FUTURE

EVENT

Frozen Organ Transplant Is Routine

Two Century Biological Lifespans

Indefinite Biological Lifespans

Reanimation for Last Cryonics Suspendee

Reanimation for Current Cryonics Suspendees

Biotech Cures for Most Heart Disease, Cancer & Aging

Fine-Tuned Mood/Motivation Transformation Drugs

Genius Drugs (>20 pts permanent IQ increase for most people)

Human Germ-Line Gene Therapy

Human Child Gestated Completely in Artificial Womb

Cloning of a Human Being

Completely Genetically Composed Children

Extinct Species Reanimation (from preserved DNA)

Cryonics Industry Revenues $1 billion/year

Nanotech Factories

Atomically Detailed Design for Self-Reproducing Drexler-style Assembler

High-Degree of Freedom Cell Repair Nanomachines

Reproducing Nanotech Assemblers

Really Cheap Fusion Power

Nukes as Cheap as Tanks

Nukes as Cheap as Handguns

Most Publications are Electronic

Most Intellectual Publications are on Web

Information Storage $0.01 per Megabyte

Computer Implanted in Brain

Human-Brain Equivalent Computers on a Desk

Human-Level A.I.

Benford

2020

2150

2300

2100

2200

2030

2010

2030

2040

2020

2050

2060

2100

2035

2100

2070

2075

2080

2100

2105

never

2015

2001

2010

2015

2030

2030

Bridge

2010

2050/2140

2080

2060

2090

2030

2020

2020

2007/2025

2050

2020

2050

2025

2015

2030/2050

2015

2040/2060

2025

2040

2015

never

2015

2008

2020

2045

2030

2050

Drexler

never

never

1967

2006-2021

2006-2021

never

?-2021

?-2021

?-2021

2006-2021

1998-2010

2006/2021

2004-2019

2004-2019

2004-2019

EXTROPY #15 (7:2) 2nd-3rd Quarter 1995

10

FORECASTS

C O M M E N T S

NICK SZABO

The first number is when something might be possible under ideal engineering, economic, and political conditions. “now” means we could have done it already. The second number is the practical prediction, based primarily on the viewpoint of starting a business with engineering and political issues secondary. Big gaps between the two numbers usually indicate major economic or cultural barriers to adaptation of the technology. Many of the political developments (law choice, etc.) will be available much sooner for those who sacrifice other aspects of their lifestyle to pioneer these new jurisdictions.

STEVE BRIDGE

My answers are based on when something will “actually happen” rather than on when it will be possible. Some of the reasons why something could be possible at a particular time but not actually accomplished are economic, some are related to social objections, and some are simply because the top minds in these fields can’t pursue all avenues of research at once. What we are trying to do here is not only to predict technological development, but to predict what people will choose to work on first.

I am equally convinced that the time line for “actually happens” is not well connected to the time line for “most people do it.” Computers have existed for 50 years or so; home computers have been easily available for at least 15 years. Yet we are still not to a point where “most” people even own a home computer, much less use one for anything more than games. Most people own a television because it is passive. Most people are not interested in actually “doing” things. We tend to forget that, because we spend almost all of our work and social time with the people who are steeped in thinking and doing. The average American and the average human are not like those around us.

ERIC DREXLER

The rate of technological advance depends on the quality of tools, both material and computational. Both are getting better faster as better tools are used to build better tools. The result of this cannot be a singularity in the strict mathematical sense, but Vernor’s term still seems remarkably appropriate. For advances that we can already describe, a steady trickle of progress over the coming decades and centuries seems most unlikely.

Actual developments will depend not only on what is possible (a matter of physics) but on races between different technologies (a matter of complex competitive processes in an unfolding history). The

