Opinions on AI

[Count von Schwalbe]

Remember what atomic power was going to be used for and what was the very first thing it was used as.

It was literally developed for the single purpose of ending WW2. And thus it was used.

I have no doubts every scientific innovation mankind has ever made has also been used in war as soon as it was developed.

Sadly, most of them were driven by the need to outdo the other guy in a ongoing or threatened war.

Fission power was entirely theoretical until WWII; without the war it's entirely possible it would not have become a reality a decade later, or more. For all the horrors of war, it's incredible at driving technological progress.

Very much so. Many modern inventions have their roots in military inventions. Even the transistor itself came from radar research.

 

[TumbleGeorge]

It was literally developed for the single purpose of ending WW2

Not exactly but yes was used and with this purpose.

 

[R-T-B]

Not exactly but yes was used and with this purpose.

Well, for the American side that was largely the rationale behind the Manhatten Project. I believe the original target was a Japanese Pacific Naval base on the isle of Truk. This was scrapped of course when that base was overtook before the bomb was ready.

 

[Assimilator]

Well, for the American side that was largely the rationale behind the Manhatten Project. I believe the original target was a Japanese Pacific Naval base on the isle of Truk. This was scrapped of course when that base was overtook before the bomb was ready.

You could almost say it was overTruk.

 

[AusWolf]

Short answer: yes. Elaboration: I am a flawed animal but I'm self aware enough to notice some of those flaws and correct some of those I notice, therefore becoming ever so slightly better once in a while.

That doesn't apply to everyone, though. A lot of people (consciously) don't notice their own flaws, let alone try to correct them.

As for the equality of law, the problem is that people actively don't want it. Hypocrisy is one of the defining factors of humanity and it goes hand in hand with virtue signalling. People want to feel better than others, I'm sure that if this possibility was be limited, many people would aggressively fight against it, even if it would be better for the vast majority of them.
“Out of the crooked timber of humanity, no straight thing was ever made.” One can hope that the past tense is a foreshadowing.

I think the problem with the equality of law is that each and every case is different. You may break the law one day due to some genuine mistake, and you may regret it. It doesn't make you equal with someone who deliberately took the same action with no regret. You could say that AI should differentiate between the two cases, but the problem is that you need empathy for that, which is a quality only humans have.

While AI excels at specific tasks and processes vast amounts of data, being much faster, more accurate, and more objective than humans, it does have limitations. Artificial intelligence lacks common sense, contextual awareness, and the ability to deal with unexpected or new situations that humans can easily navigate.

Also, the ability to create new context. When you're asked to do something at work, but you have a better idea and do that instead. This is why "AI" is a badly misused term (at least the "intelligence" part).

 

[AusWolf]

Here's some more food for thought:

A mild spoiler: AI isn't always right.

 

[authorized]

Very simply put, i'm sick and tired of calling someone (a business/gov dept usually) to be faced with "press 1 for..., or press 2 for....and so on" hold-ups. I prefer real human interaction in every interaction. And thats not even AI so you can see where i'm going with this.

AI is already there, last time I called the bank I had to tell the "virtual assistant" what I wanted...

 

[the54thvoid]

It's not the software at fault.

This has come up time and time again - ChatGPT is a language model. Anybody using the output of chatGPT who doesn't do due diligence is at fault. The prompt box literally has a warning underneath it.

So, if a user was to ask for something that requires errorless output, they need to check what chatGPT says. It's basically a learning tool for its own end. The more it is asked, the more it assimilates data and then uses algorithms to further parse the info in future conversations.

 

[Assimilator]

It's not the software at fault.

This has come up time and time again - ChatGPT is a language model. Anybody using the output of chatGPT who doesn't do due diligence is at fault. The prompt box literally has a warning underneath it.

View attachment 317009


So, if a user was to ask for something that requires errorless output, they need to check what chatGPT says. It's basically a learning tool for its own end. The more it is asked, the more it assimilates data and then uses algorithms to further parse the info in future conversations.

You and I know this, but the average person doesn't. And that's because ChatGPT isn't being sold as fallible in the press releases and news articles that heap praise on it. That's intentional, because marketing always tends towards deception, but it's extremely dangerous. And therein lies my primary concern - that the average person is being led to place their trust in something that is inherently flawed and will ultimately give them an incorrect answer, potentially at the absolutely worst time.

 

[Deleted member 229121]

Yep. Not too far off from how GPS will occasionally send someone dumb enough right off a cliff.

 

[R-T-B]

Yep. Not too far off from how GPS will occasionally send someone dumb enough right off a cliff.

It's way worse than that tbh.

GPS these days is accurate to within a very small radius, usually like 10m. While you can go off a cliff in that vicinity it's pretty unlikely unless you are reckless. ChatGPT, in standard usage scenarios, can be massively off/wrong, and it's not really uncommon.

 

[Super Firm Tofu]

Yep. Not too far off from how GPS will occasionally send someone dumb enough right off a cliff.

It's way worse than that tbh.

