AI Thinks LIke a Corporation/Death of Insects

Mark Kohut mark.kohut at gmail.com
Wed Nov 28 05:46:18 CST 2018


The anecdote of the Go Champion being STUNNED, STUNNED I TELL YOU, by an*
early* move where the rock was placed way out of range of where the action
was happening is.......something.

In chess, which I know a little, the AI's recapitulated the whole history
of various openings in days and hours as well, with increasingly
perfectible moves and, of course never a blunder but "surpises" at the
grandmaster level only in the perfection of conceived sequences (hard at
that level to be in a position for a surprise as  in Fisher's Game of the
Century).

On Wed, Nov 28, 2018 at 6:32 AM Arthur Fuller <fuller.artful at gmail.com>
wrote:

> From an article on GeeksAreSexy.net:
>
> "By starting without any human-like preconceptions, AlphaGo Zero was able
> to develop strategies more suited to its capabilities. It still needs to be
> tested against human players, but one expert who analyzed the
> inter-computer games says it used techniques he had never previously seen.
>
> Google’s hope is that such an approach might work in other areas of
> artificial intelligence, with computers that develop techniques and
> procedures that make best use of a computer’s capacity rather than trying
> to refine the way human brains approach tasks."
>
> This is analogous to the history of man's attempts to fly. Early attempts
> were modeled on birds; men with wings attached by harness, etc. Then came
> the Wright brothers, and look what happened in 115 years, 1903 to last week
> -- 852 feet back then, Mars last week (54.6 km).
>
> Similarly, by teaching itself to play GO, AlphaGO Zero bypassed human
> preconceptions about the game, and came up with moves never before seen. AI
> will do the same, I think.
>
> On Wed, Nov 28, 2018 at 9:41 AM Mark Kohut <mark.kohut at gmail.com> wrote:
>
>> "Dave, I'm sorry. I can't do that, Dave"...
>>
>> On Wed, Nov 28, 2018 at 4:22 AM John Bailey <sundayjb at gmail.com> wrote:
>>
>>> Arthur's on the money. I was an AI skeptic like David for a long time,
>>> until I learned the current prevailing method of development:
>>> essentially pitting two AI against one another, each trying to
>>> convince the other that it is "real", although the criteria for that
>>> will vary. And each learns from the other's failures, and does a bit
>>> better, and so on and so on in a reciprocal manner that is only
>>> limited by the computing power and electricity. So yes, DM, they're
>>> already talking among themselves, so to speak. But they can have
>>> centuries of conversations in seconds.
>>> On Wed, Nov 28, 2018 at 8:00 PM Arthur Fuller <fuller.artful at gmail.com>
>>> wrote:
>>> >
>>> > There is an old religious/philosophical question, originally from old
>>> Jewish theology I think: if God is all-powerful, can he create something
>>> greater than Himself? Applied to AI, this question describes what Ray
>>> Kurzweil calls The Singularity. One has only to look at AlphaGO to see
>>> this. The original AlphaGO soundly thumped the world's best GO player,
>>> after having taught itself to play the game in two weeks, playing against
>>> itself. It successor, AlphaGO Zero, played a 100-game match against its
>>> progenitor, with a result of 100 games to zero.
>>> > One can generalize this phenomenon: an AI will design and build its
>>> own successor, and once that happens, further growth will proceed
>>> exponentially. Kurzweil defined The Singularity as the moment when AI
>>> becomes smarter than its creators. Once that happens -- and I (and others)
>>> believe it surely will, then all bets, and all considerations about our
>>> well-being, are off.
>>> >
>>> > Arthur
>>> >
>>> > On Wed, Nov 28, 2018 at 5:27 AM John Bailey <sundayjb at gmail.com>
>>> wrote:
>>> >>
>>> >> I think what the article makes clear is that what "we" want from AI
>>> >> doesn't matter - as far as I know nobody on the P-list is leading that
