It might be a mistake to assume tomorrow’s training looks like today’s. Unsupervised learning is a thing and a very hot research topic, precisely because it avoids some of today’s big problems with acquiring the vast amounts of training data necessary.

Unsupervised leanring has been around for years and is already how the current wave of models are trained. It doesn't mean no data, it means no human provided labels of the data. So you still need creative new human ideas to move LLMs forward. LLMs != intelligence.

Exactly this. Even frontier models like Opus 4.6 have absolutely zero understanding of the task at hand. If you give them problem they have not encountered in training data, they will not solve it. You can however, guide them to resolve the problem, but in that case these get reduced to merely an auto complete. Don't get me wrong - models are getting better and hide very well that they don't understand anything, you can almost get fooled now.

Maybe there aren’t that many new/necessary ideas that can be mined from the fundamental building blocks of software development (languages, syntax, runtimes, static analyses, type checking, etc). Maybe people will continue to innovate by instructing models to build novel things out of those building blocks? Perhaps things we would not have thought of building before due to the effort required without LLM assistance.

I don't expect them to realise that until some time after it actually happens. When it remains a future hypothetical, it won't be accounted for.

fwiw 90% of software is reinventing the wheel. 80% of devs have an itch to "rewrite from scratch".

AI will deduplicate all of this

My experience is that 100% of AI devs are reinventing the wheel, most of the time for no better reason than "I can do it" or "not invented here"

I opened LinkedIn today (out of habit) and the first post was someone explaining how much Slack costs and that with AI every company can build their own Slack for $100. So that person decided to build an open source Slack clone using Claude Code. Granted, these were a few sane comments showing good alternatives that have already been built.

But for me it's been a signal that people have no imagination, so they are just burning tokens for no reason.

Is it not the business model of these companies ? Bill per token and stuff

NIH = more token = good for business

This is fine. How else do you learn but by taking things apart and rebuilding them? This obsession with productivity is incompatible with onboarding new talent. Having 1000 versions of the same concept is exactly what progress is.

> 1000 versions of the same concept

That sounds beyond wasteful.

It is. Humans are messy.

Why would there be a lack of original ideas? People who are born to code so to speak will do it. Information wants to be free as the saying goes. It only takes one time for an innovation for it to be to copied everywhere.

We don’t need the same volume of developers to have the same or faster speed of innovation.

And conversely if there is stagnation there is a capital opportunity to out compete it and so there will be a human desire to do the work.

Tl;Dr. People like doing stuff and achieving. They will continue to do stuff.

ps it’s too much to claim other people don’t experience creative ideas using AI. You don’t really know that’s true. It hasn’t been my experience as I have had the capability and capacity to complete ideas on my back burner for decades and move onto the next thing.

That’s the big scary point at the crux of all of this - you’ve had decades without the tooling to develop instincts. Nobody knows whether it’s possible to develop instincts with the tooling or what those instincts will look like. Creativity takes a degree of skill to execute on and the concern is that we’re potentially graduating people to painting the ceiling of the Sistine chapel before they’ve even learned to sketch.

At minimum, our current generation of leaders will have to get much better at managing resources and building people up. We have to up our games and build environments where the pursuit of deep understanding is permissible. Unfortunately with the current hiring issues, it’s totally understandable that young developers are scared to take time on tickets.

I can't repair my car, which used to be the hallmark of technical masculine skill in the era of Grease the musical, because mechanical maintenance is not primary in my lifetime. Nor do I have any idea how to manage a farm. I think the kids will be fine. On the other hand, I fear I will not survive once my Internet connection goes out.

That analogy brought another one to my mind: groceries.

We used to hunt and grow most of what we needed. For various reasons (good and bad) we are now completely removed from the whole process. That allowed us to prosper and evolve by focusing on other things. We took all of what we knew about growing/hunting, grouped it together with industrialization and abstracted it as "everything that happens before". It's a big shortcut - which is a big win for "other things" we may want to do.

AI is a lot like that. But the danger is that AI (or its dependencies - isn't Claude down this morning?) can disappear tomorrow, and then where does that leave us? It's as if all the animals and vegetation were controlled by OpenFoodstock.

I can't be the only one thinking we cannot let that happen.

Innovation is irrelevant to pushing up this quarter's numbers. No one actually values unique and novel ideas. The only thing that matters is shipping something right now that can make an impact on this quarter's numbers.

Who cares if it's derivative slop or a straight up bootleg of something else so long as the number goes up

It isn’t about training anymore. It is about harnesses.

Just look at new math proofs that will come out, as one example. Exploration vs Exploitation is a thing in AI but you seem to think that human creativity can’t be surpassed by harnesses and prompts like “generate 100 types of possible…”

You’re wrong. What you call creativity is often a manual application of a simple self-prompt that people do.

One can have a loop where AI generates new ideas, rejects some and ranks the rest, then prioritizes. Then spawns workloads and sandboxes to try out and test the most highly ranked ideas. Finally it accretes knowledge into a relational database.

Germans also underestimated USA in WW2, saying their soldiers were superior, and USA just had technology — but USA out produced tanks and machinery and won the war through sheer automation, even if its soldiers were just regular joes and not elite troops.

Back then it was mechanized divisions. Now it is mechanized intelligence.

While Stalin said: Quantity has a quality all its own.

There is no "new ideas" with AI. Claiming the opposite is a fundamental misunderstanding of the technology.

What I am excited about is the possibility of LLMs to draw conclusions from the last 150years of scientific papers.

There have been lots of instances of knowledge being rediscovered even when it was previously published but sitting on some shelf forgotten. LLMs ability to digest large volumes of data will I think help with this issue.

