Winter is coming (AI-assisted coding and its influence on Free Pascal)

[marcov]

I once heard a story about ai learning how to land an airplane. It decided that the best way to land it was to crash it ....
It certainly is the best choice in terms of speed and fuel efficiency right?

But it didn't knowingly crash it, it just optimized the hell out of some math problem. But this is not new, and AI is not new,  even in the nineties when Fuzzy Logic was all the rage (and proponents rightfully argued that FL improved regulation of certain processes in many cases), the solution was to simply have a failsafe system to take over (but less optimally) when the FL process had extreme outputs.  Best of both worlds, high efficiency over all, but still safe, so essentially a solved problem.

I can't image anyone enabling AI in anything critical without such failsafes, but such outcome in some lab setting and, worse, the big ALL CAPS reporting about it, is of course endless fodder for doom stories.

[VisualLab]

Interesting Links:
https://www.courthousenews.com/microsoft-and-github-ask-court-to-scrap-lawsuit-over-ai-powered-copilot/
https://techcrunch.com/2023/01/27/the-current-legal-cases-against-generative-ai-are-just-the-beginning/?guccounter=1
https://www.theverge.com/2022/11/8/23446821/microsoft-openai-github-copilot-class-action-lawsuit-ai-copyright-violation-training-data

Because this is a problem not with AI itself, but with people on the boards of corporations. People at the top of corporations are very greedy for money and power (it is obvious that they want to make money at all costs, i.e. preferably from other people's work for which they are not paid). These first lawsuits are just the beginning and they are right. There would be no such lawsuits if the created AI tools, along with the answers, provided information about the sources of their knowledge. Unfortunately, this area of human life (i.e. the use of AI) has not yet been legally regulated (because it is still too new).

And probably the "buried dog" is here. Because if such a "smart AI chat" provides information about how it knows what it returns, then the charm of "smart artificial intelligence" is broken. In other words, an ordinary person, fascinated by AI, is amazed at how "smart" this AI is (and how smart these corporations are to have invented such smart AI). However, if such a "smart AI" also provides sources in addition to the relevant information, the average person is no longer dazzled by AI's wisdom. Instead, such a person has the impression that he is simply using a slightly better and more extensive "search engine" (which is true). But then there will no longer be the "wow" effect that people from corporate management probably expect.

However, current AI is not a breakthrough. As AI skeptics (of which I am one) rightly noticed here, it is only and exclusively extensive software that searches and transforms existing data (created by humans) according to a set of given parameters. This software does not create anything on its own - it creates absolutely nothing, contrary to what some people may think. And there is no chance that anything will change in the near future. This current third wave, like the previous two, will reach the technology wall at some point. And for some time, people will again abandon interest in AI (as before). And when, after some time, technological barriers are overcome, some people will slowly start to come back to the topic of AI.

Also, viewing AI as a threat is a bit exaggerated (e.g. SkyNet). A much bigger threat is the decision made by humans to entrust certain areas of human life to AI without human control (supervision). Can you blame a device for causing harm to someone? In such situations, the designer or manufacturer of the device is rather blamed (because he did not take care of something at the design or production stage).

In my opinion, the reason AI is currently popular among people is:

1. people's lack of general knowledge about it as AI is designed and how it works,
2. expecting that some mysterious but magical force will do something complicated for them (which is beyond their capabilities) in an automagical way.

This is a classic case of people's lack of knowledge (and often reluctance to acquire it), combined with naivety and the desire to quickly obtain the desired results. Humans, as a species, have always been like this (to be precise: a large part of the population). And the development of humanity progressed mainly thanks to those few individuals who "wanted to learn something and do something." A huge number of people have always wanted to "take shortcuts", preferring to believe in miraculous solutions because they do not require effort (mental, physical) and a lot of time. And this is what people were promised by shamans, priests of various religions, rabbis, imams, commanders (duce, fuehrer), party leaders (e.g. communists in the USSR, PRC, etc.), charlatans, and today it is done by marketing specialists employed in corporations.

Returning to the main topic, i.e. FPC and AI. Yes, AI can be useful for performing certain tedious but repetitive activities. Unfortunately, it will not design new, fancy and complicated classes, procedures, algorithms, GUIs, etc. for the programmer. It will not do this because it is not creative - it is only reproductive, i.e., according to given patterns, it creates variations on existing topics, those it finds in its database . AI is just incredibly overrated (a bit like Python or Bitcoin). It's probably quite well suited to formulaic and repetitive work, such as website development. It can also be useful as an aid when searching for documentation or suggesting solutions to AI problems (if it finds them in its database). And that's it for now. Perhaps in some time it will be possible to have it generate code based on some schema and a set of required parameters, provided that the results are not full of errors.

[Handoko]

Also, viewing AI as a threat is a bit exaggerated (e.g. SkyNet). A much bigger threat is the decision made by humans to entrust certain areas of human life to AI without human control (supervision).

I agree AI is not the threat. The sad truth is if human has something that is very powerful, there always bad guys using it to do bad things.

https://www.youtube.com/watch?v=LHC862LOljk

[Curt Carpenter]

This software does not create anything on its own - it creates absolutely nothing, contrary to what some people may think.

