The New York Times, August 14, 2007

Until I talked to Nick Bostrom, a philosopher at Oxford University, it never occurred to me that our universe might be somebody else’s hobby. I hadn’t imagined that the omniscient, omnipotent creator of the heavens and earth could be an advanced version of a guy who spends his weekends building model railroads or overseeing video-game worlds like the Sims.

But now it seems quite possible. In fact, if you accept a pretty reasonable assumption of Dr. Bostrom’s, it is almost a mathematical certainty that we are living in someone else’s computer simulation.

This simulation would be similar to the one in “The Matrix,” in which most humans don’t realize that their lives and their world are just illusions created in their brains while their bodies are suspended in vats of liquid. But in Dr. Bostrom’s notion of reality, you wouldn’t even have a body made of flesh. Your brain would exist only as a network of computer circuits.

You couldn’t, as in “The Matrix,” unplug your brain and escape from your vat to see the physical world. You couldn’t see through the illusion except by using the sort of logic employed by Dr. Bostrom, the director of the Future of Humanity Institute at Oxford.

Dr. Bostrom assumes that technological advances could produce a computer with more processing power than all the brains in the world, and that advanced humans, or “posthumans,” could run “ancestor simulations” of their evolutionary history by creating virtual worlds inhabited by virtual people with fully developed virtual nervous systems.

Some computer experts have projected, based on trends in processing power, that we will have such a computer by the middle of this century, but it doesn’t matter for Dr. Bostrom’s argument whether it takes 50 years or 5 million years. If civilization survived long enough to reach that stage, and if the posthumans were to run lots of simulations for research purposes or entertainment, then the number of virtual ancestors they created would be vastly greater than the number of real ancestors.

There would be no way for any of these ancestors to know for sure whether they were virtual or real, because the sights and feelings they’d experience would be indistinguishable. But since there would be so many more virtual ancestors, any individual could figure that the odds made it nearly certain that he or she was living in a virtual world.

The math and the logic are inexorable once you assume that lots of simulations are being run. But there are a couple of alternative hypotheses, as Dr. Bostrom points out. One is that civilization never attains the technology to run simulations (perhaps because it self-destructs before reaching that stage). The other hypothesis is that posthumans decide not to run the simulations.

“This kind of posthuman might have other ways of having fun, like stimulating their pleasure centers directly,” Dr. Bostrom says. “Maybe they wouldn’t need to do simulations for scientific reasons because they’d have better methodologies for understanding their past. It’s quite possible they would have moral prohibitions against simulating people, although the fact that something is immoral doesn’t mean it won’t happen.”

Dr. Bostrom doesn’t pretend to know which of these hypotheses is more likely, but he thinks none of them can be ruled out. “My gut feeling, and it’s nothing more than that,” he says, “is that there’s a 20 percent chance we’re living in a computer simulation.”

My gut feeling is that the odds are better than 20 percent, maybe better than even. I think it’s highly likely that civilization could endure to produce those supercomputers. And if owners of the computers were anything like the millions of people immersed in virtual worlds like Second Life, SimCity and World of Warcraft, they’d be running simulations just to get a chance to control history — or maybe give themselves virtual roles as Cleopatra or Napoleon.

It’s unsettling to think of the world being run by a futuristic computer geek, although we might at last dispose of that of classic theological question: How could God allow so much evil in the world? For the same reason there are plagues and earthquakes and battles in games like World of Warcraft. Peace is boring, Dude.

A more practical question is how to behave in a computer simulation. Your first impulse might be to say nothing matters anymore because nothing’s real. But just because your neural circuits are made of silicon (or whatever posthumans would use in their computers) instead of carbon doesn’t mean your feelings are any less real.

David J. Chalmers, a philosopher at the Australian National University, says Dr. Bostrom’s simulation hypothesis isn’t a cause for skepticism, but simply a different metaphysical explanation of our world. Whatever you’re touching now — a sheet of paper, a keyboard, a coffee mug — is real to you even if it’s created on a computer circuit rather than fashioned out of wood, plastic or clay.

