It is clear that LLMs are the new generation of PL, the biggest SWE revolution since assembly→3GLs.
The parallels & differences are fascinating.
Take interpreted vs compiled. Previously a property of the *language*, now a property of the *program*.
1. Interpreted = Store prompt, re-run each time (non-deterministic)
2. Compiled = Use the LLM to generate a 4GL program (deterministic)
Some use cases call for 1, others for 2, others for a hybrid. Choosing well will become a core SWE skill.
[1/3]
@leaverou I'm disappointed that you're ignoring all of the practical and ethical problems with LLMs. For starters, LLMs cannot be the future because they are driving increased fossil fuel use, which will destroy our biosphere and the civilization that operates LLMs. https://www.theguardian.com/environment/2026/jan/29/gas-power-ai-climate
LLMs are a self-terminating technology: either we get rid of them, or they get rid of us. Either way they cease to exist.
@skyfaller I cannot think of a single historical precedent where tech that added significant value to humans' lives was terminated because it was too resource intensive / costly.
OTOH, there is a ton of precedent where new tech starts off as incredibly resource intensive / costly and humanity eventually finds ways to make it economical.
If anything, the acceleration of the pace of innovation that AI enables may help us get there faster!
I know of at least one tech that added significant value to human's lives and was terminated because the environmental cost: CFCs.
They were cheaper, more efficient, more practical. But they were killing the ozone layer, so they got totally banned, with an exception for medical uses.
There are a ton of pesticides and chemicals to add to that list.
About tech that starts resource intensive or costly, and never got anywhere, there are many; but nobody remembers them.
Most of innovators have real trouble finding investors to make their inventions reach the profitability stage, so their inventions get thrashed.
Remember: electrical vehicles were invented more than a century ago, but burning petroleum was much cheaper and electricity was still not so available, so they were ditched until a time where we found fuel too costly in many ways to continue using it.
I gave you a bunch of examples, and you ditch them all because you didn't like the first one.
Ok, I'll bite, but I see what you are doing here.
CFCs were cheaper, of course, but they were by far a better solution for many things, too. Spray cans, for example, are much, much more bulky and weight more nowadays because we still don't have anything nearly as good as CFCs were. I perfectly remember those cans, and I remember many people complaining about how they were selling less than half the product in a bulkier can for a bigger price. Myself included, until someone explained it to me.
@jgg What I’m saying is there is a fundamental difference between "can do X better" (for some definition of better) and "can do X that was not possible before". All of your examples around chemicals are the former.
Re:tech innovation, if the invention never got anywhere it had not yet added value to human lives, so it's not a valid answer to my question. I wasn't asking about potential value, I was asking about actual, realized value.
What does AI do that was not possible before? Using humans, I mean.
Currently, it is not nearly as revolutionary as the Internet or smartphones were, and it is costing us a bunch of orders of magnitude more in human, property rights, energy and environmental costs. Not cool.
Humans have an intrinsic value of infinite. The intrinsic value of an AI is zero: if it is not useful, it is worthless. Let's not compare. There is a reason CFCs are still used for medical uses, and nothing else.
Internet got us the ability of having instantaneous worldwide communications, and access to the hugest amount of information ever, including things common folks didn't have access even travelling. Telegrams, phone and mail were not nearly as fast or comprehensive.
About AI expertise: I wouldn't call what AIs have 'expertise'; but I don't think there is a word for it yet.
By the way, I find hard to believe that someone who obviously loves the WWW as much as you do, and has done such an amazing work empowering and divulging it, has such nice things to say about the thing that is killing the web, site by site. At this rate, very soon web pages will be the new Usenet.
I feel really sad each time I think about it. No kidding.
@jgg Literally, your entire argument about the internet can be repurposed to support AI:
"AI got us the ability of having instantaneous creative resources, and access to the hugest amount of information ever, including things common folks didn't have access even travelling. Human labor was not nearly as fast or comprehensive."
When it comes to the Internet, you can see that short-circuiting manual labor, time, and affluence was a huge win. Why can't you see it here?
@jgg Re:killing the Web, I do worry about that (for content — not apps). I don't want the Web to become a walled garden where all content is behind paywalls because it’s the only way to make it sustainable. But ad blockers are what started the shift to paywalls, not AI.
Also, every sufficiently innovative new tech comes with risks, not just benefits. But I think the potential benefits here are too huge to brush them off because the risks exist. We just address the risks!
Paywalls and walled gardens, bad as they are, would be a blessing compared to what an AI first world will be.
Nowadays, many people doesn't search the web anymore; they ask ChatGPT and believes whatever it says. They aren't in forums, they don't read any blogs, or papers. Only AI.
