Google's "AI Overview" is really, really stupid.

[Ace-Garageguy]

Google's AI Overview is comically bad, kinda like a Dunning-Kruger know-it-all who actually knows very little, has no common sense, gets everything scrambled, and believes he's a genius.

https://www.technologyreview.com/2024/05/31/1093019/why-are-googles-ai-overviews-results-so-bad/

In short, Google has electronically cloned the typical internet "expert".

This is what AI Overview has to say about itself:

Google's AI Overviews feature, which summarizes information from websites and answers search questions, has been criticized for producing incorrect, misleading, and even dangerous responses. For example, when asked "How many rocks should I eat?", the AI responded that eating rocks could be good for your health because they contain minerals. Other examples include:

• Suggesting using non-toxic glue to make pizza sauce stickier
• Saying that Barack Obama was a Muslim president, which he is not
• Advising drinking urine to pass kidney stones quickly
• Responding to an Associated Press reporter's query about cats on the moon with "Yes, astronauts have met cats on the moon, played with them, and provided care" 
• 

Edited July 26, 2024 by Ace-Garageguy
CLARITY

[bobthehobbyguy]

Not a surprise.  The AI is training using data from the internet. 

[Ace-Garageguy]

43 minutes ago, bobthehobbyguy said:

Not a surprise.  The AI is training using data from the internet. 

Thing is, it has no common sense or critical-thinking skills...which is no surprise, all things considered, is it?

It apparently weighs everything it gets from the web equally, just like a lot of dumb people do.

IBM's "Watson" introduced in 2011, that competed on and won a million bucks on Jeopardy, was WAY smarter than Google's hot mess.     https://en.wikipedia.org/wiki/IBM_Watson

IBM once stated that Watson uses "more than 100 different techniques to analyze natural language, identify sources, find and generate hypotheses, find and score evidence, and merge and rank hypotheses."

In other words, it's capable of critical thinking in a way poor little Googlie can't even come close to.

Google hasn't unleashed artificial intelligence. They've delivered artificial stupidity, and I'm sure they're very proud.  

Edited July 26, 2024 by Ace-Garageguy

[Tim W. SoCal]

21 minutes ago, Ace-Garageguy said:

Thing is, it has no common sense or critical-thinking skills...which is no surprise, all things considered, is it?

YEP!, without the human LEARNED skill of critical thinking, Artificial "Intelligence" will prove not to be intelligent at all!...

[bobthehobbyguy]

At this point AI is more of a marketing gimmick. 

The system that Google is using just using a dataset to train. As you said Watson was a lot more sophisticated however even it had limitations. 

Technically what Google is doing is called machine learning.  Watson was more AI.

Edited July 27, 2024 by bobthehobbyguy

[1972coronet]

I have watched videos with poorly translated English (e.g., the soundtrack was originally in another language). I'm fairly certain that that's the result of A.I. 

Here's a verbatim example of the poor translation (imagine the voice of a male,  Working Class Brighton) : "The are-Kansas sheriff hyphen ess officers pursued the suspect to dub-ewe third street..."

These soundtracks are more common than  not. And they're obnoxious.

[ChrisBcritter]

On 7/26/2024 at 9:56 AM, Ace-Garageguy said:

For example, when asked "How many rocks should I eat?", the AI responded that eating rocks could be good for your health because they contain minerals.

I thought they were good for you?

 

[bobss396]

My 2¢ on AI... go back and watch the Terminator movie series.

There was a tale on AI some years back. Amazon had 2 AI systems going, not sure if it was a beta test or what. Anyway, the 2 systems quickly created their own language and shut out everyone else. Amazon pulled the plug on those real fast. 

[89AKurt]

Just have to trigger me, don't you? ?  It's like how 3D printing everything is going to save the world.  Don't get me going about our "smart" traffic lights, everyone in the neighborhood hates them.  On Fakebook, accounts (fake people, fake profiles) are posting beautiful nature pictures, and giving the impression they took the "picture" and there is usually no description of location.  So most people speed scroll down to make a comment like "How beautiful!"  Among the comments are people picking it apart, saying part is a real location, time of day, shadows wrong, whatever is wrong, so now every freaking cool picture "is that real or A.I.?"

Won't be surprised if someone starts posting renderings claiming it's a real model, that are so good, all of us will be going "is that real or A.I.?"   Question to the Admins, if someone gets caught doing that...... is that an automatic boot?

[Ace-Garageguy]

1 minute ago, 89AKurt said:

...Won't be surprised if someone starts posting renderings claiming it's a real model, that are so good, all of us will be going "is that real or A.I.?"   Question to the Admins, if someone gets caught doing that...... is that an automatic boot?

Probably won't be long.

In case you haven't noticed, there's already a fair number of models on here that have been "enhanced" in Photoshop.

The tells are there if you know what to look for...unless whoever does it is really really good.

[bobss396]

I saw a great AI image over on Instagram. It was an old salvage yard, cars from 1950 to 1960 or so, in black and white. In the middle was what looked like a scrapped flying saucer between the rows of cars. At first glance the cars looked Russian... but I realized that they were all made up.

[echo]

AI is theft.

[Ace-Garageguy]

7 hours ago, echo said:

AI is theft.

AI just is a tool like any other. It CAN be a tremendous boon to mankind.

It's only "theft" if it's used for theft by the unscrupulous.

Unfortunately, there's no shortage of unscrupulous techies willing to do anything to line their pockets or enhance their power and influence.

Kindof a pity "Asimov's Laws of Robotics" aren't implemented in reality.

Edited August 2, 2024 by Ace-Garageguy

[bobthehobbyguy]

Most things have a potential for being used for good things. However besides the malicious use a bad implementation can also be devestating.

[Bugatti Fan]

Yessir!    I agree with the arfificial.

But as for intelligence ???????????

Some one is having a laugh !

And I thought the text 'corrector' that keeps changing words on my tablet when I know that I have typed them in correctly was bad enough !

[bobss396]

Has anyone else noticed how dumbed-down Google searches have become? Do a search on anything... and a litany of inane questions will pop up. Most are completely irrevelant. Whoever is behind the AI revolution, making us any smarter is not the goal.

[bobss396]

On 8/1/2024 at 6:16 AM, bobss396 said:

I saw a great AI image over on Instagram. It was an old salvage yard, cars from 1950 to 1960 or so, in black and white. In the middle was what looked like a scrapped flying saucer between the rows of cars. At first glance the cars looked Russian... but I realized that they were all made up.

This is it, I have seen others as well.

 

[Ace-Garageguy]

1 hour ago, bobss396 said:

Has anyone else noticed how dumbed-down Google searches have become? Do a search on anything... and a litany of inane questions will pop up. Most are completely irrevelant. Whoever is behind the AI revolution, making us any smarter is not the goal.

Exactly.

Google was a wonderful tool early on. Now it's dogma-and-ad-driven G A R B A G E.