I think AGI is more than possible with a heuristic based logic system, but
it's not going to be any easier than a NN based approach, and probably
can't be wholly separate from each other. It's infeasible to use a
heuristic system for representing new concepts. Where heuristics provide a
low frictio
On Friday, July 15, 2022, at 7:01 PM, immortal.discoveries wrote:
> Just giving solid advice and a good example why...
Nah, you are toeing the neural network party line. There is more inside the
class of general learning algorithms than just that one... thing. MUCH more!
--
Just giving solid advice and a good example why...
--
Artificial General Intelligence List: AGI
Permalink:
https://agi.topicbox.com/groups/agi/T9019734479a51fd8-M9f6ab68ada8915d42e9599fb
Delivery options: https://agi.topicbox.com/groups/agi/subscription
On Friday, July 15, 2022, at 6:23 PM, immortal.discoveries wrote:
> You need neural networks to create AGI.
Talk about dogma, buddy!
--
Artificial General Intelligence List: AGI
Permalink:
https://agi.topicbox.com/groups/agi/T9019734479a51fd8-M9b7451be701b59
On Friday, July 15, 2022, at 5:09 AM, stefan.reich.maker.of.eye wrote:
> I am also searching for a replacement for neural network based learning. The
> bar is high, but on the other hand, neural networks have fundamental
> limitations, and I think they will one day be completely abandoned when
>
I am also searching for a replacement for neural network based learning. The
bar is high, but on the other hand, neural networks have fundamental
limitations, and I think they will one day be completely abandoned when
something better has been discovered. We don't know what that thing is yet
th
It would be ideal if we could run a query: given a set of input and output
pairs **I -> O**, construct a function **F** of type **simply typed lambda
calculus** that connects inputs and outputs in precision of **X%**.
*Rewrite* can't do that on its own, at least not in a simple, automated way. T
On Wednesday, July 13, 2022, at 8:19 AM, ivanmoony wrote:
> Does algorithm synthesis count as an acceptable way of figuring out the rules?
Sure, but that's just a word (or two words to be precise). How to actually do
it is the big question.
--
Artificial Gen
Does algorithm synthesis count as an acceptable way of figuring out the rules?
--
Artificial General Intelligence List: AGI
Permalink:
https://agi.topicbox.com/groups/agi/T9019734479a51fd8-Mf34bed998c54970a28693ce1
Delivery options: https://agi.topicbox.com/
Is it really though? All you're doing is shifting the onus of learning from
the machine to the human whose job it is to reduce the world into discrete
rules. That's the problem AI is trying to solve - let the machine do the
hard part of figuring out the rules. This doesn't really solve the problem
Now take the world. It takes some input and produces some output. If you are
able to construct a function that mimics the world's output, you can say you
know the world well.
--
Artificial General Intelligence List: AGI
Permalink:
https://agi.topicbox.com/
Take me for example. I take some input and produce some output. If you are able
to construct a function that mimics my outputs, you can say you know me well.
--
Artificial General Intelligence List: AGI
Permalink:
https://agi.topicbox.com/groups/agi/T901973
Only by observing which inputs and outputs come in pair.
--
Artificial General Intelligence List: AGI
Permalink:
https://agi.topicbox.com/groups/agi/T9019734479a51fd8-M6a67bde031fb3e3d0f04f054
Delivery options: https://agi.topicbox.com/groups/agi/subscriptio
Very interesting - but how would it learn?
--
Artificial General Intelligence List: AGI
Permalink:
https://agi.topicbox.com/groups/agi/T9019734479a51fd8-M9d44b687b5b274d3426b1062
Delivery options: https://agi.topicbox.com/groups/agi/subscription
My ultimate goal with term rewriting would be algorithm synthesis that could
replace NN learning approach.
The story is: we define in term rewriting system a programming language such is
simply typed lambda calculus or something similar. Then we supply a set of
inputs paired with expected outpu
I've been playing with rewriting systems so much back in the day. First using
Maude, and I did make my own rewriting engines too. Kind of moved on to other
things since I never really found a good application for this.
--
Artificial General Intelligence Lis
I hate to look like I'm ignoring this, but at the moment I guess I'm not sure
what I could use it for. The logic puzzle solver examples are the most
interesting - I might have to study them more when I get the time.
It looks pretty slick though, for someone with the right application.
Also, for
17 matches
Mail list logo