AI being helpful thread

[3930dave]

There was a recent video of a "person" commenting on a Canadian situation.

First thing I found odd, was that the English captions were actually embedded in the video graphic (could not switch Off). Very odd, as the "person" was speaking English, well.

Then, I paid more attention to the graphics - standard shot, "guy" wearing a casual dress-shirt, sitting in a home library, shot mid-chest up. Ignoring the caption graphics, and looking closely at the "head"/shirt area, something was off...... it was likely an AI generated face, positioned on top of an actual video of a body/library background. It was done well enough, that w/o the forced into the video captions, many people might not have noticed the faux person "talking", but the face/shirt boundary was slightly hinkey at times, and when I paid attention, the speech (good English) and lip movement was slightly weird.

I don't recall any distinctive disclaimer on the video, concerning the face.

Most of us don't have much use for carbon-based Talking Heads, but moving into a world where you can press a button and generate thousands or millions of AI ones, all spewing a similar message, definitely takes things to another level.

^ Why Off Switches, are useful.

Rgds, D.

 

[3930dave]

ChatGPT probably saved me an hour's worth of useless searching, and ultimately giving up to buy incandescent bulbs again.

A long time ago, I think it was the head of MIT's Media Lab that coined the term "data smog".

Your LED case, a good example of cutting through the smog, quickly.

Rgds, D.

 

[quicksandfarmer]

I needed some B11 chandelier bulbs to restock the cupboard, and have been shying away from LED because nearly all of the warmer color temp ones I have tried have horrible CRI, and make colors (esp. skin) look so weird and unholy. But with astronomical energy usage in this house, I'm always looking to switch what I can to LED, so off I go searching on Amazon.

After about 10 minutes of wasting time, trying to find CRI on every bulb they list (not easy), I had a bright idea... put chatGPT to work for me!

So, I head over to chatgpt.com, enter "sort all of the medium base 2700k - 3000k 40W equivalent candle or candelabra style bulbs on Amazon by CRI", and out pops a sorted list of the highest-CRI warm-white B11 bulbs. 30 seconds later, they're in my cart, and I'm checking out.

ChatGPT probably saved me an hour's worth of useless searching, and ultimately giving up to buy incandescent bulbs again.

That said, I still had to buy some incandescents, as some of our lighting systems (e.g. Lutron RadioRA) are not LED-compatible. Hanging just one incandescent on a non-LED dimmer can make it work as-designed, even if all the rest of the bulbs on the same circuit are LED. Likewise with outside fixtures, they tend to fog up and freeze if all LED, so I usually stick one incandescent bulb in with two LED's in 3-bulb lamp post fixtures, etc.

But the problem is that Amazon has deliberately en****tified their search. They don't want you to be able to quickly find the product you're looking for, because they'd rather you end up buying something that you weren't looking for too. It's the same reason supermarkets have end-cap displays rather than just having the store carefully organized. If Amazon wanted to they could provide that same functionality without AI.

I still maintain that the only reason AI search results appear at all impressive is that other search has gotten so lousy. I'm old enough to remember when Google focused on trying to deliver exactly what you were looking for, and it seemed 95% of the time the first hit on a Google search was all you needed. The reason that's no longer true is that Google has figured out they make more money from ad revenue with bad search than with good search.

So how long before the AI companies figure out what Amazon and Google have figured out? That bad results make more money than good ones? Because you know they will.

 

[Jstpssng]

Google focused on trying to deliver exactly what you were looking for, and it seemed 95% of the time the first hit on a Google search was all you needed

Wow! You ARE old! You also just brought up my biggest lament on the "Peeves" thread... how useless search engines have become.

 

[EscapedSuburbia]

I thought maybe changing search engines would help but so far every one I've tried has been every bit as bad as Google.

 

[quicksandfarmer]

But the problem is that Amazon has deliberately en****tified their search. They don't want you to be able to quickly find the product you're looking for, because they'd rather you end up buying something that you weren't looking for too. It's the same reason supermarkets have end-cap displays rather than just having the store carefully organized. If Amazon wanted to they could provide that same functionality without AI.

I still maintain that the only reason AI search results appear at all impressive is that other search has gotten so lousy. I'm old enough to remember when Google focused on trying to deliver exactly what you were looking for, and it seemed 95% of the time the first hit on a Google search was all you needed. The reason that's no longer true is that Google has figured out they make more money from ad revenue with bad search than with good search.

So how long before the AI companies figure out what Amazon and Google have figured out? That bad results make more money than good ones? Because you know they will.

It's important to realize that AI isn't programmed to deliver the truth, it's programmed to deliver the answer that its programmer wants it to give.

More broadly, the goal of AI isn't to deliver truth or even useful answers to the people using it, it's to make money for the people creating it.

 

[garry1p]

For now AI is data rich and information poor, that will change with time.

I saw on TV this mornings news, some talking head predicting 44% of today's low paying jobs would be replaced within five years by AI. No basis of how he figured that one out I would think it will be more than just low paying jobs.

Whatever the right number, it ain't going to be pretty.

 

[pmsmechanic]

Whatever the right number, it ain't going to be pretty.

Really? I could be wrong but it seems to me that AI has mostly been trained with the plethora of false information that has proliferated over the last years. Hopefully people will eventually catch on and all the AI company's will go broke.

I know.... wishful thinking on my part.

 

[scootr]

Is Siri now the dumbest AI ever. To her credit, she was the first AI most of ever interacted with, Bless her heart.... (you southern boys know what I'm sayin)

 

[WinterDeere]

For now AI is data rich and information poor, that will change with time.

Maybe a sign that the pace of progress is always increasing, but I think we've really turned a corner on this, in recent weeks. I agree, search engine results seemed to be on a downward trend most of the last two years, after reaching a respectably high level of performance over the preceeding decade.

But it seems these AI-driven results have been getting better almost day by day, after a lot of disappointing earlier results. I'm seeing better answers to searches I make today, versus similar searches made just a few weeks ago. It's changing fast!