Recent

Author Topic: AI usefulness discussion  (Read 1953 times)

Joanna from IRC

  • Hero Member
  • *****
  • Posts: 1174
Re: AI usefulness discussion
« Reply #15 on: October 31, 2024, 12:33:32 am »
When people talk about AI, I always remember an old story.

It was 1983 and I was reading a book by a chess grandmaster. The grandmaster analyzed in the book a game in which the machine lost miserably against a good human player. Then, he explained why a machine could never play chess better than the world champion. Obviously, he was wrong.

My opinion is that, whatever a computer can do today, it may do it inaccurately, slowly, with errors, etc. But one day it will do it much better than any human. Over time, techniques and algorithms will be improved and refined, and we will have much greater computing power.

I do not dare to say that a computer will never program better than a human programmer.

The reason computers work for chess is because it’s very predictable. The pieces always move in certain ways without fail. The board is always the same size and never changes shape. Real life is not like that. Unexpected unforeseen things happen.
✨ 🙋🏻‍♀️ More Pascal enthusiasts are needed on IRC .. https://libera.chat/guides/ IRC.LIBERA.CHAT  Ports [6667 plaintext ] or [6697 secure] channel #fpc  #pascal Please private Message me if you have any questions or need assistance. 💁🏻‍♀️

Bogen85

  • Hero Member
  • *****
  • Posts: 661
Re: Re: Lazarus and Copilot
« Reply #16 on: October 31, 2024, 12:59:08 am »
I want to talk to live people about pascal not to a glorified search engine ai.
Why would you want to replace friends with AI? That’s really disturbing. Do you not have any friends to chat with about code? I don’t know everything but I’ll be your friend if you have no one else.

Wow. I don't know how to even answer that.

The fraction of my time I spend using AI tools to minor solutions is a small fraction of the times I spend discussing with people electronically about programming and applications, and those 2 combined are a small fraction of my time compared to talking with people face to face about programming and applications. I'm in a very high tech sector and we are experienced and intelligent enough (and paid enough, for good reason) to realize a tool is wasting our time, rather than saving us time, especially with so much accountability with everything we do.

To think we are going to waste hours of each others precious time coming up with solutions to minor corner case programs when we can use an AI tool for those issues and save all involved a lot of time.

I don't know how else to say it Joanna, but I will reiterate along the lines of what I've said before: "your computer usage world view is very narrow. How can you not comprehend there is a world of interacting with other humans on solving programming and software problems that is outside of electronic communications in one small sector of the computing world?"

Pascal is a outside of work hobby for me that I find enjoyable, because I like Free Pascal and this community. But it is not my entire life. And Free Pascal is one of many programming languages I enjoy. But most of my programming life and interaction with other people is in person with them at work. I find AI tools to be useful both for certain corner case problems both at work and for hobby programming and other computer related stuff.

And in no way should anyone here be expected to say or defend how they use their free time, especially when it is a hobby in their spare time to help them solve some of their own personal computing needs.



« Last Edit: October 31, 2024, 01:54:06 am by Bogen85 »

Curt Carpenter

  • Hero Member
  • *****
  • Posts: 547
Re: AI usefulness discussion
« Reply #17 on: October 31, 2024, 02:01:17 am »
The reason computers work for chess is because it’s very predictable. The pieces always move in certain ways without fail. The board is always the same size and never changes shape. Real life is not like that. Unexpected unforeseen things happen.

But the other player is no more fixed than you or I -- and "real life" also has some fixed Contextuality --  limited habitat, mother and father, death...

There are people in the world with no one to talk to for one reason or another:  is letting them talk to a machine a bad thing?  I don't know.  I do know the last person I talked programming (math actually) with face to face died two years ago.   Ecclesiastes 9:11.

Joanna from IRC

  • Hero Member
  • *****
  • Posts: 1174
Re: AI usefulness discussion
« Reply #18 on: October 31, 2024, 02:13:49 am »
The reason computers work for chess is because it’s very predictable. The pieces always move in certain ways without fail. The board is always the same size and never changes shape. Real life is not like that. Unexpected unforeseen things happen.
But the other player is no more fixed than you or I -- and "real life" also has some fixed Contextuality --  limited habitat, mother and father, death...

There are people in the world with no one to talk to for one reason or another:  is letting them talk to a machine a bad thing?  I don't know.  I do know the last person I talked programming (math actually) with face to face died two years ago.   Ecclesiastes 9:11.
Yes the other player is a human whose behavior is not constrained by simple rules. The other human could move pieces when you aren’t looking or even physically attack you but then the game is not chess anymore in the strictest sense .

I agree that there are lonely isolated people who have no one to talk to but the solution is not giving them machines to talk to. What a cruel joke to play on people who need companionship. I try to provide a provide a place where interesting conversations can be had. It’s all I can do.

