Paul Saffo from Stanford University is a forecaster and essayist. He explored AGI at the Singularity Summit 2007 in the context of popularization. Even as the public begins to join the AGI discussion, pessimism is popular right now. News reporting and commentary are lurid and pessimistic. Saffo suggested that what we need are positive and compelling visions of the future and AGI in popular fiction, but not by scientists turned writers. Poets, authors, artists, performance artists, and other creative individuals need to begin exploring the concept of AGI separately, just as previous technological breakthroughs have been explored in popular art.

In 1967 the poet Richard Brautigan self-published a book of poetry that included a positive vision of AGI. Entitled “All Watched Over by Machines of Loving Grace”, the poem is available here, and reprinted below:

I like to think (and
the sooner the better!)
of a cybernetic meadow
where mammals and computers
live together in mutually
programming harmony
like pure water
touching clear sky. 

I like to think
      (right now, please!)
of a cybernetic forest
filled with pines and electronics
where deer stroll peacefully
past computers
as if they were flowers
with spinning blossoms. 

I like to think
      (it has to be!)
of a cybernetic ecology
where we are free of our labors
and joined back to nature,
returned to our mammal
brothers and sisters,
and all watched over
by machines of loving grace.

Dr. Ben Goertzel is SIAI Director of Research and chief science officer and acting CEO of Novamente. He believes that with the right funding and capital, AGI could be developed in nine years. Novamente is approaching AGI development by focusing on virtual agents in virtual worlds like Second Life.

Goertzel defined AGI as “the ability to achieve complex goals in complex environments using limited computational resources.” AGI will require autonomy, practical understanding of self and others, and understanding “what the problem is” opposed to just rote problem solving.

Unlike some other speakers and AGI proponents, Goertzel does not believe that today’s narrow AI solutions necessarily suggest the right paths to AGI. He provided a list of items that will be necessary for developing AGI:

• AGI will not be created by generalizing narrow AI.
• Researchers need to go back to the basics; that is, by creating artificial babies and pets and allowing the technology to develop into more advanced entities.
• AI can be created in robots or in virtual reality. Robots are expensive and have high maintenance requirements and costs. Virtual reality allows for embodied agents will lower costs.
• Researchers should harness the wisdom of the crowds. Instead of only company-paid teachers, AI could learn from everyone who interact with them via virtual worlds. The populations of many of today’s virtual worlds suggests thousands to millions of users who can help.
• The particular model for minds used by Novamente was developed by Goertzel and described in his book The Hidden Pattern. Minds are patterns with pattern-recognition capabilities, including their own pattern and more complicated patterns of patterns.
• Goertzel believes a cognition engine based on the patternist philosophy will lead to the creation of AGI.
• Open-source AGI frameworks may prove to be useful and even necessary toward the development of true AGI.

Goertzel mentioned that there remain particular ethical issues. The creation of AGI presents dangers close at home, including to our loved ones and self. Determining how to create Friendly AI, more intelligent than us but with our best interests and values at heart, may be difficult.

Neil Jacobstein, chairman and CEO of Teknowledge Corporation, spoke in the afternoon of day one of the Singularity Summit 2007. He presented information about where we stand today with AI and where we might be heading in the medium term. Specialized AI applications have existed for a number of years and often outperform humans. Jacobstein did not think there was one true path to AGI.

The value in AI today consists of augmenting and replacing human skills. AI is often delivered with other components in a technology package, usually with a browser and/or Web 2.0 standards included. A study of AI solutions presented to date suggest a few things that have not worked so well with AI, including high maintenance requirements and costs, poor integration with mainstream applications (as opposed to the specialized nature of most AI packages today), and limited learning capability (most information must be entered by human operators.)

Some of the things that have worked with AI so far including real world testing and experimentation (these packages have moved out of the lab and into the real world), the ability to solve specific problems, and integration with statistics, optimization, and simulation solutions.

Today the web is undergoing evolution into Semantic Web Services, where common standards and applications provide an interface into AI. With this rise of the Semantic Web, Jacobstein sees the introduction of AI web agents and a shift from human-to-machine interactions to primarily machine-to-machine interactions.

Ray Kurzweil is an inventor and futurist, perhaps best known now for his books and lectures regarding accelerating change and the Technological Singularity. He joined the Singularity Summit 2007 via satellite. Because his work was referenced often in other speakers’ talks, he addressed some of their comments and conclusions, including a defense of double exponential growth in information technologies, rapid advances in algorithms (and therefore software), and a convergence of different technologies.

