Friday, November 25, 2016

Google DeepMind - Artificial Intelligence - Lip Reading and BackSeatDriving

https://deepmind.com/

Artificial Intelligence is like a child learning rapidly.  Learning to apply new abilities as approximations progressing toward increasing accuracy in Natural Language processing using Artificial Language systems.

Machine parsing of spoken Natural Language using AI is now common and accurate.  We see it as simultaneous translation when closed captioning is turned on in media we watch.  We see it elsewhere.  Much to my surprise my IPhone 7 recently gave me a printed text interpretation of a voice mail message.  A new OS update default based feature I suppose that I could maybe turn off with a setting?  There was one spelling error in the translation of voice to text due.  The VoiceToText:SpellChecher:InMyHead in my head (analogous to the spell checker applied to this blog entry that just caught me on a mis-spelling of "analogous") correctly interpreted the incorrect voice to text translation. 

Amazing.

It is unfortunate that the state of the art of AI is not yet at the point of selecting our POTUS based on the next step in AI to process information to a state of Knowledge.  Given its rapid advancement in childhood development perhaps we will be smart enough to recognized by the time of the next election that a Child should lead us.  Make our rational logical decisions on what is best for us.

Why not?

That is where we are going to be driven on "driverless" cars that do the driving better than us.

Not there yet but look at the trend line.  We are dumbing down.  Our machines are smartening up!  Correcting us when we are wrong.  My SpellChecker underlines "dumbing" for my attention but does not give me a correct spelling of the word.  I wish I had all these crutches available to me in college (1961-1965).  I would not have had to study English spelling and writing in grade school.

GoogleMind..............It and other AI programs learned real time translation of spoken words to text.  Now it is learning to lip read with greater accuracy than a human being.
http://www.techrepublic.com/article/google-deepmind-ai-destroys-human-expert-in-lip-reading-competition/

I spent a long time this morning reading this...reading the text..because I thought it was worth reading....even more time trying to fit it into my ObjectOriented:KnowledgeBase:FrameOfReference.
https://papers.ssrn.com/sol3/papers.cfm?abstract_id=2837929.

What I need is an AI assistant (good looking female like in the movie) that would read the link and tell me if it confirms my upper level conceptual structure of IntelligentKnowledge:Debt or not.  If not what must I examine for better FundamentalHighAbstractLevelComprehension of what a HighConceptualParentLevelThing like Debt "IS" and what it "Does" in relationship to other High (AndHigher)ConceptualParentLevelThings "Are" and "Do".

What I need is some AI thing (ApplicationProgram) to do my logic and reasoning truth based thinking analysis for me and translate my thinking into IntelligentKnowledgeBasedAction as a function of the OperatingSystem that I created for myself or........ an OperatingSystem created for me that is AI based that I choose to use because I believe in faith and reason that it is structurally and operationally better and therefore my guide.....that I will choose to over-ride and take control of if and when necessary because I am human and it is not.

However.....for those not as intelligent and knowledgeable as me.....I want the AIOperatingSystem and its ApplicationPrograms to not just read their lips but read their minds and do the steering for them while they sit in the back seat because they are to dangerous or drunk on self destructive ignorance to be trusted at the wheel???

At some time in the near future perhaps everything that Trump has said and done will be input to an AI parser that attempts to extract High Level Intelligence and Knowledge and MakeJudgment.

WTF?

World of Watson Conference: http://davebeulke.com/5-important-ideas-ibm-world-watson-conference/

http://languagelog.ldc.upenn.edu/nll/?p=29524






No comments: