Re: How long before superintelligence?
- From: Tim Tyler <seemysig@xxxxxxxxxxxxxx>
- Date: Tue, 28 Oct 2008 22:19:08 +0000
Alpha wrote:
On Oct 28, 7:45 am, Tim Tyler <seemy...@xxxxxxxxxxxxxx> wrote:
I've put together a video of me - discussing my estimate of the
time of arrival of superintelligent machines - and put it on
YouTube:
http://alife.co.uk/essays/how_long_before_superintelligence/
For human-machine superintelligence - 25-50 years; SI is likely a
certainty via GE (improvement) of the already-premiere and
unquestioned champion of SI - the human being).
For machines - perhaps 100-1000 years or maybe never [...]
Right. Thanks for giving an estimate.
AFA as your "talk", there is a lot on nonsense therein, like the ref
to Moravec's dismally estimate of the compute power of a neuron; he
obviously doesn't understand the work of Koch for example.
Christof Koch mostly seems to have reached his conclusions
via poking around in working brains. However, I can't see very
much in the way of actual estimates of computational capacity
that differ from the estimates of Moravec and Watts. Has
Koch actually produced any estimates?
And you have some incoherent aspects explicated as well to wit:
"The detailed form of proteins arises from a combination of the
nucleotide sequence that specifies them, the cytoplasmic environment
in which gene expression takes place, and the laws of physics.
We can safely ignore the contribution of cytoplasmic inheritance
"
The word inheritence should probably be "environment";
Nope - that was the way I intended it.
and it is false
to conclude that the environment actually can be ignored, as it is the
gene plus environment that leads to the phenotype; one switch gene,
switched on or off by a simple environmental event, can change one
from a dunce to an Einstein.
Assuming we are discussing the era before environmental inheritance
took off (via memes), there are good reasons for discounting it, IMHO.
The environment affects gene expression, but the issue is whether
different aspects of environments are /reliably/ inherited across
extended periods of time. Many sources of environmental inheritance
are analog, and have associated poor copying fidelity.
There are a /few/ bits of the cytoplasm which are inherited, *and*
are not themselves derived from nucleic acids - but it doesn't add
up to very many bits. The most likely candidates are centrioles
and methylation - both of which are intimately associated with DNA.
There is regular excitement about inheritance via these mechanisms -
but the evidence of inheritance that persists across multiple
generations with reasonable fidelity is still weak, even for these.
There are also too many incorrect assumptions throughout to list. One
involved in: "We will have probably machine intelligence long before
such projects gets very far off the ground." WRT reverse engineering
the human brain section of your talk. It simply isn;t necessary to
reverse engineer the brain to experiment with GE/drug approaches to SI
in humans.
We're not going to get much in the way of superintelligence via humans -
unless we fuse their brains together in a matrix, or liberate their brains
from their skulls and grow them in vats. We /could/ do those things - but,
practically speaking, it doesn't make much sense to do so.
I don't deal with genetic engineering since I do not think it will be
relevant to the origin of superintelligence. Much too little, much too late.
--
__________
|im |yler http://timtyler.org/ tim@xxxxxxxxxxx Remove lock to reply.
.
- Follow-Ups:
- Re: How long before superintelligence?
- From: Alpha
- Re: How long before superintelligence?
- References:
- How long before superintelligence?
- From: Tim Tyler
- Re: How long before superintelligence?
- From: Alpha
- How long before superintelligence?
- Prev by Date: Re: How long before superintelligence?
- Next by Date: Re: Feynman on Quantum Mechanics (for Don G.)
- Previous by thread: Re: How long before superintelligence?
- Next by thread: Re: How long before superintelligence?
- Index(es):
Relevant Pages
|
