Project Support versus Reference

Eh…I’ve seen a persistent determination to give me a wrong answer in a chat AI. After eight or so wrong answers, each followed by me telling it it was wrong, it finally admitted it didn’t know. On another occasion, it went through several wrong answers, then returned to the first wrong answer.
Which model was it?
 
Whatever my employer has licensed, and I’m not willing to name my employer. :) I can totally believe that some lie and/or are confused more than others. But I’m nowhere near actually trusting any AI yet.

The development of AI in the last year or so has been remarkable, and at this point in my business (not-for-profit healthcare), we couldn't live without it. We just got funded by a multinational to make a custom AI service that can learn new languages based on languages with the same root. In cities like mine, with 200 spoken languages, we can slowly accrue more and more languages based on simply using the app.

I was demoed an AI talking therapies app the other day (at the mere cost of £60,000 a license) that another NFP CEO and trained therapist I know said was as good as the best therapist in her organisation, totally flawless. Hospitals routinely use AI to scan for tumors in X-rays or UC in biopsies. Every day Im hearing about total game changing uses.

This is not some hit-and-miss tech, and one bum experience with a basic chatbot oughtn't to put you off the entire realm of AI. This is probably the biggest boon to productivity in the last 2 decades. It'd be like trying the internet for the first time and being given a wrong telephone number, so you never used it again.
 
It'd be like trying the internet for the first time and being given a wrong telephone number, so you never used it again.
I said “yet.” I don’t trust it yet.

I will keep on verifying everything an AI does, for the foreseeable future. If the wrongness reduces, the verifying may reduce.
 
Top