You still have the desire to live as long as you can in this virtual world — and in any simulated afterlife that the designer of this world might bestow on you. Maybe that means following traditional moral principles, if you think the posthuman designer shares those morals and would reward you for being a good person.

Or maybe, as suggested by Robin Hanson, an economist at George Mason University, you should try to be as interesting as possible, on the theory that the designer is more likely to keep you around for the next simulation. (For more on survival strategies in a computer simulation, go to

Of course, it’s tough to guess what the designer would be like. He or she might have a body made of flesh or plastic, but the designer might also be a virtual being living inside the computer of a still more advanced form of intelligence. There could be layer upon layer of simulations until you finally reached the architect of the first simulation — the Prime Designer, let’s call him or her (or it).

Then again, maybe the Prime Designer wouldn’t allow any of his or her creations to start simulating their own worlds. Once they got smart enough to do so, they’d presumably realize, by Dr. Bostrom’s logic, that they themselves were probably simulations. Would that ruin the fun for the Prime Designer?

If simulations stop once the simulated inhabitants understand what’s going on, then I really shouldn’t be spreading Dr. Bostrom’s ideas. But if you’re still around to read this, I guess the Prime Designer is reasonably tolerant, or maybe curious to see how we react once we start figuring out the situation.

It’s also possible that there would be logistical problems in creating layer upon layer of simulations. There might not be enough computing power to continue the simulation if billions of inhabitants of a virtual world started creating their own virtual worlds with billions of inhabitants apiece.

If that’s true, it’s bad news for the futurists who think we’ll have a computer this century with the power to simulate all the inhabitants on earth. We’d start our simulation, expecting to observe a new virtual world, but instead our own world might end — not with a bang, not with a whimper, but with a message on the Prime Designer’s computer.

It might be something clunky like “Insufficient Memory to Continue Simulation.” But I like to think it would be simple and familiar: “Game Over.”

The Beam of Light That Flips a Switch That Turns on the Brain

Kim Thompson, Viviana Gradinaru and Karl Deisseroth/Stanford University
In an optical switch in a mammalian neuron, red marks synapses and green shows photosensitive protein on the cell membrane.

The New York Times, August 14, 2007

It sounds like a science-fiction version of stupid pet tricks: by toggling a light switch, neuroscientists can set fruit flies a-leaping and mice a-twirling and stop worms in their squiggling tracks.

STOPPING ON YELLOW A genetically modified C. elegans worm stopped in response to yellow light that inhibits its neural activity.

But such feats, unveiled in the past two years, are proof that a new generation of genetic and optical technology can give researchers unprecedented power to turn on and off targeted sets of cells in the brain, and to do so by remote control.

These novel techniques will bring an “exponential change” in the way scientists learn about neural systems, said Dr. Helen Mayberg, a clinical neuroscientist at Emory University, who is not involved in the research but has seen videos of the worm experiments.

“A picture is worth a thousand words,” Dr. Mayberg said.

Some day, the remote-control technology might even serve as a treatment for neurological and psychiatric disorders.

These clever techniques involve genetically tinkering with nerve cells to make them respond to light.

Thor Swift for The New York Times
Karl Deisseroth and fiber-optic wires with laser light.
Raag Airan and Karl Deisseroth/Stanford University
Light stimulation every 200 milliseconds generates electrical activity, right, in an area of the brain associated with depression.

One of the newest, fastest strategies co-opts a photosensitive protein called channelrhodopsin-2 from pond scum to allow precise laser control of the altered cells on a millisecond timescale. That speed mimics the natural electrical chatterings of the brain, said Dr. Karl Deisseroth, an assistant professor of bioengineering at Stanford.

“We can start to sort of speak the language of the brain using optical excitation,” Dr. Deisseroth said. The brain’s functions “arise from the orchestrated participation of all the different cell types, like in a symphony,” he said.

Laser stimulation can serve as a musical conductor, manipulating the various kinds of neurons in the brain to reveal which important roles they play.

