Innovative future-forecasting project goes open-source
A blog-post draft sent to KurzweilAI.net editor Amara D. Angelica announcing the open-sourcing of The Uncertain Future , a Singularity Institute project on which Alyssa was research director. Co-authors named below: Anna Salamon (project manager), Michael Anissimov (writing director), Steve Rayhawk (math), and Rolf Nelson (programming).
The Uncertain Future , the world’s first web-based application for making rigorous, scientific forecasts of technological progress, is going open-source today. The Uncertain Future, a project sponsored by the Singularity Institute , was started in early 2008, with the goal of allowing anyone interested in futurism to form their own, mathematically consistent model of what the future of technology and human civilization holds.
The Uncertain Future web application focuses on the development of technology — including artificial intelligence, genetics, and biotechnology — as the key element in predicting the future. Although the more conventional political, social, and economic factors are important, we, the developers of Uncertain Future, believe that the past few centuries have shown fairly conclusively that technological growth and change is now the primary driver of our civilization.
The Uncertain Future web application takes into account a synthesis of different factors that might impact the future, such as the speed of scientific progress, the difficulty of building technology, and the possibility of catastrophe. Psychological research has shown that our naive intuitions about the future are often inconsistent. For instance, someone might say that, over the next hundred years, there is a 10% chance of humanity going extinct, and a 15% chance of humanity going extinct because of nuclear war (for more on this, see Cognitive Biases Potentially Affecting Judgment of Global Risks ). By bringing together a person’s ideas about what will shape the future, the Uncertain Future web application can help resolve such inconsistencies.
It employs a formal mathematical model, using the continuous-time Bayesian network framework. This allows for precise computation of what a prediction is, and how strongly the user thinks it will come true. It is common knowledge that you can always find some way to make a vague prediction (like a horoscope or fortune cookie gives) seem like it was accurate after the fact. Using a mathematical model greatly reduces such ambiguity.
The Uncertain Future also uses probability distributions, rather than specific scenarios. In conventional, “story-telling” futurism, one might predict event 1, and then predict how event 1 will lead to event 2, and so on, up through event 100. The trouble with this method is, if any one thing in the chain of predictions breaks (for instance, if event 37 does not lead to event 38), the entire sequence of predictions will be wrong. By using probability distributions, we avoid this problem. For example, if event 11 has an 80% chance of happening, we forecast what would come next if event 11 happened, and what would come next if event 11 didn’t happen.
The data for these predictions come from a reference section, which includes a wide range of expert opinions. This is especially useful, because the Uncertain Future is a forecasting tool, rather than an attempt to convince the user of any one specific scenario. A conventional essay about the dangers of cloning might include references that discuss why cloning is a serious possibility, but since the point of the essay is to convince the reader, it might include no references that point out the remaining obstacles in cloning technology. The Uncertain Future encourages the reader to look at expert opinions, both optimistic and pessimistic, and make up their own minds, without cherry-picking.
The Uncertain Future was developed during 2008 and 2009 by a team of researchers including Anna Salamon (project manager), Michael Anissimov (writing director), Steve Rayhawk (mathematics director), Rolf Nelson (programming director), and Alyssa Vance (research director), with the funding and support of the Singularity Institute and Singularity Institute president Michael Vassar .
The Uncertain Future was first presented at the 7th Annual European Conference on Computing and Philosophy , held in Barcelona, Spain from July 2 to July 4, 2009. The presentation, titled “Changing the frame of AI futurism: From storytelling to heavy-tailed, high-dimensional probability distributions,” outlined previous research in forecasting technology, discussed the need for a new kind of prediction software, and included a live demo. The software was made available to the public on December 12, 2009; the original announcement is on the Singularity Institute blog .
As part of our next phase in the project, starting today, all Uncertain Future source code will be freely available on GitHub , under the terms of the GNU General Public License (GPL). The GPL allows anyone to freely download, modify, and distribute the source code. We hope that the Uncertain Future code will form the basis for a wide range of other projects, both inside and outside of technology forecasting.