Even worse, website owners are contracting Cloudflare services in troves in order to block agentic AI and AI training to flooding them with traffic akin to denial of service attacks.
So costs are raising, income is tanked, user numbers in decline. There is no way that's sustainable.
AI companies are creating the final walled garden, with the hugest paywall ever. Once everybody is using AI instead the web, they will do what big tech companies always do: putting ads and raising prices. A lot. OpenAI is already testing ways to put ads, in fact.
And there will no way to turn back, because people will not be using browsers anymore.
Only AI.
@jgg As an experiment, I just fed my ChatGPT (trained on my own views) your toots. This was the response: https://chatgpt.com/share/698515ea-1410-8009-bf10-edb769c29014
I could not have put it better myself.
The only thing I'd add is about AI art: Art is not just about the technique, it’s personal expression, and in that sense LLMs are helping *more* people express themselves. I may not be able to draw what I'm seeing in my head, but perhaps I can guide an LLM to draw it. Why is that not art? [1/2]
@jgg [2/2]
Now take your response about how that’s not art, and think of how it applies to photography vs realistic painting. When photography came out, the painters of the time complained about how it was not art because it was too effortless. Yet today, we absolutely consider photography art. Not the same as painting, but def an art in its own right.
And there are photos taken with no effort in 2 seconds on a phone, and photos that took a lot to create.
Exactly the same with LLM art!
Well, that's classical AI slop; it even made a bunch of straw mans.
The main point of that slop is that chatbots answers are the equivalent to web search snippets. That's simply not true for most of people. Snippets function is to help you determine which search results you want to follow, and encourage you to do so displaying the link prominently. Chatbot answers, most of the times, do the opposite; most of times, they don't even give links; and even when they do, they are tiny, and make it easier asking for details than follow the link and read the page.
About art, that's a more debatable thing. I think the most important point here is art is a form of communication between at least two persons where skill is used as a form of expression. LLM art usually demonstrate very little skills, and send mixed signals, since it is a mix of several artists work, that the AI finds get nearest to what the prompt describes.
At this point, I have to say that I'm a bit sick of hearing people say that something has been made 100% by an AI, when the effort someone made with the prompt is roughly equal to what they would have had to do if they were to do it without the AI. When I type a text in a keyboard, nobody says the text was done by the keyboard. Nobody says the Sixtine Chapel was painted by a bunch of brushes. The one who makes something is whoever dictates every detail, not the tool. 3D artists got bugged at first by people saying their art was made by computers. Wrong. It was made with computers.
So if you dictate every tiny detail of an image, then it was not made by an AI, it was made with an AI, and it may be art.
If your prompt is minimal, then, ok, your image is made by an AI, will have minimal or none artistic value, and people will be right in calling it AI slop.
On the other hand, I hate when people asks AIs to "correct" or "enhance" their text, transforming them in boring, zero personality texts. There were too many of them before AI, and AI has made that much worse. That's the opposite of art.
@jgg I read the whole thing before sending it and agreed with all of it. What are the strawmans?
Agreed on links, I've learned that asking chatbots to cite sources is a waste of time — they will just give you links that either point nowhere or the resource they point to does not prove the claim.
That does not make them useless. Part of using a tool productively is understanding its strengths and limitations. It’s still much easier to validate than to compose. [1/n]
@jgg The problem is when the chatbot conversation is the whole research.
Same with text. If you ask AI to "correct" or "enhance" your text and just copy the result wholesale, you end up with crap. But it can be useful to incorporate some of its ideas.
Completely agreed re:art.
AI is giving the same information the web was giving me, but much less reliable, and at the price of losing the original. Replacing something by a lossy copy is always a bad idea.
About creative resources, that would be long to explain. For starters, AI generated art is an oxymoron. In other domains, it may be a help for routine work, with heavy supervision.
There was a time when the Internet was many things: email, news, IRC, the web... Until the web ate everything else (they survive, but they are a shadow of their former glory).
The AI is replacing the web. A thing made by thousands of humans with love and effort is being replaced by an impersonal thing full of whatever their billionaire owner wants to fill it.
That's not a win for the Internet, it is a huge, terrible loss. For the Internet and all of us.
Yeah, now it is fun, and has its uses.
But I am thinking long term.
I love the web, I want the web. No replacements, please.
@jgg
> "Using humans, I mean"
That’s like arguing that the Internet was not revolutionary because you could always physically travel to another place and retrieve documents yourself from the computer there.
But even humans cannot substitute AI for all use cases. LLMs have domain expertise over *every* human domain, but the number of humans that have expertise across N domains tends to zero as N increases.
Lastly, humans are not 0 carbon footprint either.