Can machine-vision systems be self-aware?
If machines can be defined as "conscious," many philosophical and religious questions arise as to whether such machines possess a soul.
by Andy Wilson
EDITOR
[email protected]
Last month, as I was perusing the Web for my Wilson's Website column, I came across a book entitled The Scientist and Engineer's Guide to Digital Signal Processing. Written by Steve Smith, president and technical director of Spectrum San Diego (San Diego, CA, USA; www.spectrumsdi.com), a research-and-development group specializing in x-ray imaging and other instrumentation systems, the book provides clear explanations of practical DSP techniques without getting lost in detailed math and theory. Best of all, it's free! You can download the book at www.dspguide.com.
After browsing around Smith's Web site, I found that he is also an expert in developing novel imaging systems for medicine, security, and industrial applications. But most interesting was a link to another of Smith's books entitled The Inner Light Theory of Consciousness. Although the title may at first appear to lend itself to a book on theological doctrine, quite the opposite is true. In fact, Smith has borrowed the title from a Hugo Award-winning Star Trek: The Next Generation episode that aired more than a decade ago.
Human consciousness
The Star Trek episode is a story of human consciousness. After being attacked by a probe, Captain Picard comes to believe he has lived on a planet called Kataan for more than 30 years. And, although Picard struggles to find the reason he has been taken from the Enterprise, he can find no evidence to support his memories while on the planet.
What's interesting about the story, as Smith notices, is that Picard is limited by his own consciousness. If, for example, Picard could monitor and understand every sensor input and output, then he would realize that an outside force is affecting him.
However, this inability or limitation known as consciousness, Smith argues, is the very reason Picard cannot determine what is in fact happening. To sum this up, Smith postulates that consciousness is the irreducible entity a computational machine perceives itself to be, as the result of an ability to observe its own high-level workings and an inability to observe its own low-level workings.
Where scientists, philosophers, and others have regarded consciousness as a positive entity, Smith's argument is most compelling. It is the very limitation of the human mind to understand how each electron in each nerve and synapse is perating that leads to consciousness.
If Smith is correct, and there is no reason to believe that he is not, then this will have a high impact on the next generation of intelligent machines. Today, imaging systems such as CAT scanners, baggage-inspection systems, and machine-vision systems use von-Neumann-like architectures to perform functions such as FFT analysis, filtering, and warping. To accomplish these functions, the programmers must have an a priori knowledge of the problem and how to program specific algorithms to solve the problem.
What the system knows
Even then, such systems are not self-aware. No machine-vision system today can possibly know which bits are flowing where and why. Neither do they need to perform these functions. Should such machines possess this ability, however, they would understand how each function is operating. However, by Smith's very definition, the meaning of consciousness would imply that the design of such machines need not require that they are self-aware, merely that they can observe their own high-level workings.
If machines can then be defined as "conscious," many philosophical and religious questions arise, such as whether such machines would possess a soul. If such machine-vision systems existed, would one dare to pull the plug without consulting a priest? Or would they simply dream of electric sheep?