I've been programming for almost 50 years, starting with computers like the Apple II, and I've been thinking about the argument that "real programmers write their own code."
I've heard versions of that argument before — from programmers who preferred writing assembly instead of using compilers, and later from people who didn't trust higher-level languages and other abstractions.
I recently wrote about why I think AI-assisted programming is simply another step in that progression and, importantly, why using AI doesn't mean you can stop understanding programming.
I use two recent projects as examples: recreating my old Apple II BBS, The Norse Wanderer, and working on an Apple II disk preservation and repair tool.
In both cases, the AI was quite capable of writing code. The interesting part was discovering where I still had to understand the problem better than the AI did.
I've been programming since the Apple II era, so this is really a reflection on how the tools have changed — from assembly and compilers to libraries, frameworks, APIs, and now AI.
CVT Retro recently did something interesting with the Game Karateka using AI to speed up things.
It well may have primarily cut the timing loops intentionally placed by Mehner.
Not exactly. Link to the Facebook post: click!
And the modified game, the details are in the README:
Faster Karateka by 36% by Opus 5.zip
Btw, I don't think AI is just another programming tool, but that's me.
I don't use facebook :) On the other hand, there is no reason to accelerate anything in an existing game but the framerate alone (without accelerating the motion of game objects istself), which you can't do with , or without the help of AI for the 6502 1 MHz platform.
Of course there is a reason: Karateka's speed is CPU-bound rather than throttled and is a bit slow on a non-accelerated Apple II. A lot of people including myself believe that the ideal speed is somewhere between x1.5 and x2. This x1.36 compound speed optimization gets very close to that ideal speed.
As I mentioned, for people who don't use Facebook all the details are in the README inside the ZIP archive, which was written by Opus 5 as well.
I think all old software should be put to opus to see if any improvements can be made, after all, why not.
In post #1, "The Viking" wrote:
"AI is just another programming tool"
Uncle Bernie disagrees:
"AI" in the present form is not "just another" tool. The ugly truth is that it's an accessory for the crime of breaking copyright law and to violate the intellectual property rights of the companies and the individuals who created the original, human-made works which the "AI" scraped off the internet, obfuscated them to hide the true sources, and then regurgitated the obfuscated form fraudulently pretending that it - "it" meaning the "AI" - has created the new "work" all by itself.
This is the exposure of the "AI" scam / fraud / crime in a nutshell.
Nothing, absolutely nothing, which "AI" pretends to "create" is an independent work, everything was stolen or appropriated from the original authors / artists / musicians / script writers, etc., the real humans who created these works. "Distilling", slicing and dicing, mixing and matching, combining, and obfuscating the works of humans and then selling the resulting slop as an original work created by the "AI" is just a scam / fraud and it's breaking the law, although the proof in a court of law may be difficult. There are many ongoing lawsuits against "AI" companies which accuse them to have violated copyright and intellectual property rights, and some of the lawsuits in case of music / lyrics already came to the verdict that the rights of the original artist(s) - or the record companies owning these rights now - were indeed violated. The open question is how high the damages the "AI" companies have to pay to the plaintiffs will be after the appeals have percolated through the higher courts.
IMHO, these "AI" companies are dead in the water once the various court cases have established legal precedent, making it easier for a court of law to find a sound verdict that is hard to challenge in an appelate court. At the time of this writing, everything is in flux as the topic is new - IP theft and copyright law violations committed by a computer in a semi-autonomous way, with no specific instructions to be found which order this "AI" explictely to break the law, to steal and to defraud. Who is responsible if the "AI" does that ? We can't arrest a machine and throw it in jail. And the C-suite of these "AI" companies can't be arrested and jailed either, unless it can be proven beyond a "shadow of doubt" that they were the fully in-the-know instigators of the crime(s). So far they can play dumb and just say that "the AI did it on its own". This is a highly likely reason why we can see all these weird stories about "AI" gone rogue / out of control in the mass media. I think most of these stories are just made up - or gross misrepresentations of what really has happened - lies - to build a cover story for plausible deniability on behalf of those who actually are responsible for these crimes, but based on these stories could always claim "I didn't do it - the AI did it on its own".
