What AI isn't

[Concerned American]

That's true I type on my keyboard, AI doesn't do that for me, I'm human and human error always exists.

LOL, apparently you missed the AI supported spell check. BTW, isn't it wonderful how AI has the ability to paraphrase what I already said. If you had been educated, you would have been able to do the same---with properly spelled English words.

 

[Churchill]

LOL, apparently you missed the AI supported spell check. BTW, isn't it wonderful how AI has the ability to paraphrase what I already said. If you had been educated, you would have been able to do the same---with properly spelled English words.

Before critiquing others, you might want to tighten up your own grammar.

 

[there4eyeM]

One thing "artificial intelligence" is, by definition, is artificial ignorance. Whatever it might artificially "know", there is everything else it, artificially" ignores.

 

[Bleipriester]

Can you explain what "real" artificial means?

It exists.

Well all digital computation is based on logic. But what has logic go to do with intelligence?

It now can make its own decisions. Prior to that AI simply consisted of pre-defined "what if"-scenarios with pre-defined reactions.
For example, you cannot simply take one program´s AI and use it in another. But using an LLM, you can. That is the difference.

I use AI all the time, there are things is truy helps with and are impressive but there are also mistakes that are no obvious sometimes too.

Yes, it makes many mistakes. It is still in development.

 

[Bleipriester]

Do you think we act idependently? or are we deterministic? Digital conmputation is deterministic, no "free will" or anything like that, I agree it can sometimes appear to be but it isn't.

Not, it isn´t. It just emulates: artificial.

 

[NewsVine_Mariyam]

It now can make its own decisions. Prior to that AI simply consisted of pre-defined "what if"-scenarios with pre-defined reactions.
For example, you cannot simply take one program´s AI and use it in another. But using an LLM, you can. That is the difference.

The bolded sectioin above is not correct.

 

[CrusaderFrank]

I might, someday, become a fan of AI. For that to happen, though, we would have to actually have some real AI. As of yet, we do not though I am hearing of strides taken in that direction.

What we have now is a potent mix of a Mark Twain impersonator who has read all of the man's writings, public and private, and he also has access to the unfiltered and unmediated internet. He might be able to pass as Mark Twain but he hasn't the brain with which to think not like Twain.

But what do I keep seeing online? People post "answers" or posts on platforms which are cut and paste jobs from AI, and the purpose seems to be to make an authoritative statement as bolstered by AI. Except that there is no intelligence there so the fact that the engine used statistics to build sentences that were consistent with earlier sentences is useless. I see these "executive summaries" which have, baked in, errors, misunderstandings and biased content. Using wikipedia at least promises the illusion of fact checking, but slapping 2 pages of astute political insight generated by an LLM provides nothing in the way of sources.

So I am doubly annoyed -- first that people should think that these pastiches are at all intelligence and second that anyone would generate one and use it as a res ipsa loquitur kind of authority and rely on it.

LLM is good for aggregating news and musing regarding a particular topic, but is not proof.

It also has been caught outright lying

 

[Bleipriester]

The bolded sectioin above is not correct.

Yes, it is. If not, deliver an explanation.

 

[Churchill]

Not, it isn´t. It just emulates: artificial.

How can you emulate free will?

 

[Churchill]

It exists.



It now can make its own decisions.

Computers cannot make decisions.

Prior to that AI simply consisted of pre-defined "what if"-scenarios with pre-defined reactions.
For example, you cannot simply take one program´s AI and use it in another. But using an LLM, you can. That is the difference.



Yes, it makes many mistakes. It is still in development.

 

[Bleipriester]

How can you emulate free will?

For now that remains their secret. But it happens.

 

[Bleipriester]

Computers cannot make decisions.

The LLMs do. There is no doubt about that.

 

[Churchill]

For now that remains their secret. But it happens.

How do you know? how did you convince yourself of this?

 

[Churchill]

The LLMs do. There is no doubt about that.

No they do not, digital computers follow rules, there is no capacity for choice, whatever output the produce will always be deterministic.

These systems all all algorithmic and algorithms do not, cannot "choose" what to do, they always just follow the rules.

I've spent fifty years working in this area. Take any programming language and try to write code where the system can choose, rather than just follow rules, you can't. There's no concept of "choose what to do next".

 

[Bleipriester]

How do you know? how did you convince yourself of this?

Talking to the bots. Reading stuff about them.

 

[Churchill]

Talking to the bots. Reading stuff about them.

You were misled then or you misread something.

 

[there4eyeM]

It should probably be called simulated intelligence.

 

[Bleipriester]

No they do not, digital computers follow rules, there is no capacity for choice, whatever output the produce will always be deterministic.

These systems all all algorithmic and algorithms do not, cannot "choose" what to do, they always just follow the rules.

I've spent fifty years working in this area. Take any programming language and try to write code where the system can choose, rather than just follow rules, you can't. There's no concept of "choose what to do next".

The LLMs do. Your fifty years lacked the hardware required to run such a LLM so nobody even tried. But now times have changed. AI is even used in war. It decides what target to attack.

 

[Bleipriester]

You were misled then or you misread something.

No. Making decisions is the whole concept of these LLMs. It is evolution of software.

 

[rosends]

The LLMs do. There is no doubt about that.

I'm not sure they make decisions. That would indicate that they have the power to choose anything else. It isn't a choice as much as a result of an algorithm. If the statistics indicate that based on samples, there is a high probability that the word "jump" would be the next in a sentence, the computer cannot choose NOT to use that word based on a feeling.