Welcome to European Tribune. It's gone a bit quiet around here these days, but it's still going.
Display:
We went through the same hysteria* with Watson/DeepBlue after it won the Jeopardy match in 2011.  One result was MD Anderson pouring $60 million into a joint venture and achieved nothing.

The MD Anderson nightmare doesn't stand on its own. I regularly hear from startup founders in the AI space that their own financial services and biotech clients have had similar experiences working with IBM.

The narrative isn't the product of any single malfunction, but rather the result of overhyped marketing, deficiencies in operating with deep learning and GPUs and intensive data preparation demands.

AlphaGo is a Go playing Expert System.  It can never be more than that because Neural Net technology does not allow it to be anything more.  Both Hinton ("We need to start over") and LeCun (They don't work in the real world") let the cat out of the bag on that one.  

* a psychological disorder whose symptoms include selective amnesia, shallow volatile emotions, and over dramatic or attention-seeking behavior.  

She believed in nothing; only her skepticism kept her from being an atheist. -- Jean-Paul Sartre

by ATinNM on Sun Oct 22nd, 2017 at 12:15:40 AM EST
[ Parent ]
When Watson will be responsible for its own marketing, then we have Artificial Intelligence.

Learning Go for 960 hours straight up to a superb level is kinda cool, but... extremely dorky.

The annual top chess engine competition is under way. No Google or IBM products there, because staying behind an open source champion would be bad marketing.

by das monde on Sun Oct 22nd, 2017 at 03:59:34 AM EST
[ Parent ]
"AI" per se is an insidious oxymoron. As I said, "man makes the machine." A machine, automation or automated processes --a type of system requires standardization of its constituents (systematic value) in order to guarantee predictability of its purposes-- are by definition artificial, not "natural," phenomena. Thus, as AK has said, "It can never be more than that", strictly speaking  because its human designer(s> do "not allow it to be anything more." The limiting factor even to parallel processing or "distributed" schemes on which astonishing! computing power rely, I think, will always be the programmatic scope of intelligence its engineers define.

I've been considering bitcoin and npm "block chain" theory and praxis as emulations of "best practices" in computing R&D. And I am not persuaded this method can or will obtain efficiencies in production or "innovation" to either human or machine benefit searches. The I/O and iteration are chiefly copies of prior "art." That is, the "system" lacks discriminating purpose.

Diversity is the key to economic and political evolution.

by Cat on Sun Oct 22nd, 2017 at 12:03:41 PM EST
[ Parent ]


Diversity is the key to economic and political evolution.
by Cat on Sun Oct 22nd, 2017 at 12:07:33 PM EST
[ Parent ]
THIS JUST IN!
Routes to Defensibility for your AI Startup
Copy of copy of copy of "a competitive strategy" is search of a business and unmet demand.

(I walk to my stacks, pull out Shapiro and Varian, Information Rules, a Strategic Guide to the Network Economy (1999) and Dolan and Simon,Power Pricing (1996)... grimace at Tho. Stewart, Wealth of Knowledge (2001)... return)

World of coders now equipped to transform 'barriers to entry' into moats.

Diversity is the key to economic and political evolution.

by Cat on Sun Oct 22nd, 2017 at 02:37:26 PM EST
[ Parent ]
If humans want Artificial God or artificial management (i.e., asking a machine to "make"/compel them), they may get that.

Corporations are AI entities already, Romney could say.

by das monde on Sun Oct 22nd, 2017 at 10:32:52 PM EST
[ Parent ]
You fundamentally misunderstand the point of AI.

Its debatable whether the aims are achievable, but it's absolutely clear that the development of independent, self-enhancing intelligence - which transcends the abilities of its developers - is a core aim.

And the aim has been attained in minor ways across many fields. Google's AlphaGo developers didn't need to understand Go to master level to be able build a system capable of beating a master.

The only difference with general AI is that the goal is to automate learning itself, not to solve specific problems in one limited domain.

Of course this requires the design of a system with discrimination and abstraction - not necessarily a purpose in human terms, but still an explicit meta-goal.

by ThatBritGuy (thatbritguy (at) googlemail.com) on Mon Oct 23rd, 2017 at 01:17:21 PM EST
[ Parent ]
props for the link: It refers to interesting (B) cases. The (A) cases being in a collection of primary and secondary research into "pricing knowledge management" I did around the turn of this century. IBM's decades-long "first mover advantage" and business model has been unraveling since the dotcom "boom". Probably SAP's, too. I no longer have the resources to follow up.

That said, all too many CS ventures, or "independent software vendors" (ISVs), share the same maintenance of effort dilemma. Apart from continually raising working capital (which determines marketing investment, inc. advertising and R&D)

The narrative business problem isn't the product of any single malfunction, but rather the result of overhyped marketing, deficiencies in operating with deep learning and GPUs and intensive data preparation demands. failure to identify and satisfy unmet demand


Diversity is the key to economic and political evolution.
by Cat on Sun Oct 22nd, 2017 at 10:03:01 AM EST
[ Parent ]

Display:

Occasional Series