How can AI benefit you?

There are truly incredibly negative notions about AI in this forum. It's like saying the word 'software' and that word is immediately equated with 'malware'. It's really that bad, I don't exaggerate at all.

So, I feel compelled to put in a few good words for AI, here is one.

------------------
Hop in MS' AI car (Copilot) and you will be able to move faster and farther than you can ever run on your own two legs

I think there are serious problems with current implementations of AI and that, I think, is only to be expected with such an emergent technology.

Haydon said:

Hop in MS' AI car (Copilot) and you will be able to move faster and farther than you can ever run on your own two legs

Click to expand...

is exactly the kind of BS that provokes negative responses.

And that drawing of a man having to sprint along the motorway desperately trying to catch up with the car that is supposed to be doing what he wants could well be regarded as an accurate depiction of the current state of play.


All the best,
Denis

Look at the state of the world. It is human intelligence we need.

Clue: real cream mmmmm, artificial cream yukkk

Anything artificial is only ever a pale shadow of the original.

The term AI is just marketing bs and anybody who believes otherwise has been conned.

I've experimented with using "Copilot" as an alternative to a Web search. The point was to be able to pose a question in natural language, rather than trying to figure out the right keywords.

The result for technical questions was likely to be wrong. There seem to be no internal checks as to whether the online data found was particular to the asked question. (Like giving an Intel-only response to a question about an AMD motherboard.)

I hope that if physicians begin using AI, there will be more safeguards. (Seems likely. There must be plenty of malpractice attorneys out there.)

I asked Copilot for a cross between a lotus and a rose, see below.

I myself would not be able to produce the drawing, but Copilot produced it in a few seconds. It would take horticulturalists years to produce the cross, if they ever can.

That's another example of how AI benefits me and others.

bobkn said:

I've experimented with using "Copilot" as an alternative to a Web search. The point was to be able to pose a question in natural language, rather than trying to figure out the right keywords.

The result for technical questions was likely to be wrong. There seem to be no internal checks as to whether the online data found was particular to the asked question. (Like giving an Intel-only response to a question about an AMD motherboard.)

I hope that if physicians begin using AI, there will be more safeguards. (Seems likely. There must be plenty of malpractice attorneys out there.)

Click to expand...

A solution usually requires multiple iterations. In the rare case that a solution only requires a single iteration, then a single search would suffice.

But if as usual the solution requires multiple iterations, then you should chat with Copilot, i.e. you should challenge Copilot to provide the next iteration, then the next and the next and the next, etc.

Copilot does commit inaccuracies, although outright mistakes are rare in my experience. But it is the chat that can be used to dig deeper, broader or narrower or whatever is necessary.

In short, it does take skill to drive MS' AI car, the Copilot.

I've been using AI that's in Photoshop and I've gotten some amazing results.

IMHO, It's taking away our ability to think on our own. I remember when the pocket calculator first came out. I was in Junior High (now called middle school) and we weren't allowed to use them. Why? Because the teachers wanted us to be able to do the math in our heads. I still hold on to that reason. I do use computers to get my work done but I don't want them to do my thinking.

AI is a disruptive technology, and can/will cause ripples in society, just like the car did when it was new. Legislation, like the highway traffic act, has to catch up, for example.

I don't think that the above is a real danger, though. The more serious issue is below, IMHO.

AI is neither good nor bad per se. It is the billions of AI use cases, good and bad, that determine whether there is a net positive. And there will be a net positive simply because of self-preservation of humanity, can't be any different.

Haydon said:

AI is neither good nor bad per se

Click to expand...

True.
But that Mal AI picture is inaccurate.
By definition, AI can't fight itself.

@Winuser Let AI enhance what you can do, just like @BunnyJ enhanced his Photoshop results.

@Ghot AI has good uses and bad uses, that's what I meant with net positive (I will edit my post in the above, if it still can be edited) Good AI is already used to crush bad AI.

