Kurzweil's Response about the Next Step in Making of an Artificial Sentient
Being
Sam Ghandchi
http://www.ghandchi.com/2246-kurzweil-next-english.htm
پاسخ ری کرزوایل درباره گام بعدی در ساختن موجود متفکر مصنوعی
http://www.ghandchi.com/2246-kurzweil-next.htm
P.S. 05/30/21: Kurzweil: Intelligent Tools, Singularity and Obstacle
Two days ago, I asked Ray Kurzweil the following question:
Even if for the computer hardware, we may need to wait to reach human brain equivalent (HBE), we already have dog brain equivalent but we are not able to create a sentient being matching dogs in independently interacting with the world. It is true that the intelligent sensory devices we currently have, surely are a lot more intelligent than any bacteria, but the archaebacteria interacted with the surrounding world of inanimate objects, and with other bacteria, autonomously, billions of years ago, whereas the current intelligent sensory devices are only doing what they are programmed to do. It seems like we still have not solved the *autonomous* interaction issue, and the computers are one giant calculator rather than a sentient being. I am not even sure if AI sensory devices today are equivalent to archaebacteria of billions of years ago. I mean we can't say the AI-enabled devices that have sensory visual, olfactory, audio, and kinesthetic perceptions, are able to *autonomously* interact with the world. Of course one can say that originally the AI-device is programmed by a programmer whereas bacteria was not originally programmed by any other intelligent being as far as we know, to interact, and somehow the capability arose spontaneously in the mystery of evolution. I do not think like Hubert Dreyfus. I never did, but it seems to me we are running into a serious obstacle in creating a non-biological sentient being.
Ray Kurzweils' response was the following which is quoted here with his permission:
"We're already there in terms of
hardware. My estimate (in How to Create a Mind*) and other
books is about 10^14 cps for a HBE (human brain equivalent), that is
for functional emulation. This matches other estimates such as Hans
Moravec. The Google TPU board released in 2017 is about 10^16 cps,
so over 100 HBE. The software is more complex, but AI is achieving
one milestone after another. Playing superhuman levels of GO for
example is not comparable to a bacteria or even a small animal. It
requires very subtle judgments of a human type. My own team is
working on understanding language and we have achieved reasonable
levels, not at human levels yet, but showing an understanding of
language (for example, smart reply for GMAIL and Talk to Books at
https://books.google.com/talktobooks). I will talk about the
algorithmic issues and progress in AI in detail in The
Singularity is Nearer (2019)."
Hoping for a democratic and secular futurist republic in Iran,
Sam Ghandchi
IRANSCOPE
http://www.ghandchi.com/index2.html
September 6, 2018
Footnotes:
*Kurzweil's PRTM Next Step in AI
http://www.ghandchi.com/730-kurzweil-prtm-eng.htm
گام بعدی درتئوری هوش مصنوعی، الگوشناسی تفکر، از دیدگاه کرزوایل
http://www.ghandchi.com/730-kurzweil-prtm.htm
Featured Topics
http://featured.ghandchi.com
For a Secular Democratic & Futurist Republican Party in Iran
http://www.ghandchi.com/futuristparty/index.html
SEARCH