zlacker

I think there is a good chance you are roughly right. I also think that the "secret sauce" of sapience is probably not something that can be replicated easily with the technology we have now, like LLMs. They're missing contextual awareness and processing which is absolutely necessary for real reasoning.

But even so, solving that problem feels much more attainable than it used to be.

>>ivraat+(OP)
i agree. it feels like scaling up these large models is such an inefficient route that seems to be warranting new ideas (test-time compute, etc).

we'll likely reach a point where it's infeasible for deep learning to completely encompass human-level reasoning, and we'll need neuroscience discoveries to continue progress. altman seems to be hyping up "bigger is better," not just for model parameters but openai's valuation.

>>ivraat+(OP)
I think the missing secret sauce is an equivalent to neuroplasticity. Human brains are constantly being rewired and optimized at every level: synapses and their channels undergo long term potentiation and depression, new connections are formed and useless ones pruned, and the whole system can sometimes remap functions to different parts of the brain when another suffers catastrophic damage. I don’t know enough about the matrix multiplication operations that power LLMs, but it’s hard to imagine how that kind of organic reorganization would be possible with GPUs matmul. It’d require some sort of advanced “self aware” profile guided optimization and not just trial and error noodling with Torch ops or CUDA kernels.

I assume that thanks to the universal approximation theorem it’s theoretically possible to emulate the physical mechanism, but at what hardware and training cost? I’ve done back of the napkin math on this before [1] and the number of “parameters” in the brain is at least 2-4 orders of magnitude more than state of the art models. But that’s just the current weights, what about the history that actually enables the plasticity? Channel threshold potentials are also continuous rather than discreet and emulating them might require the full fp64 so I’m not sure how we’re even going to get to the memory requirements in the next decade, let alone whether any architecture on the horizon can emulate neuroplasticity.

Then there’s the whole problem of a true physical feedback loop with which the AI can run experiments to learn against external reward functions and the core survival reward function at the core of evolution might itself be critical but that’s getting deep into the research and philosophy on the nature of intelligence.

[1] >>40313672

>>throwu+lb
Transformers already are very flexible. We know that we can basically strip blocks at will, reorder modules, transform their input in predictable ways, obstruct some features and they will after a very short period of re-training get back to basically the same capabilities they had before. Fascinating stuff.