This light-switch technology promises to accelerate scientists’ efforts in mapping which clusters of the brain’s 100 billion neurons warble to each other when a person, for example, recalls a memory or learns a skill. That quest is one of the greatest challenges facing neuroscience.

The channelrhodopsin switch is “really going to blow the lid off the whole analysis of brain function,” said George Augustine, a neurobiologist at Duke University in Durham, N.C.

Dr. Deisseroth, who is also a psychiatrist who treats patients with autism or severe depression, has ambitious goals. Brain cells in those disorders show no damage, yet something is wrong with how they talk to one another, he said.

“The high-speed dynamics of the system are probably off,” Dr. Deisseroth said. He wants to learn whether, in these neuropsychiatric diseases, certain neurons falter or go haywire, and then to find a way to tune patients’ faulty circuits.

A first step is establishing that it is possible to tweak a brain circuit by remote control and observe the corresponding behavioral changes in freely moving lab animals. On a recent Sunday at Stanford, Dr. Deisseroth and Feng Zhang, a graduate student, hovered over a dark brown mouse placed inside a white plastic tub. Through standard gene-manipulating tricks, the rodent had been engineered to produce channelrhodopsin only in one particular kind of neuron found throughout the brain, to no apparent ill effect.

Mr. Zhang had implanted a tiny metal tube into the right side of the mouse’s partly shaved head.

Now he carefully threaded a translucent fiber-optic cable not much wider than a thick human hair into that tube, positioned over the area of the cerebral cortex that controls movement.

“Turn it on,” Dr. Deisseroth said.

Mr. Zhang adjusted a key on a nearby laser controller box, and the fiber-optic cable glowed with blue light. The mouse started skittering in a left-hand spin, like a dog chasing its tail.

“Turn it off, and then you can see him stand up,” Dr. Deisseroth continued. “And now turn it back on, and you can see it’s circling.”

Because the brain lacks pain receptors, the mouse felt no discomfort from the fiber optic, the scientists said, although it looked a tad confused. Scientists have long known that using electrodes to gently zap one side of a mouse’s motor cortex will make it turn the opposite way. What is new here is that for the first time, researchers can perturb specific neuron types using light, Dr. Deisseroth said.

Electrode stimulation is the standard tool for rapidly driving nerve cells to fire. But in brain tissue, it is unable to target single types of neurons, instead rousing the entire neural neighborhood.

“You activate millions of cells, or thousands at the very least,” said Ehud Isacoff, a professor of neurobiology at the University of California, Berkeley. All variety of neurons are intermixed in the cortex, he said.

Neuroscientists have long sought a better alternative than electrode stimulation. In the past few years, some have jury-rigged ways to excite brain cells by using light; one technique used at Yale made headless fruit flies flap away. But these methods had limitations. They worked slowly, they could not target specific neurons or they required adding a chemical agent.

More recently, Dr. Isacoff, with Dirk Trauner, a chemistry professor at the University of California, Berkeley, and other colleagues engineered a high-speed neural switch by refurbishing a channel protein that anchors in the cell membrane of most human brain cells. The scientists tethered to the protein a light-sensitive synthetic molecular string that has glutamate, a neurotransmitter, dangling off the end.

Upon absorbing violet light, the string plugs the glutamate into the protein’s receptor and sparks a neuron’s natural activation process: the channel opens, positive ions flood inside, and the cell unleashes an electrical impulse.

In experiments published in May in the journal Neuron, the Berkeley team bred zebrafish that carried the artificial glutamate switch within neurons that help sense touch.

“If I were a fish, and somebody poked me in the side,” (in this case, with a fine glass tip), Dr. Isacoff said, “I would escape.” But when the translucent fish were strobed with violet light, the overstimulated creatures no longer detected being prodded. Blue-green light reversed the effect.

One advantage of the Berkeley approach, Dr. Isacoff said, is that it can be adapted for many types of proteins so they could be activated by light. But for the method to work, scientists must periodically douse cells with the glutamate string.

In contrast, Dr. Deisseroth’s laboratory at Stanford has followed nature’s simpler design, borrowing a light-sensitive protein instead of making a synthetic one.

