's lesson in #AIhype? On the dissection table:https://t.co/zph0b0Siro

>>
— @emilymbender@dair-community.social on Mastodon (@emilymbender) April 19, 2023
Right off the bat, the headline looks fishy (which is why I clicked through, actually). Governance, risk and compliance sound like domains where you don't want a chaos-machine, TYVM.

>> pic.twitter.com/zvjzQB8QIS
— @emilymbender@dair-community.social on Mastodon (@emilymbender) April 19, 2023
Oh, and while we're here. Take a gander at that byline. Membership is fee-based, is it? Is this in fact an infomercial?

>>
— @emilymbender@dair-community.social on Mastodon (@emilymbender) April 19, 2023
Paras 1 & 2: ChatGPT has "taken the world by storm". It's "sophisticated". Yadda yadda. Hype-ity hype hype. Main claim: "can help businesses up their game in some very specific ways, including how they manage risk."

>>
— @emilymbender@dair-community.social on Mastodon (@emilymbender) April 19, 2023
Paras 3 & 4: Ah, an "AI" explainer from someone who doesn't understand the tech. ChatGPT isn't "two different technologies". It's text munging all the way down.

>> pic.twitter.com/b630jWAs11
— @emilymbender@dair-community.social on Mastodon (@emilymbender) April 19, 2023
Note the allusion to superhuman capabilities: "deduce patterns and relationships in the data that we humans can’t readily identify". Yes, we regularly use tools to achieve tasks we can't do without said tool. Would you say a hammer can drive in a nail that a human couldn't push?
— @emilymbender@dair-community.social on Mastodon (@emilymbender) April 19, 2023
This paragraph is pretty close to right. The responses are indeed coherent seeming, plausible and palatable (as judged by the annotators OpenAI hired).

>> pic.twitter.com/u6l8UnfOGX
— @emilymbender@dair-community.social on Mastodon (@emilymbender) April 19, 2023