On Sat, Jan 9, 2021, 2:25 PM OR1 <[email protected]> wrote:

> On Saturday, January 09, 2021, at 6:10 PM, Bill Hibbard wrote:
>
> Interesting question. I don't know if AGI must/will be conscious. Hutter's
> Universal AI is at least a theoretical example of an AGI that is not
> conscious, although it would likely appear to be conscious to people it
> interacted with.
>
> "Hutter's Universal AI is at least a theoretical example of an AGI that is
> not conscious"
>
> Can you explain to me briefly what universal artificial intelligence is?
>

The expected reward of a reinforcement learner over a universal
(algorithmic) distribution of computable environments. A universal
distribution favors simpler environments, such that if the description
length is n bits, its probability is 1/2^n.
https://arxiv.org/abs/0712.3329

Hutter also proved that the optimal strategy is not computable. That
strategy, called AIXI, is to guess the simplest environment consistent with
observation. It is not computable because testing all possible environments
against observation means excluding the ones that run forever

So the best we can do is train an intelligence over a subset of practical
environments.

Now can you explain to me briefly what consciousness is?

------------------------------------------
Artificial General Intelligence List: AGI
Permalink: 
https://agi.topicbox.com/groups/agi/Ta8cf1690f2bebe0e-Mf40151b3b882761be7d86a79
Delivery options: https://agi.topicbox.com/groups/agi/subscription

Reply via email to