Sunday, July 28, 2013

// // Leave a Comment

Comments: "Strong AI manifesto", Free Access to Academic Papers - Comments on the H++ Transhumanist E-zine

String AI Manifesto

Submitted by Todor "Twenkid" Ar.. on July 28, 2013 at 7:33 am.

I support the points about the fake estimations of brain’s computing power, the quantum mechanical and parallel fake “magic”; the 486 DX or so as possibly be capable to run AGI code – on low resolution sensory data – and the over exaltation regarding “big data” – “we got faster computers, let’s throw to it 45849584958495 PBYTES”.

Human development, especially of blind and blind-deaf is an illustration that big-data is an excuse. There are other illustrations, regarding the way intelligence work, that I’ll skip here.

It does need Low-Level Sensory Data (as Sergio also suggests, “nature’s protocol”) to scale from scratch, but it doesn’t need 45894584958495849584958403 PBYTES of it to do it. That’s just a cheap excuse.

Humans get intelligent with miniature amount of input data, however it is fed and grabbed interactively and in the right sequence.

I claim, that current desktop PCs are more than enough for human-level (my term) Versatile limitless self-improver to develop, and I agree that I could ran on PCs a decade or two ago – just on lower resolution and slower.

I don’t like the term “Strong AI” – the “AI” word should be buried, it was polluted decades ago, I buried it myself 12 years ago.

A message in AGIRI list with a few words regarding the computational power, that I recalled now:

See also:

Giving Everyone a Million Dollar Education

Submitted by Todor "Twenkid" Ar.. on July 28, 2013 at 8:15 am.

Free access is nice, but there’s one big nonsense here – people DO NOT care about those papers – except a handful of them.

Giving a man who’s not intellectually gifted a book in Computer Architectures or in Biochemistry wont make her smarter, or giving someone a book in music theory wont make her a musician.

* No, there aren’t 2 billions of people who are willing to read scientific papers, in original.

* Yes, possibly there are 1000 or 10000 or who knows how many times less, most of them researchers, who are willing to read tiny little superficial digests from “Science Daily”, “Popular Mechanics” etc. from time to time. Well, probably many more read the scientific and technological columns in the electronic newspapers, which are reprinting the same news all around the world.

It’s almost impossible to make University students, even ones who are supposed to be very curious, to read and understand just one or two original papers, even if it’s regarding so advanced stuff such as Artificial General Intelligence, I have tried, with my students in the world first AGI University courses in 2010 and 2011.

Nevertheless – it’s hard to make researchers read a particular paper, there are too many already, why read precisely yours? Who are you?

PhD students read papers, because they “have to”, and very few of the most curious and gifted minds and other researchers.

Average people don’t even care, that’s not a job for every one, and free access won’t give them the intelligence and interests that are required.

Another issue is that the most of the contents of each paper is filling and “dead weight”, and the abstracts may give you the data you needed – unless you want to do additional research on the very specifics in the paper. However, if you do, you may be accused for “plagiarism”.

0 коментара: