Whatever gets predicted by tokens gets summarized by symbols, which are artifacts of language. This gets to the illusory aspects of binary as well, the rabbit hole goes deep.
Intelligence is wordless. Basic fact. And words are used in the aftermath of thought/intelligence, not prior. They are aftereffects, not effects. Words have no prior. That's a big problem.
The contraction of events and the semantic residue, into tokens, symbols, metaphors, sentences, images, segments, parameterized in any way shape/form, is irrelevant next to the semantic load the event contains in a variety of analog states.
No creativity emerges from this, just the mimicry of the tokens, symbols in their reduced, modeled parameters.
That's the problem with LLMs and frontiers, RL. It's all subject to the bottlenecks of math and symbols.
The analog load of semantic (use tasks, actions, variance, scales), the immensity of the nested forms these take between brain, body, screen, canvas, aren't replicable as creation.
It's not information or stimuli we engage with, these are convenient false reductions we're realizing are inconvenient. There's much more the brain, eye detect in reflection. The words and symbols are an aftereffect the computer detects on its way to disregarding the events.
We see this aftereffect as creativity, thinking, etc only as a clever illusion the words, tokens hand us. Nothing more.
I disagree with your view on creativity, and indeed find LLMs to be remarkably creative. Or perhaps, to sidestep the anthropomorphization issue, I find that LLMs can be used as tools to produce creative works, greatly amplifying one's creativity to the point that a "near 0 creativity" subject (whatever that is) can create works that others will view as profoundly creative.
In truth, I don't think there's likely to be a correct definition for what "creativity" is. We're all just moving each other's goal posts and looking at different subject aspects of our experience.
What I do know is that I have talked to dozens of people, and have friends who have talked to hundreds. When shown what LLMs and generative models can do, there's a significant portion of them who label this work as creative. Rather than deem these people ignorant, I would rather consider that maybe if enough people see creativity somewhere, perhaps there is something to it. LLMs creating poems, stories, connecting concepts which you wouldn't otherwise see connected, birthing ideas in images that video which had never existed before.
Of course, I'm aware that by this logic many things fall apart (for example, one might be tempted to believe in god because of it). Nonetheless, on this issue, I am deeply on the creativity camp. And while I am there, the things I get LLMs to create, and people I know do so as well, will continue to induce deep emotion and connection among ourselves.
If that's not creativity to some, well, that's life, full of creative ways to define creativity or eschew it from the world.
LLMs can't innovate form. And form/format innovation is the core of creativity. Neither can RL or frontier. That's creativity, the ability to sense wordless states beyond existing status quo cause effect exchanges. Tokens have no game here. It's that simple, it's how cutting-edges are broken through. It's always trapped behind the aesthetics input, that's just another fatal blow to it's ability to create.
I'm not concerned with "hundreds" of people's opinions. And stories, poems, these are dead/dying forms.
You may be on the creativity camp, stay there. It's generic as creativity. From my POV, someone building a hybrid between game and movies, this tech is dead dull artifactual, like a fossilized version of art.
Yes of course we have definitions of creativity, see Anna Abraham. "creativity is doing something unique and new, using existing networks for things they weren't designed for"
And we have the wordless processes that materialze as affinities and events in the imagination.
I'm very struck by coders who are neither engineers, scientists, nor creatives in the arts, yet have the false intuition and self-delusion that this code is creative. Nonsense.
I'm afraid LLMs have zilch in comprable abilities to both as they exclude words, symbols, tokens etc.
Perhaps in their basic form but you could do a variation on them like
>AlphaEvolve is an evolutionary coding agent for designing advanced algorithms based on large language models such as Gemini. https://en.wikipedia.org/wiki/AlphaEvolve
which came up with a novel improvement in matrix multiplication
It's not evolutionary, that's functionality. There are no algorithms in nature.
Evolution is a tinkering process that finds by selection, and functionality is what we post-hoc add to the success.
Anything reduced to symbols is simply functionality within models, to claim it's creative is illusory. Creativity, imagination, etc exist in a realm beyond, and preceding symbols, metaphors, etc. Creativity is 'thinking around symbolic bottlenecks'.
The matrix multiplication breakthrough is still under the glass ceiling symbols provide. A new way to do math isn't the same as visual scale invariant paradoxes. It's not creativity, the word is a metaphor, the closer word is optimization.
As a (rather obsessive, perhaps compulsive) poet, I will indeed remain in my camp :)
(I do get what you're saying. Yet, I am not convinced that processes such as tokenization, and the inherent discretization it entails, are incompatible with creativity. We barely understand ourselves, and even then we know that we do discretize several things in our own processes, so it's really hard for me to just believe that tokenization inherently means no creativity).
Keep in mind there are events, they are real. Words are just symbols we use to make the thoughts or the events cohere as memories, but they're artifacts that have no direct connection to the thoughts. Nor do any tokens, symbols, codes, etc. These are bottlenecks. They have no analog, they lack specificity. Creativity is the ecological exchange between body and ecology to make records that engage paradox wordlessly.
That's what the article's big reveal is about. The events (which can be externalized creatively) are not really creative as words. That's a big problem for the species in general. That's a glass ceiling no one is recognizing or taking notice of. And that sort of gives you an idea how poetry and code are trapped behind it.
Why is someone from Costa Rica sending US antivax propaganda posts into HN? These are the same poorly defined and badly executed narrative arguments that mythologize vaccine science from conspiracy theorists.
the Nature paper is empirically spot-on and theoretically obsolete at the same time
it reflects a pre-1990s dogmatic belief in extrapolationist macroevolution.
It treats high heritability + polygenic additive variance as if that somehow proves the Modern Synthesis was right about everything
It never once mentions developmental constraint, niche construction, active rGE as a macroevolutionary ratchet, cultural-genetic co-evolution, or any of the other mechanisms that the post-2000 literature now recognizes as necessary for understanding why human g exploded in the last million years
in other words, eugenics. we have a culling policy like in early animal domestication: herding pens outside the city walls to test whether infected animals are dangerous. only here, the culling is in the open society, without any advance warnings, which are suppressed from public knowledge as policy. it's far more dystopian than the hunger games.
In 1990 I built an SE/30 from spare parts, used an 800K fujitsu rather than a 1.2 superdrive, an Ehman grayscale page-sized monitor (remember those) all in an AT tower using a PC power supply, wired to the logic board using the Addison-Wesley wiring diagrams; I could never cover it because the Ehman card exceeded the tower dimensions, and it powered up the moment I pressed the AT's switch.
When I brought in a polaroid of the unit to that early TekServe spot on 18th street, David made a sign around it and wrote, "Why pay full price? SE/30s can be made for $1500 all-in!"
Yes! I saw that and read it but didn't favorite in time. And I'd heard some friends talking about these papers a year ago and we'd brought them up in our lunchtime chats, So when I found the link notes, I posted them.
The first one was flagged, here it is. This post here was a more specific study.