A lot depends, I think, on what you mean by "create" and "on its own."   Everything else aside, "AI" (and AI hype) is forcing us to think about those questions again in strictly human terms.

[MarkMLl]

I once heard a story about ai learning how to land an airplane. It decided that the best way to land it was to crash it ....
It certainly is the best choice in terms of speed and fuel efficiency right?

Steady there. That might not be about the "current generation" of AI, but 1980s-era AI as promoted by Lenat which was based on an enormous ontology and in principle should have been able to "reason".

Lenat and Eurisko gained notoriety by submitting the winning fleet (a large number of stationary, lightly-armored ships with many small weapons) to the United States Traveller TCS national championship in 1981, forcing extensive changes to the game's rules. However, Eurisko won again in 1982 when the program discovered that the rules permitted the program to destroy its own ships, permitting it to continue to use much the same strategy. Tournament officials announced that if Eurisko won another championship the competition would be abolished; Lenat retired Eurisko from the game.

https://en.wikipedia.org/wiki/Eurisko

MarkMLl

[Joanna]

Steady there. That might not be about the "current generation" of AI, but 1980s-era AI

Ok you may be right, maybe we should approach this from a logical rational perspective. All we need to do is know who is promoting ai and what sorts of things That they have done in the past. Isn’t Microsoft a big promoter of ai? Enough said. 

[VisualLab]

Steady there. That might not be about the "current generation" of AI, but 1980s-era AI

Ok you may be right, maybe we should approach this from a logical rational perspective. All we need to do is know who is promoting ai and what sorts of things That they have done in the past. Isn’t Microsoft a big promoter of ai? Enough said. 

Not so much Microsoft (because lower-level employees usually have as much say in a corporation as a pig in a slaughterhouse), but its management board (perhaps 2 or 3 people from the very top). It's just that some guy at the top is fixated on this topic and thinks that in some time it will bring the company a lot of money. Meanwhile, it does not depend solely on how much money such an individual from the management board will spend (not his own) money, or how convinced he is that he is right, but on the actual feasibility of a given project and its real usefulness. For now, all we know is that the management has poured huge amounts of money into AI. Whether this will translate into great profits for this corporation (and for him) will become clear in a few or a dozen years. Decision-makers in the highest positions have often spent a lot of money on projects that are of little use or have poor prospects because they thought they were right (they consider themselves infallible).

[Joanna]

There are plenty of examples of bad ideas with well financed backers that have been forced upon the entire population with disastrous results . I wish that being a bad idea was a guarantee of failure. Unfortunately it is not.

[Thaddy]

Moreover, your observation about human tendencies to seek shortcuts and miraculous solutions resonates deeply. Throughout history, humanity has been drawn to promises of effortless solutions, often at the expense of genuine understanding and effort. It's crucial to approach AI with a critical mindset and a willingness to understand its capabilities and limitations.

Indeed, it is not AI but human at fault. The whole forum is full of questions that are answered in the documentation, but some humans take a shortcut and simply  ask here... Which is a pain...

[Curt Carpenter]

Indeed, it is not AI but human at fault. The whole forum is full of questions that are answered in the documentation, but some humans take a shortcut and simply  ask here... Which is a pain...

It seems likely that your pain may well be eliminated in the not too distant future, as the (human+AI assistant) "centaur" becomes a reality and the standard context for creating programs.  No need for the programmer to ask a "forum" of any kind: (s)he will put questions to the AI assistant who will engage every question patiently, without judgmental sidebars and without the occasional bad days. 

Programmers will have a "Thaddy in a box" with a few important improvements.   

( https://www.newyorker.com/magazine/2023/11/20/a-coder-considers-the-waning-days-of-the-craft ). 

[Thaddy]

Less temper, yes 
Funny thing is that AI already indexes your and mine examples from this forum...
I have no objection to that, apart from the fact it also indexes the wrong examples that I came up with..
After  a week or so you can request your own answer.. (copilot, chatGPT based)

[marcov]

It seems likely that your pain may well be eliminated in the not too distant future, as the (human+AI assistant) "centaur" becomes a reality and the standard context for creating programs.  No need for the programmer to ask a

If you want to talk about such stuff, please comment using own experiences and real worth that you experience, rather than echoing so called visionaries that just extrapolate a graph.

[Thaddy]

ATM it brings you idea's on how to procede, not so much a finished solution in every case.
In my opinion it is really worthwhile.

[lainz]

I think AI is better than in the previous year. Now it gives compilable source code. I use Claude ai at work and is just faster than typing the code myself.

For example I created a web service classes and client, with soap parsing for Android with kotlin language without typing anything than the prompt and pasting the wsdl file.

[Curt Carpenter]

If you want to talk about such stuff, please comment using own experiences and real worth that you experience, rather than echoing so called visionaries that just extrapolate a graph.

What an interesting reaction.  I wonder where it came from?

I don't know Mr. Somers personally, but I found the essay on his experience of adopting AI in his professional practice both interesting and informative.  And I believe it was Oliver Heavyiside who said "Some learn by experience:  I prefer to learn from the experience of others."  Some wisdom there, and I don't question that Somers has had far more experience of the profession than I have, and at 80 I don't expect to accumulate it.

Whatever.  I would be pleased to read about your own views of the future of "centaurs" in  software as a profession.  Cheers.