If we are lucky, our pets may keep us as pets

January 18, 2002

Originally published 2001 at Brad Templeton’s website. Published on KurzweilAI.net January 18, 2002.

Many have debated just how the first superior or “post-human” intelligences might come to be. While some don’t think we’ll ever spawn something smarter than humans, many people in the AI, uploading, nanotechnology and related communities think it’s only a question of how and when.

The big debate is whether we will do this by creating an “uploaded” human being, or a true artificial intelligence. An uploaded human being comes about by scanning the brain of an existing person in some fashion, and using that information to create an artificial brain with the same neural pathways and other connections and systems that make a brain what it is. This uploaded person could be a copy of a living person or a re-animation of a deceased one, or the process could be done gradually, replacing a biological brain neuron by neuron, turning a living cell-based brain into one based on something else right in the skull.

The uploaded person would think and act just like a traditional human, if given a similar environment and body (biological or otherwise) to be part of. However, the temptation to enhance this new brain would be tremendous, and many predict such people would quickly become super-human. They would start by thinking faster, and expanding and perfecting their memory and other mental abilities. Then they might connect themselves with outside data networks and databases, giving them the ability to learn knowledge just by copying it, or gather more of it faster than any natural human. And of course they (we?) will enhance their brains in ways that present-day thinkers have not yet conceived.

Others expect the success will first come to the effort to create an artificial intelligence with hand-written or evolved software, running on computers like the ones we have today, or their successors. Such beings would be more alien to us than uploaded people, though they would be our children in a fairly real sense, designed at first by us, instilled with our values, and quite possibly raised by us as children are. However, clearly they will begin superior to us in many aspects (perfect memories, superb mathematical and rote skills and much more) and initially inferior to us in others. Soon they would surpass our abilities in almost all areas.

Modern humans can’t really conceive of how these “post humans” would think or act, any more than apes can write stories about human philosophy. Indeed, noted author Vernor Vinge calls this point of transition to post-human intelligence a “singularity,” a metaphorical reference to the mathematical discontinuities beyond which prediction is impossible.

I happen to slightly favor the uploading camp. To upload a mind it is necessary only to understand the lower level workings of the brain enough to recreate them in another medium. One need not understand much about the higher level activities which bring about conscious and intelligent thought. Just as a hardware engineer can build a computer which can play chess knowing only about how transistors and logic gates work. The chess software she simply copies. To build a real AI requires that we actually either understand how intelligence works–which we are not close to doing, or perhaps that we understand its mid-level functions and create something we can turn intelligent by raising it over the course of many years, just as we do with our own babies.

However, the uploading scenario presents a rather disturbing conclusion. The first super-beings may not be based on humans at all, but instead may be apes.

In the course of modern science, it is always the case that we experiment with animals first, years before we attempt anything on people. It’s the ethical way, and in many cases the only legal way. As such, as we develop the technology to scan or convert an existing brain into an artificial form, we’ll try this first on animals. We’ll start with lower ones, and then work up to our closest relatives, the chimpanzee and bonobo.

Some suppose the uploading process might begin by scanning a recently deceased human brain, which can be done with minimal legal complication. Others feel it might be better to work with a living brain, or a healthy brain that was killed deliberately for the purpose of scanning–what might be called a “destructive scan” if expressed in cold technical language. We’ll do this with apes several years before we do it with humans. Of this I am fairly certain.

Indeed a great moment of success will come when somebody first creates an artificial brain that is a copy of a real chimp’s brain, and which is shown to all outside signs to act, think and remember like the original living chimp. It is unlikely that scientific ethics or law would allow things to be done with humans until the process has been reliably demonstrated multiple times on apes. That even applies to dead humans, since many, including myself, would argue that the human with a non-biological brain is still a human being and worthy of certain rights.

Once this chimp brain is created, it will cause a flurry of research. Quite possibly, the “software” part of the brain will be published and made available for others to work on, and the hardware will be readily available too.

Indeed, the software of this chimp brain might be made available for free distribution. An “open source” ape, for all to experiment on.

And they will experiment on it. Once again, even if a human brain is similarly available, moral and legal considerations will limit what experimentation can be done, while actions on the ape-brain will probably not be nearly as limited.

Apes however are remarkably similar to humans. As you may know, chimps share 98% of our DNA. In addition, we have made intensive study of the ways in which they are different, and we will attempt to learn more.

Thus some of the first experiments on this artificial chimp brain will be to enhance it along the lines that humans and chimps are different. Humans are not so qualitatively different in our brains from chimps, though the few differences have a magnified effect in our capabilities. We have more of certain types of brain structures, and some of our structures are larger and have more neural connections. There is no component of a human brain not found in a chimp brain. Experimenters will quickly try to see what happens if you modify those aspects of the working chimp brain. They will also “graft” information from post-mortem and live scans of human brains, where available.

If the artificial chimp brains “run” much faster than biological ones, they will be able to perform these experiments quickly. They may be able to have their computer play out a thousand different experimental scenaria, each playing out years of biological scale time–perhaps in just a day of real time. They will quickly learn what works and what doesn’t, what enhances and what doesn’t. And there will be many of them.

I think quite quickly they could create a chimp brain capable of human level intelligence or beyond. It may then need training or “rearing” by real human parents, but it will be a very quick and supremely capable learner. All this will happen much more quickly than the ethical changes to occur which would allow scientists to do similar experiments on human based brains.

The chimp brain might also marry the chimp with the best that grounds-up artificial intelligence has to offer at the time. The post-human may be the result of the combination of the two. Vinge, in his 1966 story, “Bookworm, Run” (now back in print in his latest collection) imagined a computer-linked live chimp that tried to escape its creators using its near to human (and in some ways superior) level intelligence.

Of course, once these enhanced chimp brains–or perhaps best to describe them as hybrids of chimpanzee base, human improvements and AI software–do become as smart as or smarter than us, they will of course continue the research on how to enhance themselves. At this point they may decide their own ethical rules about how this takes place. But smart as they are, and with human friends as they will have, they will quickly become powerful in the human world.

Charleton Heston as Taylor in Planet of the Apes zoomed into the future to “a planet where apes evolved from men.” Perhaps this wasn’t entirely ridiculous.

Assuming they can think both better and faster than we can, it won’t be long, in fact, before they are running the “planet of the apes.” Is this a dystopian nightmare or a potential paradise? That, we can’t predict. They may feel quite a debt to their creators–humans tend to think that way–or they might hate us. They might well also arrange for humans to go through the same process that improved them, now that it is fully developed, and as such humans themselves would reach post-human intelligence.

But the chimps would have gotten their first, and the humans would not necessarily be any smarter than they are. Just more experienced at living in real time, and more used to the enhanced parts of their minds. Minsky wondered if ordinary humans would be lucky enough to be kept as pets by these superior intelligences; one wonders if he imagined that “they” might be our former pets to begin with?