levy, pierre (2010) theoretical framework for a future computational collective intelligence

Upload: jotasmall

Post on 04-Jun-2018

216 views

Category:

Documents


0 download

TRANSCRIPT

  • 8/13/2019 LEVY, Pierre (2010) Theoretical Framework for a Future Computational Collective Intelligence

    1/6

    1

    Theoretical Framework for a FutureComputational Collective Intelligence

    Pierre Levy (01/03/10) em Pierre Levy Posteroushttp://pierrelevy.posterous.com/theoretical-framework-for-a-future-computatio

    1) Difference between Collective Computational Intelligence

    and Computational Collective Intelligence

    To avoid any misunderstanding I would like to begin with a clear

    distinction between what I mean by Collective Computational

    Intelligence and Computational Collective Intelligence.

    a) Collective computational intelligence involves collaboration

    between software agents, with a new level of computational

    intelligence emerging form their collaboration. These technologies

    involve swarm intelligence, ant colony simulation, web services, grid

    computing, distributed cloud computing and multi-agent computing in

    general.

    b) Computational collective intelligence is a more multidisciplinary

    field. Its subject is the understanding of human collective intelligence

    and its augmentation by the means of ubiquitous distributed

    automatic symbol manipulation. Even if computational collective

    intelligence involves the use and development of collective

    computational intelligence, its scope is broader because it is not

    concerned only by computer engineering but promotes a strong

    collaboration between computing on the one hand and humanities

    and social sciences on the other hand.

    2) Human Symbolic cognition

    The main difference between animal cognition and human cognition is

    that human beings are able to manipulate cultural symbolic systems

    like languages, musical systems, rituals, legal systems, technologies,

    etc. Animal cognition is obviously characterized by categorization and

    memorization of sense data. But animal memory is mainly limited tothe individual and its categorization stays implicit. By contrast, in

  • 8/13/2019 LEVY, Pierre (2010) Theoretical Framework for a Future Computational Collective Intelligence

    2/6

    2

    human cognition, sense data are explicitly categorizedby the use of

    symbolic systems and symbolic memory can be shared inside

    communities and between generations. As symbolic systems are

    created and maintained at the social scale, human collective

    intelligence involves a distributed and socially coordinated processing

    of symbols (recording, transformation and transmission). This

    personal and collective symbolic processing bears three

    interdependent main fruits: reflexivity, dialogue and narrativity.

    a) Reflexivity

    The main source of reflexivity is the syntactic complexity and

    semantic diversity of human languages. Reflexivity implies:- the ability to create and recreate symbolic maps of the

    phenomenal world (the sensori-motor experience),

    - the self-referential ability to map one's own symbolic

    cognition,

    - the ability to be aware of one's own ignorance, to question

    one's own mental models and to ask questionsin general.

    b) Dialogue

    Dialogue is based on the partaking of symbolic systems - and notably

    common languages - among human communities. Dialogue includes:

    - exchange of ideas and narratives,

    - negotiation about meaning and practical decisions,

    - common understanding of complex models,

    - common norms for reasoning and interpretation

    - representation of the other's cognitive processes.

    c) Narrativity

    Narrativity is based on the ability to model complex causal

    phenomena at the symbolic level. This involves the identification of

    actors, objects, qualities and processes and the evocation of their

    complex interaction through the virtual space-time of narratives.

    Through all its various forms (that are not limited to the traditional

    tales) narrativity is the main human tool for knowledge organization.

  • 8/13/2019 LEVY, Pierre (2010) Theoretical Framework for a Future Computational Collective Intelligence

    3/6

    3

    Based of symbolic manipulation, reflexivity, dialogue and narrativity

    make the natural human collective intelligence more powerful than

    the animal collective intelligence.

    3) The Digital Medium

    Information technology augments natural human collective

    intelligence by enhancing its ability to record, process and transmit

    symbols. Information technology gives rise to an artificial human

    collective intelligence. Artificial human collective intelligence began

    with the use of writing systems, it grew with the printing press and

    electronic media, it is now related to an universally distributed digital

    medium.

    The main idea behind computational collective intelligence is that the

    digital medium is integrating all previous media and that it can

    augment in an unprecedented way human collective intelligence by

    harnessing the power of ubiquitous digital data storage and automatic

    manipulation on these data. But as the digital medium is only one or

    two generations old, we still don't have the cultural symbolic systems

    that will help us to exploit fully the new availability of automaticsymbol manipulation. In particular, we know how to augment

    numerical calculus and logical reasoning by the way of automatic

    computation. But what about hyper-reflexivity of computational

    collective intelligence processes? What about trans-cultural and trans-

    linguistic dialogue on shared digital data? What about new kinds of

    hyper-narrativity harnessing the ubiquitous availability of multimedia

    data and computational power? We have already a digital medium

    but still no hypercortex.

    4) The IEML Semantic Sphere

    a) Addressing layers of thedigital medium1) The first layer assigns addresses to bits in the physical storage

    systems of automatic symbol manipulators. It is the operating

    systemsthat take care of this addressing layer. This layer has been

    set-up in the middle of the XXe century.