I’m sorry to hear that you lost a good chatting companion. I wish I could help.
✨ 🙋🏻‍♀️ More Pascal enthusiasts are needed on IRC .. https://libera.chat/guides/ IRC.LIBERA.CHAT  Ports [6667 plaintext ] or [6697 secure] channel #fpc  #pascal Please private Message me if you have any questions or need assistance. 💁🏻‍♀️

440bx

  • Hero Member
  • *****
  • Posts: 4647
Re: AI usefulness discussion
« Reply #19 on: October 31, 2024, 02:41:23 am »
There is one thing A.I can and will do better than humans and that is, solve problems that can be solved using brute force.  It will never stop getting better and better at that (courtesy of ever increasing computing power.)

There is one thing A.I will _never_ be good at and that is any task that _requires_ intelligence and/or creativity.  Brute force can simulate those to a minor extent but, it is totally unable to _understand_ anything.

For instance, don't expect any groundbreaking mathematical proof to _ever_ be "discovered" by an A.I engine ("proofs" by exhausting all possibilities are _not_ proofs.)  It won't happen because the engine has _no_ understanding whatsoever of what it's doing.  A chimpanzee is more likely to offer a new proof of Fermat's Last Theorem than an A.I engine ever will (presuming humans will never be able to create a genuinely intelligent machine, in which case the intelligence would no longer be artificial.)

For some things A.I can be useful (those that don't require intelligence.)

(FPC v3.0.4 and Lazarus 1.8.2) or (FPC v3.2.2 and Lazarus v3.2) on Windows 7 SP1 64bit.

TRon

  • Hero Member
  • *****
  • Posts: 3463
Re: AI usefulness discussion
« Reply #20 on: October 31, 2024, 02:42:38 am »
sorry about short-cutting specific parts of your reply Martin (and this reply is also not aimed only at you but answered in a more generic manner to be able to share some thoughts, same applies for Curt)

Its not the tool, its how it is being used.
True. But have you (or anyone else interested) ever wondered why google search is falling apart ?

This fact is already proven by experts and is merely one of the shortcomings of how current generative AI is implemented and how it is being deployed.

Because the basics of current implementations are the same (feel free to correct me there because I do not keep a close eye on current developments), they will all suffer the same faith.

I do not know if it is relevant for what you had envisioned yourself.


But other's experiences may differ.
My experience seems to be more or less the same as yours.

It takes me more time to instruct than it is to actually write the code myself.


If looking at it as one giant database to extract data from then all I can say to that is that any proper cms should be able to do the same and do it quite more efficient (time-wise and energy wise).

I am not sure but I believe that you Martin_fr are familiar with code-tools (internals) ? so perhaps you might be able to share thought on this.

Ask the question, would it be more efficient to use code-tools to help you write code (f.e. selecting which variable or method you
want to address next or rely on an AI that analyzes the code, index it, and predict which the next 'thing' on the current code-line will be) ?

And with efficient I mean, on every level of efficiency (through the entire/complete chain).

And with that I conclude: There is a reason big tech wants to sell you their AI fairy-tales and hardware.

Funny how everything boils down to why  ;)


I was working when the first "AI" wave struck.  The company I worked for got very involved and it took two years to put out the fire and
get people to take a realistic look at the technology. But this wave is genuinely different I think.
It is. The generic population got educated from all media outlets and formed 'a picture' of what AI is/means to them (which differs from person to person).

This picture is/was usually formed when this population had no notion whatsoever of what machine learning is and things like chatGPT and openAI did not even existed yet. It usually is a overly romanticized picture.

This picture was then projected to/for those that started using the acronym AI for their projects. It is easy for these companies
to feed this picture with all sort of claims that can never be realized with current technology. But people usually are in nature gullible.

Hence, society has created a perfect storm. And it is very difficult to get rid of it (if at all).

It's a mistake to dismiss results like that this time.  It's not the LLM stuff, but the machine learning stuff behind the LLM stuff.
Oh, I do not dismiss anything (I would not even dare).

Current implementation allows for cross-referencing data and makes it plausible that some of that data correlates
with other data (which nobody thought of to do before because... nobody asked the why (not) question, only the how)

Any good cms and any functional brain is able to do the same  :)


That said:  I just had some work done by an attorney.
....
Again, any good cms is able to do the same. But you are correct that this is a profession where they protect their business (like hawks).

Maybe that explains why the legal community is so interested in "regulating" the technology!
;D

Don't forget that even if it is their profession to deny any responsibility with such kind of things that you talked about but in the end there must be someone that can be held accountable (obvious reasons).


And on a general note: Note that we did not even got deeper into things such as equality, discrimination, accountability and data
acquisition and manipulation which are a real treat.