Kurzweil kept his comments short and then opened up the rest of his time to questions from the audience.

Yudkowsky returned to the stage to discuss the challenge of Friendly AI. The problem, to Yudkowsky, is increasingly difficult because it is difficult to pick out, of all mind possibilities, the one that we would consider friendly. In some sense, we would like to develop AI that can create expert AI, and an AI-Theory AI to create AI, and a Friendly-AI-Theory Friendly AI to create Friendly AI.

One approach to protect against unfriendly AI is to create a moral trajectory for which Friendly AI would aim, rather than hard-coding a particular morality into Friendly AI. He used an example of ancient Greeks, if they had had the ability to create AGI. Their AGI would not necessary be considered friendly by today’s morals, and our own developed AGI might not be truly friendly if based on today’s morals. Therefore, Friendly AGI should be able to seek further moral development, to avoid the constraints otherwise placed by our existing morals.

James Hughes is the the executive director of the Institute for Ethics and Emerging Technologies and host of Changesurfer Radio. He discussed preparation for the future and how futurists can have an impact on public policy.

Hughes listed a series of assumptions he makes about AGI and the Technological Singularity, such as AGI is likely, but probably dangerous and radically alien. While he does not adhere to either a utopian or dystopian view of the Singularity, Hughes did suggest that there are real risks involved, and that our assumptions about AGI may need to be analysized in some detail.

For example, do we really want a perfect utilitarian AGI, or a perfectly truthful AGI? These AGI might not turn out in ways that are beneficial to humans.

While AGI is likely, Hughes suggested that we might also want to augment human intelligence in ways that attempt to allow us to keep up with AGI. In this way we might be able to keep up with technological advance including AGI so that we can deal with any potential problems and consequences.

Other considerations Hughes highlighted included robot/AI rights and the possibility that we will need to regulate and license these superintelligences to better control their use.

Christine Peterson is Vice President of Foresight Nanotech Institute and she talked about how we can prepare for the advent of AGI (what she called “preparing for bizarreness”), with a focus on upcoming risks.

According to Peterson, risks, and their potential solutions, can be approached in a top-down monolithic way, or a bottom-up way, in which solutions are decentralized, voluntary, experimental, collaborative, open, and transparent.

Peterson suggested automated or semiautomated systems designed to protect us without threatening us at the same time are an appropriate long-term security goal. Perhaps the group best qualified to debate these concerns are the open source programming community.

Steve Jurvetson is a Managing Director for the venture capital firm Draper Fisher Jurvetson. He explored two possible paths to AGI, while emphasizing an evolutionary, bottom-up approach versus designing AGI directly, based on the experience of DFJ investigating and investing in nanotechnology.

Evolved systems could result in robust, resilient, and adaptive technologies, such as AGI. It will not be necessary to understand every thing that is going on in these systems. Combined with exponential progress, the resulting systems may outstrip our ability to model and develop a theory of how the system or its subsystems works.

Michael Lindsay from the X Prize Foundation announced the possible development of an upcoming series of education X Prizes. After a video about past and current X Prizes, Lindsay provided new details about their latest incentive competitions.

Education is a challenge, especially in the United States. The public perceives the school system as broken. The system is antiquated and tailored toward groups of children rather than an individual student.

X Prize will focus on educational software improvements in upcoming education prizes. Competitors will creates adaptive educational software with near real-time student assessment and diagnostic capabilities with interactive game-like interfaces. The subjects to be covered first are Algebra 1, Basic Literacy, and Second Language Acquisition.

A learning achievement goals of two sigma (average student achievement greater than 98% of classroom students) was suggested.

Can software improve learning? A recent study suggests current software solutions do not, but the X Prize Foundation would like to show that breakthrough software could do so in the near future. Some possibilities include networked gaming settings, intelligent software, peer-to-peer, and teacher-assisted.

While no specific prizes and amounts were announced, Lindsay asked for audience comments regarding development of future prizes. The responses were, perhaps expectedly, opinionated and passionate. When asked by Lindsay if the problem of education might be solved within ten years, a minority of audience members raised their hands.

Dr. Charles L. Harper, Jr. is Senior Vice President for the John Templeton Foundation. He provided three big questions people might want to think about regarding AGI and the Technological Singularity:

• Big Question #1: What do slugs know of Mozart?
• Big Question #2: How serious is the “dilemma of power”?
• Big Question #3: How important is the”transformation of desire”?