We will still need to reproduce and verify conclusions but will be interesting to see what might come from this.

While that’s kind of true in some sense, I think there’s an argument to be made for the contrary: that the mechanism for generating new ideas in humans is not quite as special as we would like to think.

In other words, creativity in humans is arguably just as derivative as in machines.

I think this can be falsified by just considering the history of humanity. It wasn't that long ago that human language literally did not even exist. And our collective knowledge wasn't all that much more than 'poke him with the pointy end'. Somehow we went from that to putting a man on the Moon, unlocking the secrets of the atom, and more. And if you consider how awful we are at retaining/sharing information and just general inefficiencies due to the fact that we're humans and not just logical information processing machines, we did all of this in little more than the blink of an eye. This is something that seems to certainly be rather special.

All that humanity has achieved happened due to the simple loop of identifying a desire/need and finding a way to satisfy it. Also known as reinforcement learning. The only thing that really differentiates humans from machines is... history. We've been learning and passing on our knowledge to successive generations over millennia. Nothing really special there; give the machines a few years to learn and see what happens.

What needs do machines have? What desires do they have?

None, yet. But you can be 100% sure it's something we'll eventually succeed in adding, as it's through the guidance of desires and needs that intelligence really expresses.

Not sure how that follows but okay

What a ridiculous take lmao.

What are your contributions again?

And what exactly is ridiculous about it?

Reinforcement learning requires a well defined goal and a well defined way to quantitatively measure progress along that goal. In reality these don't exist without a hand of God guiding you. In the case of machine learning that hand of God is our own. Even given infinite processing power, you could not construct a reinforcement learning system that would mimic humanity's progress - it simply is a nonstarter due to the nature of reinforcement learning itself.

Conceptually, it's really not as hard as you make it seem. There are layers, but once you peel them away there's only one thing left, which all living things share: the drive to survive (maintain internal state parameters within a certain range by accessing nutrition, protection from environmental elements, security from other survival-seeking entities, reproduction to pass on genes, etc). No need to bring God/gods into it.

There's also no need to specifically mimic humanity's progress; that's just an accident of survival facilitated by opposable thumbs and language ability. We've already made machines with the base abilities, and emulated the drive (see evolutionary algorithms[0] for example). We just need to put it all together in a few units and let them "loose" to evolve on their own for a while. It took humans ~300,000 years to get where we are today; I'm positive that it'll take machines a small fraction of that. Nothing special.

[0] https://en.wikipedia.org/wiki/Evolutionary_algorithm

Very little, certainly approximately 0%, of what humanity has done has been driven by base survival instincts. You're describing the process to mimic a roach, not a human.

You do work, don't you (or at least are in school)? Do you do it for sheer fun? Or to be able to afford things that allow you to survive?

Even those who do something like art "for fun" do it because it sates an internal need (not all actions need to make sense, because the inherent randomness of evolution is messy and leaves artifacts). Though also the desire to create some form of legacy can be considered a kind of survival: to be remembered by others beyond one's own lifespan.

I’m not claiming an LLM is structurally or functionally equivalent to a human brain. I just said that what we call “creativity” is in fact a very derivative thing.

i don't think all sides of this discussion agree on what a "new idea" is. i am a very creative person but i've never had a truly original thought and i don't know how having one would be possible

It depends on what layer you look at I think, shoulders of giants and all that..

that's only partially true.

AI can innovate in synthetic-realm of novel ideas, while real-world novelty will remain untouched.

There are different types of novelties

What is a "real-world novelty" and what prevents AI from touching it?

If AI could innovate it wouldn't be a public product. It would be a cash cow. Why give your customers the ability to come up with new and amazing ideas when you can just keep it for yourself and launch a thousand products? USA is a capitalist society. It doesn't share profitable ideas.

And if AI was really about productivity they'd be talking about doing more faster with the same workforce, not reducing the workforce.

if you like, the business model is called Innovation-as-a-Service :)

That's perfectly aligned with capitalistic motivations

I hear this sentiment a lot but it doesn’t ring true for me.

What is an idea really and what’s your definition of new?

If i get a LLM to spit out, I dunno, a deployment system written in haskell that uses bittorrent or something, none of those bits are new, but certainly there will be unique challenges to solve in the code and it’s a new system.

Where is the line for new? Is it in combining old ideas? If not then does any software have “new” ideas? It’s all combinations of processor instructions after all…

Technically all the problems that almost any given business needs to be solved today has already been solved umpteen times over the years. There are no new problems that can't be solved by porting and/or combining old solutions.

That's the literal definition of stagnation. That is not compatible with growth.

Also that's not a new idea, that "everything worth inventing/exploring has already been". It's precisely what AI reinforces, and that goes against human nature (and capitalism) as that statement has historically proved.

Oh no, there will always be things that're still worth inventing/exploring. They're just very few, and a tiny subset of that few will be applicable to today's businesses.

"Everything has already been invented." - Some 19th-century scientist who had no imagination to see the wave of technological innovation that was coming.

Iterations and remixes. That's what innovation is. That there's sometimes a huge jump doesn't change the nature of it.

It's the opposite. The less competent the average developer the more valuable coding LLMs become (as the only way for those bad developers to generate ok code). Eliminate the developers and even bad coding LLMs become valuable.

Hey man is there a private space where thinkers like you hang? Would like to join :)

I wrote a similar post in this thread.

I don't know... but I'd rather mix in with the cultists. A dissenting opinion is important in their echo chamber if we're ever going to give sanity a chance.