GPS these days is accurate to within a very small radius, usually like 10m. While you can go off a cliff in that vicinity it's pretty unlikely unless you are reckless. ChatGPT, in standard usage scenarios, can be massively off/wrong, and it's not really uncommon.

Not so much GPS, but Google Maps

He Drove Into a Creek and Died. His Family Blames Google Maps. (Published 2023)

Philip Paxson’s family said in a lawsuit that he was following Google Maps directions last year when he drove off a collapsed bridge in Hickory, N.C., plunged into a creek and drowned.

www.nytimes.com

People just follow blindly what they're told.

So many things can be totally safe and helpful until you add the human element.

 

[thesmokingman]

Yep. Not too far off from how GPS will occasionally send someone dumb enough right off a cliff.

Just make sure you're driving a Tesla lol.

Spoiler: off cliff

Two adults and two children rescued after #Tesla plunges off cliff in #California

After a #Tesla plunged 250 feet off a cliff on the Pacific Coast Highway in #California, first responders used a helicopter to help rescue survivors. #NBCNew...

www.youtube.com

 

[R-T-B]

Teslas ironically, use "AI" in their infamous self driving nonsense, so yeah, that's pretty ontopic.

 

[Assimilator]

It's way worse than that tbh.

GPS these days is accurate to within a very small radius, usually like 10m. While you can go off a cliff in that vicinity it's pretty unlikely unless you are reckless. ChatGPT, in standard usage scenarios, can be massively off/wrong, and it's not really uncommon.

The follow-on problem is that ChatGPT isn't just wrong, it's confidently wrong. Rather than responding with "sorry, I don't know" it'll make up whatever stupid bullshit it can, and present that bullshit as if it's the truth. This is compounded by the fact that quite often its bullshit is almost correct, except for one error that invalidates the whole thing. And because it's being setup as this oracle of truth, people are going to be seriously caught out when they trust it and it gives them bullshit back.

 

[R-T-B]

The follow-on problem is that ChatGPT isn't just wrong, it's confidently wrong. Rather than responding with "sorry, I don't know" it'll make up whatever stupid bullshit it can, and present that bullshit as if it's the truth. This is compounded by the fact that quite often its bullshit is almost correct, except for one error that invalidates the whole thing. And because it's being setup as this oracle of truth, people are going to be seriously caught out when they trust it and it gives them bullshit back.

That is a major danger yes. I wish people would take it more seriously because all it takes is one person trusting it too much to lead to literal casualities.

 

[AsRock]

An AI is only as good as it's dataset / trainer.

(Hint, that's humans).

So it's bound to always have error's ?.

 

[R-T-B]

So it's bound to always have error's ?.

At least the current tech we are using? Yep.

 

[the54thvoid]

In a strange sense, a true AI would need to inhabit a physical form, or at least have access to observe things beyond it's physical base. Feeding info to a subject, be it human or machine, is more or less indoctrination. A true AI would need to be able to remove itself from such influence, and 'discover' its own truth. We're a very long way from creating such an autonomous intelligence.

 

[R-T-B]

In a strange sense, a true AI would need to inhabit a physical form, or at least have access to observe things beyond it's physical base. Feeding info to a subject, be it human or machine, is more or less indoctrination. A true AI would need to be able to remove itself from such influence, and 'discover' its own truth. We're a very long way from creating such an autonomous intelligence.

I agree with that assessment completely.

 

[Assimilator]

Feeding info to a subject, be it human or machine, is more or less indoctrination.

What about when parents keep details of all the evil things in this world from children? Is that also indoctrination?

 

[lexluthermiester]

A mild spoiler: AI isn't always right.

Was there any doubt of this?

What about when parents keep details of all the evil things in this world from children? Is that also indoctrination?

No, that's sheltering, a completely separate and distinct problem and very nearly as bad.

 

[AusWolf]

The follow-on problem is that ChatGPT isn't just wrong, it's confidently wrong. Rather than responding with "sorry, I don't know" it'll make up whatever stupid bullshit it can, and present that bullshit as if it's the truth. This is compounded by the fact that quite often its bullshit is almost correct, except for one error that invalidates the whole thing. And because it's being setup as this oracle of truth, people are going to be seriously caught out when they trust it and it gives them bullshit back.

Are you saying that ChatGPT is a politician?

Was there any doubt of this?

Not on my part, but in the general public, I would assume so.

 

[sLowEnd]

AI has massive potential for both good and bad. I can't see if it'll be a net positive or a net negative for humanity as a whole, but I hear it's a huge deal for the field of medicine. If it can help eradicate, or produce effective treatments for previously incurable illnesses, then that would be fantastic.

 

[Count von Schwalbe]

Benevolent synthetic intelligence created - leads to the singularity and saves humanity from ourselves. My hope.

Something interesting for you:
"Mechanocracy" by Miles J. Breuer. A very thought-provoking book, even if it is a work of fiction 90 years ago.

Should be public domain in most countries but I refrained from posting a link just in case.