In 2003, Georg Nagel, a biophysicist then at the Max Planck Institute of Biophysics in Frankfurt, and colleagues characterized channelrhodopsin-2 from green algae. This channel protein lets positive ions stream into cells when exposed to blue light. It functioned even when inserted into human kidney cells, the researchers showed.

Neuroscientists realized that this pond scum protein might be used to hot-wire a neuron with light. In 2005, Edward Boyden, then a graduate student at Stanford, Mr. Zhang and Dr. Deisseroth, joining with the German researchers, demonstrated that the idea worked. And in separate research published last spring, Mr. Zhang and Dr. Boyden, now at the Massachusetts Institute of Technology, each found a way to also silence neurons: a bacterial protein called halorhodopsin, when placed in a brain cell, can cause the cell to shut down in response to yellow light.

The Stanford-Germany team put both the “on” and “off” toggles into the motor neurons or muscle cells of transgenic roundworms. Blue light made the creatures contract their muscles and pull back; yellow let them relax their muscles and inch forward.

Dr. Augustine and associates at Duke next collaborated with Dr. Deisseroth to create transgenic mice with channelrhodopsin in different brain cell populations. By quickly scanning with a blue laser across brain tissue, they stimulated cells containing the switch. They simultaneously monitored for responses in connecting neurons, by recording from an electrode or using sensor molecules that light up.

“That way, you can build up a two-dimensional or, in principle, even a three-dimensional map” of the neural circuitry as it functions, Dr. Augustine said.

Meanwhile, other researchers are exploring light-switch technology for medical purposes. Jerry Silver, a neuroscientist at Case Western Reserve University in Cleveland, and colleagues are testing whether they can restore the ability to breathe independently in rats with spinal cord injuries, by inserting channelrhodopsin into specific motor neurons and pulsing the neurons with light.

And in Detroit, investigators at Wayne State University used blind mice lacking photoreceptors in their eyes and injected a virus carrying the channelrhodopsin gene into surviving retinal cells. Later, shining a light into the animals’ eyes, the scientists detected electrical signals registering in the visual cortex. But they are still investigating whether the treatment actually brings back vision, said Zhuo-Hua Pan, a neuroscientist.

At Stanford, Dr. Deisseroth’s group has identified part of a brain circuit, in the hippocampus, that is underactive in rats, with some symptoms resembling depression. The neural circuit’s activity — and the animals’ — perked up after antidepressant treatment, in findings reported last week in the journal Science. Now the team is examining whether they can lift the rats’ low-energy behavior by using channelrhodopsin to rev up the sluggish neural zone.

But human depression is complex, probably involving several brain areas; an easy fix is not expected. The light-switch technologies are not likely to be used for depression or other disorders in people any time soon. One concern is making sure that frequent light exposure does not harm neurons.

Another challenge — except in eye treatments — is how to pipe light into neural tissue. Dr. Deisseroth’s spinning mouse demonstration suggests that fiber optics could solve that issue. Such wiring would be no more invasive, he said, than deep brain stimulation using implanted electrodes, currently a treatment for Parkinson’s disease.

An even bigger obstacle, however, is that gene therapy, a technology that is still unproven, would be needed to slip light-switch genes into a patient’s nerve cells. Clinical trials are now testing other gene therapies against blindness and Parkinson’s in human patients.

But even if those succeed, introducing a protein like channelrhodopsin from a nonmammal species could set off a dangerous immune reaction in humans, warned Dr. Howard Federoff, a neuroscientist at Georgetown University and chairman of the National Institutes of Health committee that reviews all gene-therapy clinical trial protocols in the United States.

In the near term, Dr. Deisseroth predicts that the remote-control technology will lead to new insights from animal studies about how diseases arise, and help generate other treatment ideas.

Such research benefits could extend beyond the realm of neuroscience: The Stanford group has sent DNA copies of the “on” and “off” light-switch genes to more than 175 researchers eager to try them in all stripes of electrically excitable cells, from insulin-releasing pancreas cells to heart cells.