Take a gander at the OpenWorm project. It's a great example of how simple neuronal activity is (given details like the connections, number of receptors, and transmitter infrastructure). SOTA models of neuronal activity are simple enough for problem sets in undergraduate biomedical engineering programs.
Sure, to your point, we don't know. But the worm above (nematode) swims and seeks food when dropped into a physics engine.
My main point is that the scale of the human brain is well beyond the capabilities of modern imaging modalities, and it will likely remain so indefinitely. Fascicles we can image, individual axons we cannot. I guess, theoretically, we'll eventually be able to (but it's not relevant to us or any of our remote descendants).
If they freeze the vesicles that deliver transmitters and make them analyzable, you've got all the information you need. In terms of a modern ANN, it's the connections (axons) and the weights (transmitters/receptors in tandem).
That said, this article doesn't get to the point in the free section. How are they collecting the information? Slicing is inherently destructive. Someone's got to manufacture an entirely novel imaging modality. Perhaps they could scan millimeters ahead of the slice at a resolution high enough to image receptors. Not possible currently.
I have a good analogy. 10 years ago, I was convinced that a 24-inch 1080p monitor at arm's length was perfection. There could never be any reason to improve over it. I could do everything I ever wanted to, to a standard I would never need to improve upon.
Yet here we are. The simplest and most obvious improvement is a 24" 4k monitor at 200% scaling. Basically, better in every way.
There's a discussion to be had about whether you need the better setup, which I think is your point, but there's no denying you'd want it (all other variables the same).
Yes. Not a drug in the typical sense. Simply replacing the protein in which the mice are deficient (that was identified by measuring levels in human spinal fluid).
Peter often reiterates that he doesn't recommend the use of open models for OpenClaw. They're much weaker when it comes to prompt injection, being the main reason. I'd be interested to understand the security measures put in place here.
the complexity of advanced connectomes is so far beyond our imaging capabilities that we have no way of knowing how far away from understanding intelligence we are
see the open worm project to get an idea of what artificial neuronal architecture requires to express anything meaningful. (and an interesting ethical perspective on digital consciousness.) my point being that the number of neurons is fairly meaningless. you could take neuron models and link them circuit-style to play doom at the 10^2 scale if you wanted. from a cellular neurophysiological perspective, there's nothing particularly special here (as opposed to sentience/intelligence that's a paradigm shift beyond our understanding). and, in my opinion, absolutely nothing to be even the slightest bit worried about ethically.
This will be a Harvard Business case study on market share.
Claude Code was instrumental for Anthropic.
What's interesting is that people haven't heard of it/them outside of software development circles. I work on a volunteer project, a webapp basically, and even the other developers don't know the difference between Cursor and Claude Code.
It's impressive, even if the books and the posts you're talking about were both key parts of the training data.
There are many academic domains where the research portion of a PhD is essentially what the model just did. For example, PhD students in some of the humanities will spend years combing ancient sources for specific combinations of prepositions and objects, only to write a paper showing that the previous scholars were wrong (and that a particular preposition has examples of being used with people rather than places).
This sort of experiment shows that Opus would be good at that. I'm assuming it's trivial for the OP to extend their experiment to determine how many times "wingardium leviosa" was used on an object rather than a person.
(It's worth noting that other models are decent at this, and you would need to find a way to benchmark between them.)
I've heard Codex CLI called a scalpel, and this resonates. You wouldn't use a scalpel for a major carving project.
To come back to my earlier comment, though, my main approach makes sense in this context. I let Opus do the abstract thinking, and then OpenAI's models handle the fine details.
On a side note, I've also spent a fair amount of time messing around around in Codex CLI as I have a Pro subscription. It rapidly becomes apparent that it does exactly what you tell it even if an obvious improvement is trivial. Opus is on the other end of the spectrum here. you have to be fairly explicit with Opus intructing it to not add spurious improvements.
That's a good point. I'm not familiar enough with the various moats to comment.
I was just talking at a high level. If transformers are HDD technology, maybe there's SSD right around the corner that's a paradigm shift for the whole industry (but for the average user just looks like better/smarter models). It's a very new field, and it's not unrealistic that major discoveries shake things up in the next decade or less.
Yes, I only plan/implement on fully AI projects where it's easy for me to tell whether or not they're doing the thing I want regardless of whether or not they've rewritten the codebase.
For actual work that I bill for, I go in with intructions to do minimal changes, and then I carefully review/edit everything.
That being said, the "toy" fully-AI projects I work with have evolved to the point where I regularly accomplish things I never (never ever) would have without the models.
I'm suprised there isn't more "hope" in this area. Even things like the GPT Pro models; surely that sort of reasoning/synthesis will eventually make its way into local models. And that's something that's already been discovered.
Just the other day I was reading a paper about ANNs whose connections aren't strictly feedforward but, rather, circular connections proliferate. It increases expressiveness at the (huge) cost of eliminating the current gradient descent algorithms. As compute gets cheaper and cheaper, these things will become feasible (greater expressiveness, after all, equates to greater intelligence).
I'd encourage you to try the -codex family with the highest reasoning.
I can't comment on Opus in CC because I've never bit the bullet and paid the subscription, but I have worked my way up to the $200/month Cursor subscription and the 5.2 codex models blow Opus out of the water in my experience (obviously very subjective).
I arrived at making plans with Opus and then implementing with the OpenAI model. The speed of Opus is much better for planning.
I'm willing to believe that CC/Opus is truly the overall best; I'm only commenting because you mentioned Cursor, where I'm fairly confident it's not. I'm basing my judgement on "how frequently does it do what I want the first time".
Personally, I'm fascinated by the opening for protocol languages to become relevant.
The previous generations of AI (AI in the academic sense) like JASON, when combined with a protocol language like BSPL, seems like the easiest way to organize agent armies in ways that "guarantee" specific outcomes.
The example above is very cool, but I'm not sure how flexible it would be (and there's the obvious cost concern). But, then again, I may be going far down the overengineering route.
I'd be interested to hear some use cases people have for large contexts on an 8B model. Other than sentiment analysis or summarization (this release implies agentic use). My experience with the general intelligence of agentic interactions is that everything is unusable before 32B for any context greater than 4k tokens.
You just need a robust benchmark. As long as you understand your benchmark, you can trust the results.
We have a hard OCR problem.
It's very easy to make high-confidence benchmarks for OCR problems (just type out the ground truth by hand), so it's easy to trust the benchmark. Think accuracy and token F1. I'm talking about highly complex OCR that requires a heavyweight model.
Scout (Meta), a very small/weak model, is outperforming Gemini Flash. This is highly unexpected and a huge cost savings.