FM-2030	Miller	More	Szabo
1990s+	N+20 if (S>N+30)	1999-2008	2020-2030
2010-2020	never	2015-2040	2040/2100
	S-50	2020-2045	2090/2150
	S+6*DAF	2025-2055	2050/2200
2020	S+10*DAF	2030-2100	2400/2410
1990s+	S+10*DAF	2015-2040	2090/2130
1990s+	N+10	1998-2010	2040/2050
	S-10 to S+10*DAF	2020-2060	2010/2050
1990s	N+20	2010	now/2020
2010-2020	S-5 to S-2*DAF	2015-2035	2100/2120
2010	S-5 to S+4*DAF	2010	now/2010
		2015-2020	2060/2100
	N+5 to S+1		2010/2020
2010-2020	N+30 (if S>N+30)	2015-2020	now/2020
2010-2020	S-3 to S+1	2015-2030	2070/2080
	N+7	2000-2015	2100/2100
2010	S+2*DAF		2160/2180
	S-3 to S+1	2020-2030	2120/2140
2010-2020		2010-2020	2200/2210
		2040-2050	2100/2150
			2200/2250
1990s+	N+10 to N+30	1999	2000/2005
late 1990s	N+5 to N+30	1999-2002	2000/2005
	N+1 to N+10	2015	2010/2010
2010	N+1 to N+10	2020-2050	2010/2020
2010	S-3*DAF	2030	2040/2050
2010		2040-2150	2150/2200

11

EXTROPY #15 (7:2) 2nd-3rd Quarter 1995

EVENT	Benford	Bridge	Drexler
Uploaded Minds	2060	2125	2006-2021
Uploads Running 1000x Faster than Humans	2080	2125	2006-2021
Big Fraction of Economy Off Earth	2200	2100	2006-2021
Big Fraction of Economy out of Solar System	2800	3000	2011-2026
Comet Mining, Javelins, Drugs, etc. (robotic space industry)	2080	2075	2006-2021
First Person on Mars	2050	2025	2006-2021
First Person in Another Solar System	2400	2085	2011-2026
Reproducing Comet Eaters		2070
Reproducing Asteroid Eaters	2150	2045	2006-2021
Reproducing Starships	2300	2200	2006-2021
1,000,000+ People Using Anon. Electronic Cash	2010	2020
30%+ of Labor Telecommutes	2015	2030
Untaxable Economy Using Electronic Cash $100b/year	2020
Ocean Colonization	2020	2020/2045
Most Education Privatized	2005	2050
Most Law Enforcement Privatized	2010	2095
Most Law Choice Privatized	2020	never on Earth
National Defense Privatized	never	never on Earth
Betting Markets a Big Policy Influence	never	never

dates that follow represent a single set of basically similar scenarios, in which advances in computation and molecular machinery support one another strongly and win certain technology races. The key assumption is that molecular manufacturing is not vastly more difficult to develop than it now seems to the handful of people doing atomically-detailed design and simulation of molecular machines.

A reliable schedule for technological advance is impossible to produce, but these dates wouldn’t greatly surprise me.

MARK MILLER

The time of onset of many of these events are related to each other. Phrasing the predictions in terms of absolute time makes predicting even more hazardous than it need be. If I think that A will happen between 10 and 40 years from now, but that B will follow A by between one and two years, I communicate much more to the reader by saying that than to predict that B will happen between 11 and 42 years from now. Think of AB as a fairly rigid structure hanging at the end of a long stretchy rope. The relative positions of the elements of the structure carries information. However, phrasing all predictions in terms of offset probability distributions from all the other predictions would drive both reader and writer crazy. Accordingly, I will use the

following variables:

N = Now, 1995 S = Singularity DAF = Design Ahead Factor

I could define what event constitutes Singularity, such as general molecular assembly capability. However, in this exercise such a definition is unnecessary. By predicting other events as clustered in time around Singularity, Sigularity effectively becomes defined as a kind of center of gravity of the onset of these other events. I consider the distance between here and Singularity to be the most uncertain distance in the mix. I predict Sigularity as occurring between N+10 and N+40.

Finally, how sudden Singularity is, i.e., how closely clustered the various events are, depends on how much design ahead has occurred in anticipation of Singularity. The sooner Singularity occurs, the less design ahead will have anticipated it, and the more spread out it will be. To account for this, I introduce a Design Ahead Factor which I define as 10/(S-N). If Singularity occurs in 10 years, DAF is 1. If Singularity occurs in 40 years, DAF is 1/4.

One can always make yet more complex models, but I fear that, as Karl Popper would put it, my precision already vastly exceeds my accuracy. Well, you’re better off if you try than if you don’t.