Haydon said:

@Winuser Let AI enhance what you can do, just like @BunnyJ enhanced his Photoshop results.

@Ghot AI has good uses and bad uses, that's what I meant with net positive (I will edit my post in the above, if it still can be edited) Good AI is already used to crush bad AI.

Click to expand...

I found this picture via Google and I expanded it with Photoshop AI and I added the blackbirds with AI,

Haydon said:

A solution usually requires multiple iterations. In the rare case that a solution only requires a single iteration, then a single search would suffice.

But if as usual the solution requires multiple iterations, then you should chat with Copilot, i.e. you should challenge Copilot to provide the next iteration, then the next and the next and the next, etc.

Copilot does commit inaccuracies, although outright mistakes are rare in my experience. But it is the chat that can be used to dig deeper, broader or narrower or whatever is necessary.

In short, it does take skill to drive MS' AI car, the Copilot.

Click to expand...

I have provided corrections to Copilot when I get a bogus result. Copilot has acknowledged the error.

But if I try the same search later, the correction doesn't stick. Apparently, Copilot isn't configured to "learn" from corrections provided by users.

I can only guess that if you don't get seriously false information from Copilot, you are doing something other than asking narrow technical questions.

Example: suppose that I asked about RAM for an AMD TRX50 motherboard. Copilot offered some DDR5 RAM that would have been good for an Intel Z790 board. Sadly, the TRX50 requires RDIMMs. The suggested UDIMMs wouldn't work at all. I'd call that an outright mistake.

Your last sentence is a fairly serious insult.

Haydon said:

Hop in MS' AI car (Copilot) and you will be able to move faster and farther than you can ever run on your own two legs

Click to expand...

I can hope in my car and do the same thing. IMHO that is not a good example of AI.

@bobkn Copilot is a tool, and as any other tool, it requires skill to use it to get the best results. Why is making such an obvious statement a 'fairly serious insult' ?

Copilot is still under (very, very rapid) development. One of the things that it still cannot do, is to change its parameters interactively.

Haydon said:

@Winuser Let AI enhance what you can do, just like @BunnyJ enhanced his Photoshop results.

Click to expand...

Why? I don't need any pictures or works of art created with AI hanging on my walls. I also don't need AI doing my thinking for me. As a test I used Copilot to create a picture of a buffalo and a love letter to my wife. I didn't keep either one of them. I'm content in the fact that I'm a lousy artist and letter writer.

Haydon said:

@bobkn Copilot is a tool, and as any other tool, it requires skill to use it to get the best results. Why is making such an obvious statement a 'fairly serious insult' ?

Copilot is still under (very, very rapid) development. One of the things that it still cannot do, is to change its parameters interactively.

Click to expand...

Maybe MS should wait until AI is functioning properly, before releasing it.

MS already released a similar product 2-3 years ago. It didn't last more than a few months before it went drastically... off the beam. It started swearing all the time and got seriously racist as well. They pulled it rather quickly after that.

I have no problems with "new" software... that works properly.
I do have issues with software that's half baked, and is released as working.

The OP provides a metaphor of how AI can benefit you. Here is another one that can benefit garden designers by providing design inspirations that they can monetize, for example.

Now if there was only somewhere to "buy" fish that swim in the air.
And... that's the first broken rainbow I've ever seen.
And the little girl has three legs.

Haydon said:

@bobkn Copilot is a tool, and as any other tool, it requires skill to use it to get the best results. Why is making such an obvious statement a 'fairly serious insult' ?

Copilot is still under (very, very rapid) development. One of the things that it still cannot do, it change it parameters interactively.

Click to expand...

You don't feel that blaming Copilot's apparent shortcomings on PEBKAC is insulting? That's an interesting lack of empathy.

Copilot will have to do a lot better in providing correct answers to highly specific questions before I'll begin to regard it as a tool rather than a toy. Modifying itself when receiving corrections (after acknowledging the corrections are accurate) would be a major step in reliability.