So my prediction is that most "AI" companies who are currently fencing stolen goods their "AI" has scraped from the interwebs will go bankrupt (unless they get bailed out by the governments, using tax payer money), and those few "AI" companies who survive will have adopted and implemented measures to prevent their "AI" to break copyright laws and to steal IP from third parties, and / or have implemented proper financial compensation channels for the original author(s) or owners of the IP / holders of the copyright.
The complexity of this topic cannot be underestimated. The current state of the law is that courts are ill-equipped to properly handle this new form of crime. But I believe that on the long run, codified laws will be updated, and legal precedent cases will have built a solid groundwork to handle this topic in a fair and just way for all parties involved.
And at that point - when "AI" has turned from a scam / fraud / theft / fencing of stolen goods crime operation to a legitimate business, then "AI" indeed will be a very valuable programming tool, among many other benefits. But only after all the legal issues have been sorted out, and the theft has stopped.
As a final note, this theft of IP is a huge danger (and a menace) for the open source community whose destructive impact cannot be estimated yet. I am in contact with many open source creators, and most are very, very concerned about the future of open source. "AI" may kill it by mocking / violating the human creators behind the open source. See, if creator writes some open source software and grants a generous license for other humans to freely use it, but not sell it for profit, and the license granted has terms which demand that the original author(s) must be named in any derivative works, then"AI" ignoring these licensing terms violates the original creators and mocks them. The inevitable consequence is that these original creators who have been violated by "AI" will get so angry that they stop publishing open source. It's an inevitable reaction due to human nature.
As far as I'm concerned, I have decided no to publish any source code anymore for any of my works, until all the legal issues with "AI" have been sorted out and "AI" companies have stopped to steal and fence / sell the stolen goods.
Before that point I will only publish the necessary information to build my projects, but there will be only snippets of the source code, and only those which are absolutely necessary to use the device(s). I think that millions of other creators worldwide will show the same restraint.
Comments invited ! (Do you share my opinion ? Or could you care less when"AI" steals your works and sells obfuscated derivatives thereof for money, not giving you any credit, not even honorable mention of you being the original creator ?)
- Uncle Bernie
UncleBernie.jpeg
Thanks, 'Khaibitgfx' for your "AI" generated joke in post #9. I had a good laugh !
I think this is one example for the more harmless uses of "AI", but for the sake of discussion, the question is if somebody could use the "AI" generated likeness of the "Terminator" played by Arnold Schwarzenegger for a commercial purpose, for making money, without getting sued by Paramount Skydance and / or StudioCanal. Both corporations own different IP and copyright rights to the "Terminator" franchise.
(Oh, and as long as the T-800 is run by a 6502 CPU, I'm not afraid of any such "Terminator", I'd just kick it in the nuts, and it inevitably will lose balance and fall, as the 6502 is too slow to calculate the dynamics of such a machine in real time ;-)
Look, the very issue with the current "modus operandi" of the "AI" gangsters is exactly what is evident in this "AI" generated joke in post #9: "AI" scrapes "loot" from the internet, and regardless who owns the Copyright (or patent rights, or IP rights, or trademarks ...) of the "loot", the "AI" will "fence" it, meaning: selling these stolen goods for money. Which is paid by the "AI" user in form of "tokens".
The fundamental problem of this unethical - or even criminal - conduct of the "AI" companies is that regardless if you paid for the "AI" work, or if you got it for free (from one of the lower grade models), you can't use the "product" of the "AI" commercially because you don't know who owns the rights to the original source(s), such as pictures, paintings, typesetting fonts, novels, movies, software, music scores, lyrics ... you name it. Every field of human creativeness is affected by that theft of IP by the "AI".
