One for Progaemmers and Gerdeners

Of all the "AI"s send to plague us, Gemini is by far and a way the most useless.

Have you had it tell you that you are mistaken yet ?

No, but it often tells me it may be wrong :-)

In message snipped-for-privacy@news.individual.net>, Jeff Gaines snipped-for-privacy@outlook.com writes

Please let us know how you get on with that.

Adrian

A couple of days ago I had TV subtitles engaged for a mumbling drama. Turned to BBC news , good subtitles, then the weather . Looked like AI fantasy, none of the ,wreck a nice beach instead of recognise speech , sort of stuff ie logical/understandable "mishearing" . I've since engaged subtitles for BBC tv weather and its been fine. A live try out of AI speech recognition, never to be repeated?

Some of what AI is now reporting as fact is the often repeated bullshit on social media.

Chatgpt recommended sourcing a replacement pivot pin for a sash window: "Don't buy them from Wickes or B and Q, they are not the right ones"

(based on a photo of a UPVC sash window with a detached spring).

if you appened "-ai" to your google query it currently drops the useless summary saving screen estate.

I'm currently seeing searches as 50% less efficient than in 2021 (when I last looked properly). That is a *lot* of electricity for much less output. I wonder who is paying ?

Wow, thank you, works a treat :-)

You can also add &udm=14 to the URL:

Theo

Whenever you wonder that, the answer is not too far away.

It isn't just Google:

We never needed AI for that: I started complaining years ago about "Internet 'Facts'". The majority of internet users are uneducated or unintelligent, or both; when they want to know something (like: all the time) they. google it (like: "it's just so easy!"),

Since Google trawls the entire Net for its answers, without a Stupid filter, dumb buggers are educating: each other - and thus creating "facts". If you're educated you stand a chance of mistrusting such facts and checking them. But most people are too busy to do that (because something else has pinged their screen even while looking something up!). But if you express a doubt, they will raise their voices at you termerity to do so and tell you to "Lookit up!" (which to them simply means: google it.)

But certainly: AI can create Internet Facts 1,000,000,000 times faster than mere morons.

Brought to you by

Disgusted, Disillusioned and Disheartened of Tunbrdige Wells ;-)

Some of the impact of AT is less obvious. Yesterday, I realised I was beginning to think of Wikipedia as an authorative, trustworthy source.

"Everything is relative".

I tried ChatGPT with a question about how to do some complicated electronics communications thing I'd been working on for a few years. I wasn't expecting much, people in this field don't make their IP public domain.

It produced something which was technically completely useless, but which with a few tweaks would have made a good pitch to not very technical management. Buzzwords galore.

SWMBO had an appointment with an ENT consultant. He asked if she'd mind if AI was used to produce a report. He was wearing a microphone which recorded everything said, and he did deliberately (it seemed to me) say some things in a more concise manner than you might expect, but then we'd never met him before, so who knows.

The report was very well drafted, couldn't fault it. I don't know how much if any editing was done, but it was produced quickly.

It's not just AI but the promoting of sponsored links for items you may be looking for and then the only company in the Uk that stocks it will be number 100 in the search list.

I listened to a consultant on the radio saying that AI was great for the same thing BUT she also said although it saved a lot of time writing patient notes/records it had the be reviewed immediately after the appointment to correct any mistakes. Someone I know, a victim of crime, had a Police interview which was recorded and told that AI had been used to produce the report. When reviewed the report was somewhat bollox. As you indicate, perhaps to use this technology in this way you do have to structure the language used and maybe the interviewer needs to precis the ramblings of a victim.

I have various Access files but have never heard of 'tree data'. It seems AI hasn't either. What do you mean by it?

Sometimes the consultant's report looks nothing like what was said in the consultation, and that then goes on your records which you have to correct. Perhaps having a transcript will avoid making those kinds of mistakes? Or will it just apply its own bias, based on the training dataset?

Theo

I'd guess the defendants need to raise a complaint for anything to happen.

Or maybe the Bar Standards Board is more proactive . . . certainly should be.