You sent me great arguments about quantum computing and AI, but Gil Kalai is arguing that quantum computers are never going to work in the first place.
So my last post seems to have hit a nerve. I received several e-mails and InMails expressing a variety of opinions. Some agreed that true AI is not possible without quantum computers. Others wanted to argue about how best to define artificial intelligence. The best response was an e-mail from an extremely nice gentleman who argued that the brain is not quantum mechanical.
In his message, entitled “Quantumness, brains, and AI,” he suggested:
… the overwhelming consensus among physicists is that the human brain is not quantum mechanical, but is classical. There was considerable interest in this topic in the 1990’s, and two representative publications describing these conclusions are:
(FYI, both of those links point to the same author’s research.) I would respond that this may just be the current limits of our understanding. Also, I don’t think there is one hundred percent agreement in the scientific community. I do find his support somewhat compelling, though. Our current hypothesis is that the human brain is classical. So let me pose a follow-on question: If the human brain isn’t quantum mechanical, does that change your position on whether or not quantum computers are required to realize true artificial intelligence?
I wanted to share those links with you, and thank the sender for weighing in. I know you’ll be shocked to hear that the physics side of the house needs to conduct some more research into this subject.
Hey, Joey, this guy over here says we’re wasting our time.
In addition to reader comments, I’ve just finished reading a piece in Quanta by Katia Moskvitch. In it she discusses “The Argument Against Quantum Computers” with Hebrew University (Jerusalem) mathematician Gil Kalai. Kalai believes that quantum computers are “something of a mirage.” (Cue the e-mails in my inbox this afternoon).
His argument in a nutshell:
All physical systems are noisy, and qubits kept in “superpositions” will inevitably be corrupted by any interaction with the outside world. Getting the noise down isn’t just a matter of engineering. Doing so would violate certain fundamental theorems of computation.
Kalai has spent some time studying noise related errors and their correlations. He tried to identify which types of correlations will cause failures. His work has led him to believe that the noise levels will never be sufficiently reduced to allow for reliable, practical error correction.
Don’t bring me down, Gil.
In that case, why are IBM, Intel, Microsoft, and Google spending so much time and money to build a quantum computer? The answer may lie in the criticism of Kalai’s work, which he shared in the interview:
My critics also point to two things that they find strange in my analysis: The first is my attempt to draw conclusions about engineering of physical devices from considerations about computation. The second is drawing conclusions about small-scale quantum systems from insights of the theory of computation that are usually applied to large systems. I agree that these are unusual and perhaps even strange lines of analysis.
So why is he so certain? Because he believes the effort required to reduce the error rate to a useful level will increase exponentially with the number of qubits. He also believes that 20 qubits are enough to test his prediction. He does admit that he’s a teensy bit nervous about the potential to be proven wrong. But in his own words:
. . . I don’t need to be certain, I can simply wait and see.
To sum up, I see your “the human brain is classical,” and raise you “a robust quantum computer is not possible.” For all of us working in this nascent industry, I sincerely hope that Gil Kalai is wrong.