Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]

Groups > comp.programming.threads > #4681 > unrolled thread

Ben Goertzel on artificial intelligence

Back to article view | Back to comp.programming.threads

Contents

  Ben Goertzel on artificial intelligence Horizon68 <horizon@horizon.com> - 2018-10-21 12:24 -0700

#4681 — Ben Goertzel on artificial intelligence

Hello,

Read this:


I just read what wrote Ben Goertzel on artificial intelligence, he wrote 
the following:

"My own intuition is that the shortest path to AGI will be to use deep 
neural nets for what they’re best at and to hybridize them with more 
abstract AI methods like logic systems, in order to handle more advanced 
aspects of human-like cognition."

That's good to know.

And he wrote also this:

"In the 15 years since I first introduced the term “artificial general 
intelligence” (AGI), the AI field has advanced tremendously. We now have 
self-driving cars, automated face recognition and image captioning, 
machine translation and expert AI game-players, and so much more.
However, these achievements remain essentially in the domain of “narrow AI”"


Read more here:

https://singularityhub.com/2018/07/22/from-here-to-human-level-artificial-general-intelligence-in-four-not-all-that-simple-steps/


I think he is not correct to say that it is still narrow intelligence, 
read my following writing to understand better why:


Artificial intelligence is becoming "powerful" because we are now
capable of "general" artificial intelligence, read carefully what
follows, it said that:

"Artificial intelligence up to now, including DeepMind’s AlphaGo, have
been limited to perform well only for a specific task - hence the term
‘narrow artificial intelligence.’ But researchers are beginning to
embrace the arrival of artificial general intelligence, an AI that can
perform any task that a human can. In their recent paper, Google
DeepMind proposes a method into training AGI by using genetic algorithm."

And:

"So how well does PathNet perform in action? DeepMind’s experiments show
that PathNet is faster than fine-tuning, and does positively transfer
trained knowledge in supervised learning classification tasks and
reinforcement learning tasks - successfully addressing the three
requirements of an AGI."

Read more here to notice it:

http://www.deepbio.co.kr/2017/03/20/Weekly-paper-DeepMinds-PathNet.html


And read about the following other important step that is Generative
Adversarial Network:

What’s a Generative Adversarial Network? Leading Researcher Explains

Read more here:

https://blogs.nvidia.com/blog/2017/05/17/generative-adversarial-network/




Thank you,
Amine Moulay Ramdane.

[toc] | [standalone]

Back to top | Article view | comp.programming.threads

csiph-web