Society & Environment

The Future of Human Nature: A Symposium on the Promises and Challenges of the Revolutions in Genomics and Computer Science (Conference): Session Two

 

Series: Pardee Center Conference Series
Dates: April 10, 11, and 12, 2003
Location: Frederick S. Pardee Center for the Study of the Longer-Range Future, Boston University, Boston, MA

Session Two

Marvin Minsky

What’s Wrong with People?

I will start by talking about what is wrong. If you make a list of serious problems, you see that most of them are due to there being too many people. That includes problems of garbage disposal, disease epidemics, the depletion of natural resources, the disappearance of biodiversity, and the distribution of wealth. One solution would be to reduce the size of people from six feet to six inches. That way you could get a trillion people on the planet with less pollution.

Or we could reduce the size of neurons. Neurons might be filled with stuff they do not need, and you could simplify them. Or if you could reproduce people’s minds on computers, you could probably store the total amount of human memory on a CD. There’s no evidence that people have more than 100 megabytes of knowledge. Nobody knows how to figure it out, but we have 50 trillion synapses. In any case, although neuroscience is doubling every few months, we still do not know simple things, like how memory works in any high-level sense.

There are also alternatives to traditional ways of having children and living in families. When we understand what the genome contains, we could make forty-six people. A fairly simply form of genetic surgery would be to decide which are most important to you. Instead of implanting a whole nucleus, you ought to be able to implant carefully selected chromosomes, so that forty-six people altogether could have fifteen or twenty children. Then everybody would have lots of relatives and live in a big family, and each person would reproduce for every two in previous generations.

My major point is that although there are many serious problems, we are not smart enough to solve them. Right now it takes about 100 years to learn biology. People do not live long enough to understand and solve biological problems. Maybe we need to figure out how our minds work and put them in computers. That has lots of advantages. You could live forever, since you can replace parts. And there would be all sorts of wonderful enhancements you can get through fairly simple biotechnology. People could communicate better, assuming you can translate between one private mental language and another, and international travel could then be banned, because you could simply e-mail yourself to any place without the risk of spreading diseases.

One of the problems with smart machines is that the first thousand of them would be wildly insane. After all, a very large percentage of people are quite crazy, infested as they are with systems of memes called religion. Still, we may well ask why we do not have artificial intelligence. So far, there has been a lot of progress making machines that understand things in particular domains and specialized areas. But around 1980, people discovered that, with the exception of certain little pockets of mathematics, computers could not solve hard problems. No computer can understand a first-grade children’s story. People then tried inventing machines that would get smart, but nothing came of it.

Computers do not really think. A lot of people are trying to figure what consciousness is, but, in fact, it may not be anything at all. Instead, there are twelve or sixteen things that the mind does, each of which is pretty complicated, like remembering what you just did. Other functions include envisioning different possibilities, explicitly formulating plans, or comparing results. Looking for a single answer or a single function is not going to be successful. Still, there are lots of fads. The worst is building those stupid little robots, which concentrate on doing something with minimal degrees of freedom.

Similarly, the interest in logic is misguided. It is too rigid and inexpressive. Nobody has gotten a logical system to make even childish analogies. And most of thinking involves using analogies. Machines, to be effective, should do their thinking in a natural language, like English, where each word may have a dozen meanings and contains metaphors that have been evolved by millions of people for thousands of years. Ambiguities allow you to change an approach a little bit and not get stuck in a problem. Logic is appropriate in the world of mathematics, but not when you are learning by example and analogy.

There has been some recent work on how to get common-sense knowledge into computers. At first they tried logic, but quickly ran into problems. If you try to make an orderly hierarchy without sufficient cross connections, you get two rather similar things very far apart on the tree, because of the differences between function and structure, for example. I concluded that you should classify knowledge by the kind of problems it can solve. As yet, we do not have a classification of that sort.

Thinking must have evolved into many ill-defined states. Bacteria have many programmed reactions, but cannot solve a problem by imaging two different actions, envisioning the results, and then comparing them. That is what humans and some primates do on the deliberative level. We need all these different levels, not some magic bullet that tries to do everything.

My own general scheme looks a little like Freud’s, who was the first to make a sophisticated architecture of how the mind works. In all cases, there is an expectation of what should happen, but then there is a bug. I am not looking for a general, elegant solution that has just a few parts. Instead, I think in terms of a large computer system that works pretty well but has bugs. Then people fix the bugs. That is what evolution does.

Christine Peterson

Preserving Human Nature: Peaceful Coexistence Among Diverse Entities in a World of Hyper-Advanced Technology

It is possible to make potentially useful projections regarding technological developments in the 50-to-250-year time frame, but strong discipline is needed to avoid our natural tendency to focus on nearer-term issues. Organizers of the Conference on the Future of Human Nature are hereby encouraged to continually redirect the group into discussing the desired time frame. This will be difficult given the senior level of many participants—not to mention their independent natures—but it will be necessary in order to make any progress on the challenge before us.

Serious forecasts in the target time frame must include what we would regard today as extraordinarily advanced technology. If our scenarios do not “sound like science fiction,” we will have failed at our task.