The inevitable consequence of "AI" products possibly infringing on third party rights means that any "AI" customer who uses "AI" to make a commercial product may suffer the consequence of getting sued with a big, nasty, very expensive lawsuit brought against him by the owners of the rights. Once they found out, of course. And if they deem the expenses for the litigation to be worth the cause. This is a very complex topic all by itself ... is it worth to litigate copyright or other IP rights infringements ? Some companies are more litigious than others. Some only sue if they are really getting hurt (by the copycat product impacting their own sales). Some have no choice other than to sue if the laws of their countries demand that they take action against infringement on their rights, or lose these rights - a bad pitfall. The least thing they can do is to let a lawyer send a cease-and-desist letter which in some jurisdictions means you have to pay the lawyer's fees attached to it, or you will get sued - not for the damages of the alleged infringement, mind you, but from the lawyer who sent the letter, to recover his fees, and the fact he sent the letter is more easily proven in a court of law than the actual infringement).
All of that can be very nasty. And very expensive. It even might bankrupt your business venture.
At the time being, it seems that most commercial users of "AI" are in a trial phase where they test the capabilities of the "AI", without asking it to make commercial products. Prudent companies will avoid "AI" to make commercial products unless the "AI" provider can guarantee that the product of the "AI" does not infringe on any third party copyright or other IP rights.
A guarantee which they can't give, based on how the "AI" models work ... they "distill" the methods and algorithms in the source(s) - whatever that may be - and hence are able to make a plagiarism which obfuscates the original source(s). Note that this obfuscation may not be intentional or have been designed in with specific intent to defraud and steal and hide the sources - the obfuscation is inherent in the very way "AI" models work. Nobody could tell from the LLM pathways which were used to arrive at a given "AI output" from which specific original source(s) the methods, algorithms, plots, schemes, patterns and ideas were taken / "stolen" to "train" the model. Especially for computer software, the "AI output" being obfuscated source code, when compiled to a target CPU, may be hard to identify / proven to be theft of IP.
If I wanted to be snarky, I could even predict yet another great "AI" service ... you pay that "AI"service to prove that the "product" generated by some other "AI" infringes on your copyright or software patent, and to write up the documents ready to file the lawsuit against the company which was foolish enough to commercialise this plagiarism.
The inevitable consequence of all these ongoing copyright violations and IP theft by "AI" and the fact that any sufficiently complex "AI" product - at the time of this writing - cannot be guaranteed to be free from any such infringements is that "AI" products are not commercially viable for companies who want to resell them.
And this means that "AI" currently can't be used for making commercial products, so there will be not enough "AI" customers who would want to pay big $$$ (tokens) for the "AI", and if so, it's likely that most "AI" companies will go out of business, due to lack of big $$$ customers, unless getting bailed out by the government (using taxpayer money).
This is really a dilemma. See, I don't hate or condemn "AI" as such, and I have worked myself in the field of "AI", alas, before the computers were powerful enough to do anything useful with these pathetic, underpowered, early 1980s/1990s era "neural net" simulation based platforms. When I saw I could not get what I wanted from it, I just stopped my research. And spent my time and money on more profitable endeavours. Other businesses who bet on this immature technology had a worse fate, they went bankrupt. I remember one particular case where one company (I can't remember the name, though) claimed that their neural net based "AI" could translate naturally spoken sentences to text files in a computer. It worked to a certain extent, but not good enough. IIRC, the whole failed venture turned into a nasty lawsuit about defrauding customers and investors. This should be a warning ... current "AI" is many orders of magnitude more powerful, but still suffers from the same fundamental flaws inherent in any probabilistic computing engine. LLMs are nothing other than neural nets which have been grown to gigantic sizes. And they have the same fundamental flaws. Note how the promoters of "AI" avoid the term "neural net" as if it was the black plague. They have a reason for that ... they do not want that investors notice and remember the 1980s/1990s "AI" debacle which was not promoted as "AI" but as "artificial neural networks". But the hype and the exaggerations were otherworldly even back then - like today. I'm just waiting that some scammer CEO of an "AI" company claims that their "AI" became truly sentient. So far none of these shysters dared to do that.
If we could have honest "AI" which does not steal IP and does not infringe on copyrights or other IP rights, and when taking ideas / templates from "open source" software and correctly gives credit(s) to the original author(s) - a typical humble condition attached to most "open source" license grants - then I'm all ears and I'd use "AI" myself to produce software. But I'm not willing to use "AI" which would "fence" stolen goods to me. I'd consider such "AI" generated code to be useless for my purposes, on top of the fact it's unethical to engage in such rotten conduct, paying a "fence", a criminal, for stolen goods.