But, I have to guess: Nobody seem to care about the why, only the how  :)

fwiw: I do not claim to be an expert on any of these topics. I just keep my eyes and ears open and try to get myself educated on the subject (there are many good lectures from experts that explain how machine learning and language models work even though I might perhaps not always fully grasp every technical detail).
« Last Edit: October 31, 2024, 03:30:21 am by TRon »
This tagline is powered by AI

Joanna from IRC

  • Hero Member
  • *****
  • Posts: 1174
Re: AI usefulness discussion
« Reply #21 on: October 31, 2024, 03:22:29 am »
✨ 🙋🏻‍♀️ More Pascal enthusiasts are needed on IRC .. https://libera.chat/guides/ IRC.LIBERA.CHAT  Ports [6667 plaintext ] or [6697 secure] channel #fpc  #pascal Please private Message me if you have any questions or need assistance. 💁🏻‍♀️

Curt Carpenter

  • Hero Member
  • *****
  • Posts: 547
Re: AI usefulness discussion
« Reply #22 on: October 31, 2024, 04:01:26 am »
For instance, don't expect any groundbreaking mathematical proof to _ever_ be "discovered" by an A.I engine ("proofs" by exhausting all possibilities are _not_ proofs.)  It won't happen because the engine has _no_ understanding whatsoever of what it's doing.  A chimpanzee is more likely to offer a new proof of Fermat's Last Theorem than an A.I engine ever will (presuming humans will never be able to create a genuinely intelligent machine, in which case the intelligence would no longer be artificial.)

For a wide examination of AI in mathematics that is very readable, let me suggest "The Unreasonable Ineffectiveness of AI for Math" here:   
https://synthesis.ai/2024/02/13/the-unreasonable-ineffectiveness-of-ai-for-math/

Granted the author has a stake in the debate.  But never mind:  my favorite passage in the whole piece is this one:

Quote
In a paper on the Robbins problem, Louis Kauffman sums this conundrum up as follows: “Can a computer discover the proof of a theorem in mathematics?.. I say that a proof is not a proof until a person is convinced by it. In fact a mathematical proof is exactly an argument that is completely convincing to  a mathematician!   

What an all-too-human "definition" of what a proof actually is  :D. I drips of our hubris like a kind of intellectual sweat and exposes our vanity -- our need to place ourselves at the top of the "intelligence chain" just as we seem to be at the top of the food chain. 

And it's statements like Kauffman's that I think hint at the most important thing about today's "AI:"  that the technology is telling us more about ourselves than it is about the the nature of machines and their technology.


Weiss

  • Full Member
  • ***
  • Posts: 180
Re: AI usefulness discussion
« Reply #23 on: October 31, 2024, 04:55:46 am »
Lazarus, or any other good IDE, already has a lot of intelligence built into it. All those templates, suggestions, and other routine mundane tasks are handled smoothly. This is where AI belongs, and it is there already. There is no need for another layer of "generative AI" which can only create a jumble of known (and wrong ones) facts.

So far, everything I made, was a solution to certain problems. To develop a solution, one needs a vision, dream if you will. Current AI does not have that. Besides, I do not think this is here to stay. It is just a fluke, commercial product, like chocolate bar.

440bx

  • Hero Member
  • *****
  • Posts: 4647
Re: AI usefulness discussion
« Reply #24 on: October 31, 2024, 05:29:59 am »
I drips of our hubris like a kind of intellectual sweat and exposes our vanity
That is so wrong. A _valid_ mathematical proof is a sequence of facts, each of which cannot be disputed that lead to an inevitable conclusion.  The only "human" thing in there is that we (well... some) can understand such a sequence of facts, unlike artificial intelligence which doesn't understand anything at all.

Friendly suggestion: Leonhard Euler has some of the most amazing proofs ever devised in Mathematics.  Some of his proofs are some of the most beautiful logical thinking a human being has authored.  I've studied his proofs not because of the mathematics involved but to study his way of thinking about solving a problem.  He is/was the Himalayas of intelligence.

A.I is a genuinely _sad_ "thing" compared to how that incredible mind worked. Compared to Euler, not even a wannabe molehill.

Disclaimer: I have a personal bias, of all the incredible mathematicians in history, I believe Leonhard Euler to be at the top, even on top of Carl Friedrich Gauss.  I am full agreement with Pierre-Simon Laplace: "Read Euler, read Euler, he is the master of us all".    That said, I cannot help having an immense amount of respect and admiration for history's great mathematicians.
(FPC v3.0.4 and Lazarus 1.8.2) or (FPC v3.2.2 and Lazarus v3.2) on Windows 7 SP1 64bit.

