openai.com for instance has a 'unsolved' type of task: reading and 
understanding english descriptions of programs and turn them into code.
Looking at some of the descriptions reveals that it requires a lot of 
imagination and applying wisdom to contexts.

I'm personally interested in how well a machine might perform planning 
advertising for target audience or estimating the value of a brand for 
instance by looking at Facebook likes, how people react to the brand name 
and similar.

The more I think about how I solve problems the more I conclude its a mix 
of 1) creativity 2) wisdom (remembered details) 3) applying and following 
patterns such like a stack machine 4) trying different strategies such as 
how to reach the target or knowing the starting point what could I do ... 

Do you have any hints about where to start reading about which solutions 
exist right today to understand when computers will do my job?

Companies like sourced.tech are trying to understand huge amounts of code 
and links 
like http://news.mit.edu/2016/faster-automatic-bug-repair-code-errors-0129 
indicate that AI might be useful to apply to real world problems fixing 
bugs today ..

I personally have spend huge amounts of time *reinventing the wheel* in the 
past - for instance when implementing a XDEBUG protocol plugin for Vim 
requiring some creativity - Vim strings cannot represent 0 bytes - but the 
protocol is using them as well as stupid work (fixing regex based XML 
parser library) ... But in the end Vim is flawed because VimL code can be 
interrupted when on resize autocommands without you as programming having a 
chance to setup locking.

So the perfect AI would return a reply such as:
1) a quick fix would be fixing this VIML that way
2) a real fix would be rewriting Vim in a different language (rust/go) and 
rewrite all the event handling because its broken by design...

There are a lot of fun tasks such as "write a completion engine" for all 
editors out there or turn an editor into something usable within a browser 
(I know about llvm -> JS) -> which I know can be done triviialy once you 
spend enough time on the tasks...

Once new forks arrive (neovim) (same happened to emacs in the past for 
legal/license reasons) ... the funnier it gets and the more AI could assist.

I don't fear more intelligent AI - neither do I fear more intelligent 
people :) Given the same environment things get more and more similar - 
same will apply to humans/AI.

chess/go -> soon StarCraft (Blizzard) -> See deepmind people and machines 
have been showing similar trates (over training) for quite a while as well.

Marc Weber

-- 
You received this message because you are subscribed to the Google Groups 
"opencog" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to [email protected].
To post to this group, send email to [email protected].
Visit this group at https://groups.google.com/group/opencog.
To view this discussion on the web visit 
https://groups.google.com/d/msgid/opencog/12c7c94d-eb24-41aa-ba57-ef04736345a9%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

Reply via email to