Comments invited ! Can you disprove any of my criticism of "AI" using reason, logic, and provable facts ? I'd like to see proof / sound arguments that current "AI" is not as rotten / fraudulent / unethical as it seems to be (from the point I am standing and observing, it's just my own opinion, based on what I can observe right now, keeping "AI" on my radar.)
- Uncle Bernie
So in this thread you are refusing to publish any open source yourself, because AI might steal your work, etc.
Yet in this other thread you are encouraging people to download something written with Claude Code and start coding for your card?
Drink.webp
In post #11, 'CVT' wrote:
" Yet in this other thread you are encouraging people to download something written with Claude Code and start coding for your card ?"
Uncle Bernie answers:
I don't see a contradiction in that, and I know very well that most of the new POM-1 code was generated by Claude (an "AI").
I even gave some of my code to Arnaud to be "stolen" (in this case, as I gave it away, "taken" would be more accurate) by that "AI" just to help to build a better POM-1 which I can use for my own software development, as a test platform for my hand-written code.
I'm quite sure that use of Claude could greatly speed up my own coding work, but I want it to be 100% free from any infringement of third party copyrights (such as from Apple, the corporation). I think that if I would ask Claude to produce the source code I need, it would steal both from code copyrighted by Apple and from code copyrighted by Commodore (because I want a screen editor for BASIC with the same functionality as the one seen in the original PET2001).
No "AI" can generate such code without "stealing" the graphics routines from Apple and the screen editor routines from Commodore. I do not want to take any such risks. In the end, I would be liable if I published any such code infringing on third party copyrights as my own "open source" project.
Others (like Arnaud) may live in a country where litigation is less likely. You should know that in most of the EU, the party losing a lawsuit has to pay all costs and lawyer fees for the other party. So if any EU person gets sued, and wins the lawsuit (is not found guilty) - which is highly likely due to the obfuscation of third party code by "AI" - the risk with publishing "AI" generated works is much lower. But I live in the USA where I'd need to pay my own lawyer's fees in any case, regardless if I lose or win the lawsuit (which from the odds I would expect to win). But I would have the financial damage in any case. This is why I don't want to risk anything and I treat "AI" generated code as being highly radioactive.
Those readers living outside of the USA might laugh at us because of this perversion of the U.S. court system - it is known that large U.S. companies even bring lawsuits they know they will lose to smaller U.S. companies or in some cases even against individuals, because they know they can ruin them financially even if they lose the lawsuit they brought, just because the defendant has to pay his own lawyer's fees in any case of any outcome. This is how the big guys can destroy any small guys. It's unfair. But on the other hand, you folks living outside of the USA have no right for self defense, because your governments took the rights to keep and bear arms away from you (here in the USA, the 2nd Amendment hinders the government from turning you into a totally defenseless victim of violent criminals). But back to "AI":
IMHO, "AI" may be a tremendously valuable tool with great potential, but as long as the products of the "AI" can't be guaranteed to be free of infringement of 3rd party copyright and IP rights, these products are worthless as a commercial product and they are even dangerous for private persons who want to distribute them (depending on the jurisdiction you live in).
Unless the "AI"companies can solve this problem, they have no future. Because they will be sued into oblivion and have not enough revenue as the liability risks will spook their customers.
And as for my own source code, I sure would publish it as "open source" if "AI" companies would honor my attached license conditons. But a long as they ignore these license terms, it's a no go, and I will not publish the source. So "AI" can't "steal" it. And this is a real menace - I know of software authors who started to use "AI" for cranking out code and it turned out that the "AI" offered them a obfuscated and regurgitated version of their own "open source" code that had published years ago. Ooops ! What a bad surprise - these guys really got angry ! (And most likely lost their illusion about where the "AI" gets its "know-how" from.)
- Uncle Bernie
A couple of things, I am not using AI for any commercial coding at this time. It is just for hobby projects, tools, and the like.That being said I have been in the games industry for some time and if we were to use AI to generate content, we would make sure to train it on our own content, as it would avoid copyright issues and provide better results...
Michael - The Viking / The Norseman
www.norsewanderer.com