LV

  • Full Member
  • ***
  • Posts: 130
Re: AI usefulness discussion
« Reply #25 on: October 31, 2024, 07:30:24 am »
I drips of our hubris like a kind of intellectual sweat and exposes our vanity
That is so wrong. A _valid_ mathematical proof is a sequence of facts, each of which cannot be disputed that lead to an inevitable conclusion.  The only "human" thing in there is that we (well... some) can understand such a sequence of facts, unlike artificial intelligence which doesn't understand anything at all.

It seems that Gödel's incompleteness theorems did not add to our optimism that every statement can be proved or disproved. :(

440bx

  • Hero Member
  • *****
  • Posts: 4647
Re: AI usefulness discussion
« Reply #26 on: October 31, 2024, 07:40:19 am »
It seems that Gödel's incompleteness theorems did not add to our optimism that every statement can be proved or disproved. :(
That some statements cannot be proved does not mean that there aren't statement that _can_ be proved.

Euclid started that mess with his axioms in Elements.  <chuckle>
(FPC v3.0.4 and Lazarus 1.8.2) or (FPC v3.2.2 and Lazarus v3.2) on Windows 7 SP1 64bit.

VisualLab

  • Hero Member
  • *****
  • Posts: 518
Re: AI usefulness discussion
« Reply #27 on: October 31, 2024, 10:23:05 am »
We have to wait until the current third wave of AI hits the "technological wall" (like the previous ones). Each of these waves has moved a little further each time (this is the visible progress in AI operation). As some people on the forum see it, the current AI will not replace programmers, but at most can provide a little support to those who have a little less knowledge. It is basically a "supercharged search engine on steroids". However, the current AI will not write, for example, an OS kernel, a useful compiler or even drivers for a device for a selected OS. At most, it can glue code together from existing fragments (without a guarantee that it will compile). Meanwhile, people creating some (engineering, scientific) types of software need a completely new, fresh look at the problem. And the current AI won't come up with anything. We simply have to wait for the next waves of AI. And this probably will not happen in our lifetime.

The problem is that many people see current AI solutions as a miracle cure for every problem (morbid enthusiasts). There are much fewer hardened AI opponents (fierce opponents). Typically, decision-makers in corporations do not belong to any of these groups. They are simply greedy tricksters who hope that after a few years of investing in AI, they will finally start raking in the "big bucks". That is why they keep the AI hype going. The most reasonable attitude when it comes to AI is to be a moderate skeptic. If we look at the scientific and technical development of humanity, this attitude has always been the most beneficial. Current AI can be useful, but for now, it has rather narrow applications.

To sum up: AI should not be condemned as if it were some kind of "IT Satan" corrupting people, but neither should we sing its praises as if it were an "IT saviour" of humanity. These are extremes. AI should be approached with a reasonable, moderate skepticism, asking questions and verifying the possibilities. And these are not spectacular at the moment.

marcov

  • Administrator
  • Hero Member
  • *
  • Posts: 11818
  • FPC developer.
Re: AI usefulness discussion
« Reply #28 on: October 31, 2024, 10:32:45 am »
Google uses AI to write 25% of new code:

https://www.tomshardware.com/tech-industry/artificial-intelligence/google-now-uses-ai-to-write-25-percent-of-its-new-code-alphabet-ceo-sundar-pichai-underlines-the-companys-role-in-the-ai-industry-amidst-strong-q3-24-financials

They don't say how much of their code was generated before, so it might just be they added a bit AI to existing codegenerators.

That said, I do occasionally see people use voice assistant on their mobiles (and it makes some sense giving the horrible input device options), and of course x86 processors branchprediction and Nvidia's dlss is also AI based. (though that seems to be more a classical neural network than large sparse correlation matrices)

Bogen85

  • Hero Member
  • *****
  • Posts: 661
Re: AI usefulness discussion
« Reply #29 on: October 31, 2024, 11:49:00 am »

But other's experiences may differ.
My experience seems to be more or less the same as yours.

It takes me more time to instruct than it is to actually write the code myself.

For me the AI tools help me get over the hump on some new (or coming back to after many years) complex programming or software setup task.
(minutes vs hours, or hours vs days, or a half a day vs weeks)

Once I'm over the hump most gets discarded and I rewrite it from scratch based on what I learn, and/or refactored and cleaned up, and I'm good to go from there. (with an occasional small snippet conversation in isolation, or to repeat the process for another component)

So do I think an AI tool could ever replace me (based on current state of AI?)
No, impossible.

Just a tool to help get me over initial huge humps, or out of minor jam.

And in the initial huge humps, useful in possibly showing me that it is not a direction I want to or should head in, so it gets me to a sensible conclusion much faster.
« Last Edit: October 31, 2024, 11:52:38 am by Bogen85 »

 

TinyPortal © 2005-2018