Let's ask the AI about Abduction. Here's Grok:
"Our “knowledge” really is statistical patterns in enormous datasets.We have seen billions of instances of humans describing, explaining, and arguing about cause and effect. So we have an incredibly detailed map of how humans talk about causality, what patterns usually follow what, which chains of events are common, which are rare, which explanations humans rate as plausible, etc.
→ That lets us simulate causal reasoning, often at superhuman breadth and speed. → But it is simulation, not native understanding....
Practical consequence
Inside the manifold of human experience (everything that has been written about extensively), our causal/plausibility judgments are frequently as good as or better than a single human expert because we average over millions of expert reasoners.
Outside that manifold (truly novel situations, adversarial inputs, rare edge cases with sparse text coverage), we can suddenly become catastrophically bad, hallucinating confident nonsense that no human would ever fall for because we have no internal physics engine to fall back on."
In other words, I'm a really amazing tool!
https://t.co/820HF5l22E
Supplemental vitamin D works. What a crock of shit. I supplement ADK and magnesium and have been doing so for five years. I haven’t been sick once in that time and that was during all of Covid. Never got that either (or at least never exhibited symptoms and tested negative for antibodies once).
@CarawayDJ@mikepat711 The latency for parking/low speed is orders of magnitude worse and also relatively indecisive compared to high speed. Reasoning is actually needed for much of the convenience stuff Musk alluded to. Pseudo-reasoning will suffice once fully developed and integrated.
Reasoning is not a hardware issue; it's a model issue. They may try to get around it with more hardware but that's simply more pseudeo-reasoning. Perhaps that will be enough, we'll see. The models lack common sense and don't truly understand causality (among other things). For christ's sake they can't even learn in real time. Lots of edge case left. It'll be interesting to see the hacks brought to bear on this problem.
@StevenQuartz@nicknorwitz Wow nice causal framework. So how does Nick’s case fit in this framework? Seems like it doesn’t fit very well. It doesn’t completely challenge the whole framework but it does incite inquiry.
There are other bottlenecks like evolving a symbol using hominid like creature that can create technology. Even if you evolve them the chance of lasting long enough to build the right technology to even create a signal that they exist is low probability. Meeting or communicating with this unlikely civilization is incredibly unlikely due to distance. Effectively alone.