EXTROPY #15 (7:2) 2nd-3rd Quarter 1995

12

FM-2030	Miller	More	Szabo
	S+7*DAF	2040-2100	2300/2400
	S-3*DAF	2045-2100	2450/2450
	S+20*DAF+20	2100-2200	2150/2200
	S+20*DAF+(50-200)	3000	2400/2500
			2040/2060
		2050
2010-2020	N+15 to S+2*DAF	2025	2040/2060
2030-2050	S+10*DAF+20	2150-2400	2200/2400
		2050-2070	2140/2180
	S+2*DAF		2350/2400
1990s	N+10 to N+30 to never	1999-2006	1997/1999
1990s	never	never	2000/2050
	N+20 to never	2010-2115	1997/2005
	never	2010-2050	now/2040
	N+10 to S+50		now/2040
	S+50 to never		now/2150
	S+20*DAF+20 to never		now/2150
	never		now/2200
	S+20*DAF+20		2000/2100

I assume “Human-level AI” means without uploading. This may not in practice be a clean distinction at the time, just as there is no longer a clean distinction between synthesized images and texture mapped sampled images. The practical way really impressive “synthesized” images are usually created is to also mix in (via texture mapping) lots of images sampled from the world. Similarly, by the time we are synthesizing human-level AI for practical purposes, we will probably be mixing in uploaded components of evolved intelligence.

“Uploaded Minds” and “Uploads Running 1000x Faster than Humans” will happen about the same time: The hard part will be getting a good upload. At the level of technology at which that will be achieved, the extra factor of 1000 will not be a computer power issue. There will, however, be user-interface issues in both directions. How does a sped up upload interact with a world working 1000 times slower (subjectively) than he’s used to, and how do non-uploads interact with a 1000 times faster person?

For the first problem, part of the answer is what I call “bodies as user-interface metaphor”. Our cognition is the result of long evolving to control a body which in turn effects the world. Computer mice work by keying into notions of pointing, grasping, and carrying. Uploads will have no need for physical bodies, but to make good use of their evolved minds to affect the world, they will largely bring about these effects by controlling a simulated and somewhat abstracted body, in a somewhat more symbolic world.

THE FORECASTERS

Gregory Benford: Benford is a physics professor at University of California, Irvine, where he conducts research in plasma turbulence and in astrophysics. He is a Woodrow Wilson Fellow and a Visiting Fellow at Cambridge University, and has served as an advisor to the Department of Energy, NASA, and the White House Council on Space Policy. Benford is author of over a dozen novels, including Jupiter Project, Against Infinity, and Timescape. A two-time winner of the Nebula Award, Benford has also won the John W. Campbell Award, the Australian Ditmar Award, and the United Nations Medal in Literature. gbenford@uci.edu

Stephen Bridge is the President of the Alcor Life Extension Foundation, the world’s largest cryonics organization. He is 46 years old. Steve graduated from DePauw University in 1970 with a B.A. in Theater and from Indiana University in 1974 with a Master’s Degree in Library Science. Steve was a public librarian from 1974-1992 and has been involved in cryonics since 1977. He co-founded Cryonics Magazine in 1981. He has long had an interest in science, science fiction, life extension, and the future. As a former librarian, he knows a little bit about everything. steve@alcor.org

Eric Drexler: Eric Drexler extracted a Ph.D. in Molecular Nanotechnology from MIT. He wrote Engines of Creation (Doubleday), co-authored Unbounding the Future (Morrow), and Nanosystems: Molecular Machinery, Manufacturing, and Computation (Wiley Interscience) — named the outstanding computer science book of 1992 by the Association of American Publishers. He began studies of molecular nanotechnology in 1977. drexler@netcom.com

FM-2030: See Profile in this issue.

Mark Miller: At Datapoint Mark built the first commercial distributed windows system. He was co-architect of the Xanadu distributed hypermedia server. At Xerox PARC, Mark co-authored (with Eric Drexler) the agoric open-systems papers on market-based computation and market-oriented programming. Currently he is co-director of the Agorics Project at GMU, Chief Technical Officer of Agoric Enterprises, Inc. in Fairfax, VA, and a founder of Agorics, Inc. in Los Altos, CA. mmiller@netcom.com

Max More: See Contributors on p.59.

Nick Szabo has worked at JPL scheduling communications on the Deep Space Network, and at IBM and Sequent on operating systems software. He currently resides in the Netherlands, working at Digicash bv on privity technologies. szabo@netcom.com

13

EXTROPY #15 (7:2) 2nd-3rd Quarter 1995