The “Disruptive Technologist” Q&A with Caltech’s Virgil Griffith
Clay Farris Naff: You say you want to create a machine that feels. Yet, surely, some machines already possess awareness. (A thermostat is aware of the ambient temperature, for example.) What’s the difference between awareness and feeling?
Virgil Griffith: Merely having a representation of something is not sufficient for being aware of it. … Without closing one of your eyes you don’t know whether your right or left eye is the one observing something. However, in your head there are neurons that only light up depending on which eye is observing a particular object.
CFN: Is self-awareness implicit in the ability to feel? For that matter, do you have some understanding of what a self is?
VG: The answer is a resounding “No.” You can have awareness/feelings without self-awareness. The most palpable examples are riding a rollercoaster or playing an intense video game. In either situation you are completely absorbed in an all-encompassing experience, but nowhere are there any thoughts like, “My name is Virgil. I’m a human male. I live in a post-industrial society with patriarchal tendencies.”
CFN: It seems Hollywood expects the Singularity any day now. But what does that mean?
VG: People use the single term “Singularity” to mean different things. The broadest definition is anytime when the rate of technology’s change on society exceeds our ability to predict how technology will change us. Most narrowly, it’s when a machine intelligence is the smartest entity on the planet.
CFN: How different would you expect conscious AI to be from human consciousness?
VG: What machine consciousness could be like is an expanded version of the question, “What is it like to be a bat?” And the short answer is that we don’t know. At the very least, the bat example shows that the sensory experience can be utterly different. Would introspection be different? We have no idea.
Science aside, what will really convince people [of artificial consciousness] is when there’s a machine that looks into their eyes and convinces them that it is alive and it loves them. People have already married virtual characters. It’s a thing. They were married in Guam!
CFN: Sober thinkers worry about the Singularity. Does it raise legitimate concerns in your mind?
VG: There are those who worry about non-benevolent AI. Some of my colleagues may excoriate me for this, but I personally am not concerned. If the machines don’t like us, we’ll probably disassemble them and make ones that do.
CFN: Since the 1940s, humankind has been building weapons that can destroy civilization, and our greenhouse gas wastes may do the job even if those don’t. Do you see a benign role for a superintelligent artificial consciousness? Rather than a destroyer, could it be a savior?
VG: We could use AI-related techniques to make scientific discoveries faster. There’s even a prototype [of a machine that can design and build machines] from a group at Cornell. As far as I know there aren’t many people explicitly working on this idea. But I think it’s a great approach.
—Clay Farris Naff