Everything gets old, dirty, faulty...

• new car vs old car
• baby vs old woman
• new building vs old
• new road vs...

You get the idea.

Disorder [entropy] increases over time.

Phenomena in the world include fluid flow, fall due to gravity, oxidation, vibration, surface tension, electrical conduction and a LOT more.

Materials, in the form of structures, exhibit phenomena - it's how matter 'behaves'.

None of material behavior, involves computation.

Our brains are materials..........

Real bodies+brains deal with the real world in terms of phenomena, accumulate first-hand, interactive, continuous, 'experience'.

Brains don't compute [explicitly].

Experience is not computable!

Computation, numbers, written/read words, musical notation, math... is ALL a human invention! It's all symbol-based, and only carries SHARED meaning.

Cats don't compute. Elephants don't play chess [thanks, Rodney].

A von Neumann chip does explicitly compute - via explicit specification of quantitities.

ANY computation can be reversed!

Look up {fluid,particle,cloth,hair} simulation cache.

Eg. watch 0.00-0.30, 5:00-5:30, 8:30-8:50 of this clip: https://www.youtube.com/watch?v=M-VR3OEFDM0 [what you see is simulation being played backwards and forwards].

Reality cannot be reversed. Computation can.

Reality => not reversible.

Reversibility => not real.

Symbolic computation certainly can produce intelligent behavior. But that is not all there is.

Ex: Rube Goldberg machines, Braitenberg vehicles.

So..... ALL of AI to date, which relies on explicit computation, isn't, can't, be 'real'!

LLMs cannot become conscious the way biological animals are.

Multimodal LLMs don't deal with the 'real' world.

Symbolic AI, or neuro AI (data-driven), or neuro-symbolic - NONE of them can lead to 'real' AI, ie. one that mimics biological intelligence.

Why? Because data/computation is being used to mimic/substitute for reality - using a baseless assumption (called the PSSH).

What to do?