Main Page

   Note for workshop participants: WorkshopMinutes

The AGI roadmap is potentially a very powerful tool in the effort to focus the AGI research community; it aims to play a vital role in lending credibility to this emerging field of formal science, providing a starting point and central knowledge repository for new entrants to the field, and increasing available AGI research funding by way of demonstrating that research in this area is both active and viable.

   However, we need you! Please create an account and join the effort to organize and advance this field and make progress toward the goal of AGI!

 This site is changing very quickly, so be sure to check back regularly for updates.  

Contents 1 Welcome 2 A Call to Arms 3 What are the Goals of the Roadmapping Effort? 4 Editing and Contributing to the Roadmapping Effort 5 A Roadmap Toward Human-Level AI Systems Embedded in Broad, Approachable, Physical or Virtual Contexts 6 General AGI Roadmapping Related Content 7 Guilty Parties

Welcome

The ambition of this website is to provide a forum for discussion and construction of a Roadmap toward Artificial General Intelligence. Its creation is inspired in part by Itamar Arel and Scott Livingston's IEEE Computer column in the March 2009 issue. The closing paragraph of the draft is given below.

This is a determined effort to organize and focus the efforts of many independent researchers; we can't do this without you! Create an account and join the discussion!

Please note that this website is quite new and undergoing rapid development. Please check back regularly!

A Call to Arms

As we approach the twilight of the first decade in the twenty-first century, one may feel inclined to settle with the great achievements of man that shine forth in all aspects of modern life: the stunning marvel of the Internet, the subcultures that rise and fall in it; the pervasive little white earbuds seen in many a pedestrian; and the rapidly accelerating mobile technologies that are changing the way we interact. We have passed the year 2001 without the formidable Hal, but not for lack of human potential. By designing one or more roadmaps to artificial general intelligence and creating important benchmark problems that define steps on the path to this "grand dream," the original goal of AI may yet be achieved.

Will the common goal of AGI be achieved? The answer depends at least in substantial part upon whether a quality roadmap is formed.

What are the Goals of the Roadmapping Effort?

The goal that we are working toward is defining and refining goals that AGI researchers can work towards. Very broadly speaking, we hope to make progress in answering the following questions:

• Defining the problem
  • Bring focus and coherence to the efforts of the AGI community
  • Promotes consistent terminology across researchers and practitioners in the field
  • Clarify the differentiating elements of AGI with respect to "narrow" AI
  • What type of behavior ought one to expect from an AGI implementation?
  • What attributes and characteristics are required of an AGI implementation?
• Testing potential solutions
  • Facilitate progress of performance metrics and evaluation methodologies
  • How can one test an AGI implementation to verify that it does or does not meet the requirements for AGI?
  • By what method or measure can one AGI implementation be said to be 'better' than another?
• Identifying concerns that must be addressed
  • What ethical considerations must be taken in the development and pursuit of AGI?
  • What social impact would a functioning AGI implementation have, and how ought that be handled?
• Fostering discussion of the work ahead
  • Communicate what the AGI community is all about to both the public and the community itself
  • What questions ought to be included in this list?
  • What are the goals or milestones that researchers ought to come together and pursue?

Generally, we seek to promote scientific progress and discovery as it relates to AGI.

Editing and Contributing to the Roadmapping Effort

We need contributions from you. However, it is extremely important to remember that in a young field like AGI there is a need to carefully manage the conflict that is bound to occur within the community. We acknowledge that there are many issues being discussed on this page that by no means represent a unified opinion from the community. In light of that, it is important to remember our goals while editing or contributing to a page: feel free to add to the content of a page, but if you feel that something needs to be changed or removed, this must be discussed on the talk page (accessed via the discussion link at the top of every page). We do not want to have editor wars where multiple people change pages back and forth in an attempt to establish factual content of a page.

The goal of this wiki is to foster the creation of one or more viable roadmaps leading to powerful AGI. We are not attempting to be the unique source and we strive to acknowledge and inform about all of the approaches and schools of thought in the field.

A Roadmap Toward Human-Level AI Systems Embedded in Broad, Approachable, Physical or Virtual Contexts

See [1] for a PDF document laying out a concrete plan for creating an AGI roadmap based on a specific set of assumptions about the path to be taken to AGI.

This is not intended as the only plausible AGI roadmap; it is intended as a plan for creating one among the many roadmaps to AGI that are likely possible.

This section is intended to grow into an AGI roadmap based on the above document. Content that questions of the assumptions of this document, in most cases, should go somewhere else on this wiki. The point of the material in this section is to flesh out a roadmap based on certain assumptions.

This page gathers a bunch of Useful resources for defining tasks to evaluate developing AGI systems

The detailed roadmap is intended to grow within the following pages:

• Identifying the end product that will constitute the focus of the roadmap
• Identifying the environments and tools that will be used to enable the following of the roadmap
• Milestones | Specifying major incremental capabilities or milestones that build toward the end product
• Specifying methods of evaluating and measuring these incremental capabilities
• Identifying technology alternatives and their projected timelines

General AGI Roadmapping Related Content

This section gathers general discussion on issues related to the creation of AGI roadmaps.

This site is still rapidly evolving, so much of this is subject to change. Have a suggestion about a new section or an alternative organization of the existing material? Let us know on the talk page.

Currently, here are the main areas of the wiki:

• The Big Picture: items that affect all AGI researchers, regardless of methodology.
  • Defining the field: Properties that define intelligence and AGI, formal requirements and AGI must meet (if in fact any can be defined), milestones to be met, the general course of the field as a formal science.
  • Legitimizing the field: Increasing the perceived legitimacy of AGI, making progress toward increasing the research funding available to AGI, steps that must be taken before the general population accepts AGI as viable.
  • Challenges facing the field: What major issues must be overcome before AGI can be considered a formal science?
• AGI Research Methodologies and Issues
  • Research Methodologies: Possible approaches used for AGI implementation, cognitive architectures, language interpreting and symbolic representation, model-free approaches.
  • Testing, training, and evaluation: Training and testing environments or tasks for AGI, metrics used to objectively quantitatively measure performance and capability of AGI (if indeed such a thing can exist).
• AGI Moral and Social Issues
  • Ethical Considerations: What must an AGI do in order to be qualified a 'friendly AGI?' Ought we create AGI at all?
  • Social Impact: What must the world do to be ready for AGI? What must AGI researchers do to increase social acceptance of and desire for AGI? How would a successful AGI implementation change the social and political landscape?

Guilty Parties

This website is currently being maintained and moderated by Bobby Coop; it was originally created by Scott Livingston and Itamar Arel in February 2009. The intention, however, is for new contributors to join and help make this a true convergence point for discussion and efforts toward creating an AGI roadmap.