  • 8/13/2019 LEVY, Pierre (2010) Theoretical Framework for a Future Computational Collective Intelligence

    4/6

    4

    2) The second layer assigns addresses to the automatic symbol

    manipulators in theglobal digital telecommunication network. It is of

    course the Internet protocolthat takes care of this addressing layer.

    Even it has been invented at the end of the 1960 decade, the

    Internet emerged as an universal addressing system only during the

    years 1980, at the time of the personal computing revolution.

    3) The third layer, the WWW, assigns precise universal addresses to

    Pages (URLs), allowing the hyperlinking from any URL to any other.

    The WWW has been widely adopted from the middle of the nineties of

    the last century. A powerful generalization of the WWW is theWeb of

    data. Through the data formats XML, RDF and OWL, the Web of data

    insures a better interoperability between data addressed by URLs.The Web of data (particularly regarding RDF and OWL) is still in a

    phase of implementation.

    4) I'm pleading here for the adoption of a fourth layer of addressing:

    the addressing ofconcepts. This new addressing layer would be based

    on IEML (Information Economy Meta Language), an artificial symbolic

    system that I have designed and that is theoretically able to express

    any kind of meaning in a formal computable way. The main differencebetween the layer of data - addressed by URLs - and the layer of

    semantic metadata - addressed by IEML USLs (Uniform Semantic

    Locators) - is that URLs are opaque by design and that USLs are

    transparent by design.

    b) Transparency by design

    All the nodes (USLs) of the IEML semantic sphere are variables of the

    same semantic group of transformation and the computable circuits

    between these nodes describe paradigmatic relations (set-subset

    relations, etymologic relations, several kinds of symmetric relations,

    serial relations) and syntagmatic relations (relations between the

    elements of hyper-propositions and hyper-narratives).

    The IEML semantic sphere presents itself as an immense closed

    structure of intersecting syntagmatic and paradigmatic pipelines that

    branch out from nodes (IEML USLs) of which each is a separate

  • 8/13/2019 LEVY, Pierre (2010) Theoretical Framework for a Future Computational Collective Intelligence

    5/6

    5

    variable in a system of symmetrical and calculable transformations.

    One can think of the syntax of IEML as a virtual machine capable of

    computing the vast network of fractal complexity of the semantic

    sphere. This syntactic machine needs to be provided with a

    dictionary. The dictionary establishes a correspondence between IEML

    and natural languages and sets the details of internal connections of

    the semantic sphere. Each point, junction or node of the IEML

    semantic sphere is at the center of a multitude of pathways of

    calculable transformations. Along these pathways of transformation,

    each "step" from one intersection to another is the variable of a

    discrete function. Step by step and from one to the next, these paths

    connect each point to the entire mass of all the other points. In the

    centrifugal direction, an intersection point is at the singular origin of astar of transformation that generates the entire sphere. In the

    centripetal direction, a junction-point functions as a universal

    vanishing point of the noosphere, since there is a computable path of

    transformation that leads to it from any other point.

    c) A model of computational collective intelligence

    I propose a model of computational collective intelligence in which anonline creative conversation actualizes the junction of (a) the

    symbolic cognition based on the natural cortex and (b) the

    computational symbolic cognition based on a mechanical

    hypercortext. The structure of the natural symbolic cognition has

    already been sketched. The computational symbolic cognition has

    hopefully the same structure as the natural cognition.

    The Web of data is the equivalent of the sense data of personal

    natural cognition. It is of course multimedia. It is also opaque (or

    implicit) because its addresses are not the variables of a transparent

    group of transformations.

    The IEML semantic sphere is the equivalent of the discursive explicit

    symbolic layer in the natural cognition based on the biological cortex.

    But as it is based on the digital medium, it has been designed from

    the beginning to exploit fully its distributed computing power and to

    augment human reflexivity, dialogue and narrativity at the scale of

  • 8/13/2019 LEVY, Pierre (2010) Theoretical Framework for a Future Computational Collective Intelligence

    6/6

    6

    the current online creative conversation. It can be used to describe,

    map and organize the digital multimedia memory from any semantic

    perspective or universe of discourse. It is also able to make these

    different perspectives and universe of discourse automatically inter-

    operable. Moreover, its circuit structure allows for the design of

    varied information economy games organizing the circulation of

    symbolic energy. These information economy games will take as their

    imput the activity of players tagging data and assessing their value.

    Computational collective intelligence implies the categorization of

    multimedia data by semantic metadata and the multimedia

    representation of semantic information economy.

    My guess is that we would be able to build a hypercortex augmenting

    decisively our human collective intelligence only if we grow a

    transparent hypertextual sphere of semantic metadata, the nodes of

    its graph being the variables of the same group of transformation.

    Needless to say, the growth of an hypercortex enabled by a semantic

    sphere of transparent metadata would be a huge cultural and

    technical international endeavor. IEML is only the proof by a real

    example that the success of such an endeavor is possible.