For purposes of this essay, human nature can be thought of as the set of characteristics our species has shared for millennia in the past. Rather than specify these in more detail here, we’ll borrow the famous judge’s quote on another topic: “We can’t define it, but we know it when we see it.” The case presented here is that it should be possible for our species to continue into the long-range future, where, by definition members of “our species” are entities sharing what we think of today as “human nature.”

Assumptions

The assumptions we need to make in order to have any coherent discussion of the 50-to-250-year time frame include the belief that some tools we use today will still be applicable, e.g., the laws of physics, the laws of economics, and the laws of human nature—this last defined above to be roughly stable for current purposes. Our understanding of these laws changes over time, but to have a discussion now we need to assume that some of today’s tools will still be useful in the future.

Technologies of Matter and Information

Our understanding of human nature today includes the fundamental tendency of some members of our species to do creative engineering: to make new technologies, both physical and informational. This, combined with the laws of physics and economics, is expected to lead to a capacity for total control of the structure of matter, down to the individual atoms. Results should include systems of molecular machinery which are more complex than those evolved by nature. Time estimates on the nearest end of our 50-to-250-year time frame; many expect it even earlier. Terms for this include: molecular nanotechnology, molecular manufacturing, and “strong” nanotechnology. It can be regarded as a highly advanced form of artificial “dry” biotech: molecular machine systems under external design and control, with a level of complexity equivalent to, and even beyond, today’s most complex such systems (ourselves).

One result of this technology should be computers of mind-boggling raw computational ability: at a minimum, the power of a billion of today’s desktop machines in the volume of a sugar cube. Raw power does not automatically translate into machine intelligence, but combined with evolutionary strategies in software, we can expect computational entities with human-level intelligence to arrive on the near-term end of our 50-to-250-year period. Unlike humans, these entities should be able to “think” together in a tightly integrated way, so we should assume that shortly after arrival they will surpass us in raw intelligence.

In considering the future of directed evolution of our species in the time frame of interest, we need to keep in mind these other technological developments. Our species’ technical ability to change the structures of ourselves and our offspring should be far in advance of germ-line genetic engineering and human somatic cell cloning. It should be possible to construct tissues and organs without using biological mechanisms at all, if desired. Changes more extreme than those encoded, or even encodable, in DNA should be technically possible.

A major benefit of these abilities is that it should become unnecessary to implement genetic changes on future generations, regardless of how disastrous a given gene is. Rather than tamper with DNA, the needed change could be implemented directly, enabling health without altering genes. The problem gene would remain in place, with its undesired operation compensated for in other ways. One early application we should expect is computer interfaces with the central nervous system; crude efforts at this are underway even today, in attempts to enable the blind to receive visual signals. Given the level of technology expected in the 50-to-250-year time frame, we should assume that these interfaces will become seamless, and that what will appear externally to be a standard human body may contain computational power many orders of magnitude beyond today’s humans. A minor result is that sense data reaching such an entity should be assumed to be recorded internally, regardless of copyright rules.

Eastern vs. Western Attitudes

In speculating on the social factors affecting the adoption of various directed evolution technologies, we can build on observations of early patterns seen today. Germ-line engineering, reproductive cloning, and even stem cell research are controversial in Western countries, roughly correlating with the prevalence of Christian-based values. In contrast, countries whose belief systems have other bases, especially much of Asia and the Middle East, are not finding these technologies so objectionable and are moving forward. (An exception is the Chinese Academy of Sciences ban on reproductive cloning, for reasons of “ethical morality.” In contrast, stem cell research is actively encouraged in China, Saudi Arabia, and Israel.)

However, as sketched above, in the 50-to-250-year time frame we can expect to move beyond controversial biological techniques to those which fix problems without changing DNA or harming embryos. Western ethical objections may decrease.

Today we are seeing ambivalence in the West toward technologies that promise major changes in the human body, such as the distinct-but-related goals of extending human life span and improving human performance significantly. Today’s U.S. administration includes both ethics advisor Leon Kass, who opposes life extension, and nanotechnology initiative leader Mihail Roco, who advocates human performance enhancement. From this we can speculate that deep controversies can be expected in the West on whether “improving” the human body should be publicly funded, or even permitted. But again, it is not clear these issues will be controversial in much of Asia, where a “full steam ahead” attitude may well prevail. If these technologies are seen as militarily or economically important—highly likely—the West may feel forced into moving forward in parallel, regardless of disagreements.

The Goal: Peaceful Coexistence

Given the seeming inevitability of a wide variety of entities in the 50-to-250-year time frame—including traditional humans, augmented humans, and machine-based intelligences—an obvious goal is to work for peaceful coexistence. This would include ensuring that the use of augmentation technologies is voluntary, and that the physical security and assets of humans are protected against coercion. A subgoal would be that traditional human families and communities continue to be able to live as they choose, without either physical force or confiscatory taxation levels making it impossible for them to live by their traditions.

How can this be accomplished in a world with entities that are far more intellectually (and, presumably, economically) powerful than traditional humans? Our species already has some experience in handling such entities: our governments. The best answer found to date seems to be the use of checks and balances. Additional insight can be obtained from the field of strategy known as game theory. Preliminary theoretical work has been done on this issue by nanotechnology theorist K. Eric Drexler and is now being written up for publication.

