The End of Moore’s Law – Interview with Ed Schlesinger - Exame Informática, March 1, 2012
Will Intel still exist in five years? And in 10 years? Who will lead the production of self repairing chips?
Ed Schlesinger, director of the Center for Memory Intensive Self-Configuring lntegrated Circuits (MISCIC), which is managed by the Defense Advanced Research Projects Agency (DARPA), predicts a bright future for devices that self-configure and repair their own flaws.
How are created self-configurable integrated circuits?
At our center we have with phase change materials. These are materials used in rewritable CDs which can move from an amorphous state to a crystalline state. All this is already known. By creating circuits with these materials we can turn on and off electrical connections on equipment. The goal is to create circuits that change their function – and that do it in the hardware itself and not through software. Maybe in the future my computer, mobile phone or any other device can change, not because there is a new software but because there were changes in the hardware.
How are these materials called?
The phase change materials can be chalcogens… and we also work with germanium. There are colleagues who work with strontium oxide, which has different mechanisms in relation to how they resist to changes.
How far can we go in miniaturization?
As much as some would think that technology has reached a remarkable level of sophistication. I think we’re just on the surface of what we can do with it… we begin to manipulate materials at the atomic scale and can do several interesting things.
Let me guess. We put three atoms together and we have a processor/CPU?
It may not be three atoms... a single copper atom for example has properties of copper, but it doesn’t behave in the same way as copper as a metal. If I have a multitude of atoms for example, 10 to the 23rd, which is what is usually a cubic centimeter of copper – in that case it has all the properties of copper. Now imagine you have a cube of copper to which I take out two atoms to place other kind of atoms. This new material can no longer be considered copper metal because it is below the limit at which atoms change its properties. It will have a different behavior from that which is common in copper atoms but it cannot be considered copper metal because we withdrew atoms. If I can control the concentration of atoms and, possibly, their positions, I can create materials that have properties that were never thought of before. And this is only an example.
Only with a new industry is possible to move in this direction.
Yes, it must appear a new industry. But improving the quality of life and economic development requires innovations. I was a child when the movie 2001: Space Odyssey of Stanley Kubrick came out. In that movie, the shuttle that takes the astronaut to the space station is from PANAM and today that company does not exist. One of the characters makes a video call to Earth with a CFtT TV that has the symbol of Bell that was also dismantled. People in 1968 never thought that in 2001 these companies would not exist anymore. And also never imagined it would be possible to have a phone for video calls… and never thought about the emergence of companies like Google, Intel, Microsoft…
Intel will continue to exist in 2020?
Maybe yes, maybe not.
What is still left to discover in order to atomic computing to proceed?
There are several technologies that are being explored – from the physical manipulation of the sensitive probes or self-producing tools. It is about going from electromechanical systems to technologies that allow for exerting forces or pressure, using light, magnetic forces or heat.
Moore’s Law will never fail?
Gordon Moore’s Law tells the dimensions of the transistors and also of the economic scale. In addition to predicting even smaller transistors, the law predicts that the manufacturing cost is reduced. Moore does not just say that next year someone will be able to put twice as many transistors on a CI, also says that the transistors of the next year will cost less than this year. Therefore, consumers will want to buy the latest transistors because they are cheaper and do more things. Except that Moore’s Law will also end… and we are already seeing signs of that end.
It is becoming difficult to maintain valid Moore’s Law. To keep it means increasing the cost of manufacture. And the Moore’s Law will eventually come to an end. If not in the next five years, it will be in the next 10 or 15 years. Then, maybe we will have something that we can call Moore’s Law, but that took different directions.
The manipulation of atoms can help to join the electronic human organs?
Yes, probably there will be created new interfaces between humans and electronics. There is an area that goes be the name of Brain-Computer Interface which makes us predict that in the future can be created a unique interface between electronics and the human body.
Maybe, I’ll be able to buy memory and somehow connect it to myself… and if I want more computing capacity I will also be able to buy something that also connects to me through an interface. We can imagine devices so small that enter into cells to fight diseases or help create systems that fight disease cell by cell.
A configurable device can also be an equipment without failures?
It will be possible to create circuits configured by humans and also channels that are self-configured. In the first case we have a circuit in which the manufacturer chooses a particular setting. And until you go back to the factory this device remains with the configuration. But even more interesting are the circuits that take self-configured settings in response to events or information that comes from outside. We can imagine a circuit that measures different signs and that adapts to receive the one who is stronger – all without human intervention.
It is a revolution in hardware but it also implies a revolution in software!
Without a doubt! It is also about what we can do more efficiently with the software or with the hardware… We tend to forget what already exists. An example: the smart cars. It’s not a question of when they will appear – it is an evolution that technology will provide us. Today most cars leaving the factories turn lights on when senses it’s dark. To most people are not smart cars, because it is just the inclusion of light sensors. But the cars also have water sensors to active windshield brushes and anti-skid systems and so on… The cars are already smart. And since there are cars that can park themselves, we must get used to the idea that one day cars will drive themselves.
A car with autonomous driving won’t be something much more viable than the computer that repairs itself?
I believe that the computer that repairs itself will start with some systems, where some simple but important things can be done to fix some details. And the user won’t even notice because it is a very complex system. Probably we already have things like that and do not care. For instance, a cabinet that turns off when we use too much electricity. We can imagine similar things happening in a microscale, allowing to protect the system and avoiding the user having to press a button that stops the operation in order to reestablish contact. Again, I say that is an evolution, we begin by introducing one or two things in a technology until one day we ask ourselves: how could we live without circuits that self-repair themselves?
Besides leading the research center at DARPA, Ed Schlesinger directs the Department of Electrical and Computer Engineering at Carnegie Mellon University. His area of research is applied physics with special emphasis on the area of sensors and configurable chips.