Sometimes when I read these discussions I get frustrated by my inability to
distingish between genius and nonsense. It is a basic property of all
intelligent agents that they cannot detect higher intelligence than
themselves. Your score on an IQ test depends on how many questions you get
right. But the person scoring the test must know what the correct answers
are, so their IQ must be higher.

We can still measure how fast you got the answers. People marveled when
vacuum tube computers in the 1950s could multiply 20 digit numbers in under
a second. But we could still check the answers in a few minutes. With
Navier Stokes it is not so easy. I can read the proof but all I can do is
gloss over the equations. AI has been known to make up answers and express
them condfidently when it actually doesn't know, because it is trained on
human data and humans do the same thing. People like to sound smart by
throwing together lots of big words to express grammatically correct
nonsense.

It's an old problem. It's why it takes decades to award a Nobel prize. It's
why we don't have an AI test better than the Turing test, where the highest
possible score is human level.

With regard to E8 or various string theories of everything, they all fail
to make experimentally verifiable predictions. The reason we don't have a
TOE is that by Wolpert's law, no system can have complete knowledge of the
system that contains it.  Quantum mechanics describes a deterministic wave
equation where observers see random particles. Relativity describes
observers that see space and time differently from each other. Whatever TOE
unifies both also has to explain why time has a direction when both quantum
mechanics and general relativity are symmetric with regard to time.

I like the multiverse theory because it is algorithmically simple. If the
physical constants like G, h, c, and the masses and charges of protons and
electrons were any different, then stars would not form or create planets
with the right elements to form complex molecules that could encode
information and self replicate. Those other universes exist, but we
necessarily observe one where intelligent life is possible.

-- Matt Mahoney, [email protected]

------------------------------------------
Artificial General Intelligence List: AGI
Permalink: 
https://agi.topicbox.com/groups/agi/T906c878e4ba411be-M383eebf14f32dfe1f8b1d7ec
Delivery options: https://agi.topicbox.com/groups/agi/subscription

Reply via email to