Summary

In the 50-to-250-year time frame, we can expect advanced molecular manufacturing and machine intelligence to far surpass near-term biological directed evolution techniques. Military and economic competition will drive these technologies to be used by nations which desire to remain in positions of technological dominance. The goal becomes to protect the safety and assets of traditional humans, i.e., those who exhibit what we today call human nature.

Special thanks to Foresight chairman K. Eric Drexler for advice on this essay.

I will try to be as outlandish in my projections as I can here, but I nevertheless think that my ideas are probably too conservative. Things will change radically. The question we need to consider, therefore, is not “Can we change human nature?” but “How can we preserve it?” Whenever one thinks about progress in the next two hundred years, one begins to sound like a science fiction writer. This is not entirely unfortunate. Many of them spend a lot of time thinking about these subjects, and some of them are not stupid. Some of them have backgrounds in physics and biology.

I am not prepared to say exactly what human nature is. I will just say that it is a set of characteristics that we have had for quite a while. So assuming that people are not going to change very much in the next 250 years, let us start with some basic things we already know about people. First, people want more money. Second, there is also the tendency of a subset of our species to do creative engineering and push technology forward. If you add these things up, what you get is technological advance to the limits allowed by nature. Estimating time limits is more difficult. But applying Moore’s Law to technological progress, we estimate that we will achieve total control of the structure of matter down to individual atoms by 2017.

In the near future, we can look forward to molecular manufacturing. Today we have already achieved atomic precision at a very minute scale. We can also make large, complex structures that are not atomically precise. The goal now is to do both—to build anything we want, no matter how big, and control it down to the atomic level. The most exciting application of this technology will be to the human body. We will be able to tackle all diseases and aging. To do this, we will be inspired by the way living systems work. After all, we are systems of molecular machines, with little machines inside us. We are beginning to understand and control these processes.

Beyond that, we will arrive at the point where we will not have to model ourselves on natural machines any longer. Instead we could design quite different machines. The difficulties, of course, are great, but the payoffs are immense, both economic and military. Presently we are able to design machines whose parts consist of individual atoms and molecules. Machines such as these give us enormous raw power. We will soon be able to put the equivalent of one billion desktop computing machines into the volume of a single sugar cube.

Combining this kind of raw power with evolutionary strategies, we will be able to understand how evolution created intelligence, and, being able to duplicate it on whatever level of detail is required, we could do it all over again. That would give us human-level machine intelligence that is far smarter than we are. We could also construct tissues and organs without using natural biological mechanisms to do so. An artificial liver, for example, would do everything that a natural liver does, but not look at all like a liver. We would not have to worry about how these artificial organs might change the way we look, since all of them would be internal.

These developments bring up certain ethical questions. One of them concerns passing permanent genetic changes on to our offspring. One solution would be confining ourselves to non-genetic changes to our children and deferring genetic changes until they are old enough to decide for themselves. Such changes might involve the integration of machines into humans or implanting chips into people. When nanotechnology is fully developed, we can expect seamless integration. So the human body, which will continue to look standard on the outside, would be quite different internally. Human enhancements could include immense computational abilities and increased recording capacities for sense data, including new ones.

Attitudes towards these controversial biological technologies vary greatly from country to country. Asia, in particular, does not share American reservations about technological advances. But other Western countries, like Sweden, are not as inhibited as we are. There are economic and military reasons why the West will eventually feel that it must move forward with technology, especially in the area of improving and enhancing human performance in general. We certainly want augmentation and enhancement to be voluntary options. Many people presently do not want to avail themselves of these possibilities, but this point of view is falling by the wayside. I personally think that parents should not make such decisions for their children.

As long as there is a choice, however, there will be imbalances in power, intelligence, and wealth. And with imbalances in power comes the possibility of physical coercion. We will want to protect weaker entities in our society from stronger ones, just as we do today. The police and the military exist to protect weaker members of society from those who would want to coerce them. People should be able to choose to continue to live in the way they wish. We should also try to preserve traditional families and communities. They need to be able to live without interference by physical force or very high levels of taxation. The Amish are an example. The best way of ensuring this goal is through a system of governmental checks and balances, where the government as a segmented entity plays the different parts off against each other.

In summary, in the time frame we are looking at—which is not very far off—we can expect advanced molecular manufacturing, machine intelligence, and genetic engineering techniques that will far surpass what we see today. The challenge is to protect and preserve human diversity.

 

 


This is a chapter from The Future of Human Nature: A Symposium on the Promises and Challenges of the Revolutions in Genomics and Computer Science (Conference).
Previous: Session One  |  Table of Contents  |  Next: Session Three
 

 

Glossary

Citation

Longer-Range, F. (2014). The Future of Human Nature: A Symposium on the Promises and Challenges of the Revolutions in Genomics and Computer Science (Conference): Session Two. Retrieved from http://www.eoearth.org/view/article/51cbef0a7896bb431f69c03a

0 Comments

To add a comment, please Log In.