>>> >> charge, but certain people are and we shouldn't talk about the
>>> >> "progress" or "evolution" of a particular technology as if it's
>>> >> ahistorical and inevitable.
>>> >>
>>> >> A practical example: there's a lot of talk about the ethics of
>>> >> automated cars, and what their algorithms should take into account
>>> >> when deciding who dies in a crash. From all I've read/heard the
>>> >> discussion comes down to utilitarian ethics, and what would be the
>>> >> greater good in such a situation. But utilitarian ethics treats people
>>> >> as mathematical variables and is far from the only ethical model that
>>> >> could be applied, but it's the model that makes most sense from a
>>> >> programming standpoint, and perhaps the standpoint of a legal
>>> >> corporation trying to cover its posterior.
>>> >>
>>> >> Maybe the problem in AI thinking like a corporation is that
>>> >> corporations are very good at a lot of things (perpetuating their own
>>> >> survival, decentralised functioning, reorganising themselves to adapt
>>> >> to challenges, reducing individual culpability) but not so good at
>>> >> others (pretty much everything covered in the history of ethics).
>>> >> On Wed, Nov 28, 2018 at 4:08 PM David Morris <fqmorris at gmail.com>
>>> wrote:
>>> >> >
>>> >> > Does anyone think AI would be better with a chaos quotient?  I
>>> don't think so.  So Predictable Intelligence is our real goal. We want
>>> *smart* servants, not intelligence.  So, of course predictable AI will
>>> support corporate structures.
>>> >> >
>>> >> > it seems to me that AI is essentially imitative, not creative, not
>>> spontaneous.  It isn't really intelligent. We don't want it to talk back or
>>> even question us.  We won't ever tolerate that.
>>> >> >
>>> >> > David Morris
>>> >> >
>>> >> > On Tue, Nov 27, 2018 at 9:47 PM Ian Livingston <
>>> igrlivingston at gmail.com> wrote:
>>> >> >>
>>> >> >> Yep. Chiming in with gratitude, Rick. Thanks.
>>> >> >> My answer to the concluding question is pending, though I tend
>>> toward the
>>> >> >> latter proposition.
>>> >> >>
>>> >> >> On Tue, Nov 27, 2018 at 1:58 PM John Bailey <sundayjb at gmail.com>
>>> wrote:
>>> >> >>
>>> >> >> > Thanks Rich, great read.
>>> >> >> > On Wed, Nov 28, 2018 at 3:41 AM bulb <bulb at vheissu.net> wrote:
>>> >> >> > >
>>> >> >> > > Really excellent article, thank you Rich.  Working for a
>>> company that is
>>> >> >> > making massive investments in AI - this puts things in
>>> perspective..
>>> >> >> > >
>>> >> >> > > -----Original Message-----
>>> >> >> > > From: Pynchon-l <pynchon-l-bounces at waste.org> On Behalf Of
>>> rich
>>> >> >> > > Sent: dinsdag 27 november 2018 15:45
>>> >> >> > > To: “pynchon-l at waste.org“ <pynchon-l at waste.org>
>>> >> >> > > Subject: AI Thinks LIke a Corporation/Death of Insects
>>> >> >> > >
>>> >> >> > > thought you guys would be interested
>>> >> >> > >
>>> >> >> > >
>>> >> >> >
>>> https://www.economist.com/open-future/2018/11/26/ai-thinks-like-a-corporation-and-thats-worrying
>>> >> >> > >
>>> >> >> > > like everything else these days we're dazzled by the science
>>> not knowing
>>> >> >> > or caring about context, origins
>>> >> >> > >
>>> >> >> > > and this
>>> >> >> > >
>>> >> >> > >
>>> >> >> >
>>> https://www.nytimes.com/2018/11/27/magazine/insect-apocalypse.html?action=click&module=Top%20Stories&pgtype=Homepage
>>> >> >> > > --
>>> >> >> > > Pynchon-L: https://waste.org/mailman/listinfo/pynchon-l
>>> >> >> > >
>>> >> >> > > --
>>> >> >> > > Pynchon-L: https://waste.org/mailman/listinfo/pynchon-l
>>> >> >> > --
>>> >> >> > Pynchon-L: https://waste.org/mailman/listinfo/pynchon-l
>>> >> >> >
>>> >> >> --
>>> >> >> Pynchon-L: https://waste.org/mailman/listinfo/pynchon-l
>>> >> --
>>> >> Pynchon-L: https://waste.org/mailman/listinfo/pynchon-l
>>> >
>>> >
>>> >
>>> > --
>>> > Arthur
>>> >
>>> --
>>> Pynchon-L: https://waste.org/mailman/listinfo/pynchon-l
>>>
>>
>
> --
> Arthur
>
>


More information about the Pynchon-l mailing list