4 ms·

Supremacy is the word that seems awkward here. "We might have something to show by the end of the year" seems to fall a little short of supremacy. And there is

Supremacy is the word that seems awkward here. "We might have something to show by the end of the year" seems to fall a little short of supremacy. And there is something about a double exponential rate? It still seems like we have a double exponential rate of "might show something". Of course I do look forward to seeing the quantum computer playing chess or something.

jchw 7y ago
Quantum supremacy is actually a more specific concept than you might be thinking: https://en.m.wikipedia.org/wiki/Quantum_supremacy https://en.m.wikipedia.org/wiki/Quantum_supremacy
hedora 7y ago
There is such a thing as a double exponential rate: https://en.m.wikipedia.org/wiki/Double_exponential_function https://en.m.wikipedia.org/wiki/Double_exponential_function If their claims are correct, using the Wikipedia example, if they are as fast as a cell phone in year one, then they should be faster than something like a million of the biggest super computer in year two. In year three, the incremental improvement will be incomprehensibly faster. So, they are making some pretty strong claims. Their current claims are definitely stronger than the ones that led to the AI winter. (edit: and, if it could do physical simulation poorly in year one, by year four or five, it would be simulating an incalculable number of universes starting from the Big Bang, at much faster than real-time. I guess they could then recursively contain comparable quantum computers, which would progress even faster than the host machine?)
- perl4ever 7y ago
  As shown in the graph in your link, the double exponential has a sharper "elbow" and therefore it seems to me that if you start at the same y-value, then the double exponential will lag behind the single one for a period of time, which in practice might be any length.
scottlocklin 7y ago
Some boob at google's QC group overfit 3 data points, two of which were questionable, and called it a "law" named after himself. I look forward to people attempting to justify why this is still a "law" when there is still no quantum supremacy 10 years from now...
- lowdose 7y ago
  >Nevin’s Law is currently more of an affectionate term for a rule coined by Google‘s Hartmut Nevin. At the company’s Spring Quantum Symposium this May, Nevin made the claim that quantum systems are increasing in performance at a doubly-exponential rate. This means, rather than doubling in performance with successive iterations as was the case with classical computers and Moore’s Law, quantum technology is increasing in performance at a much more dramatic rate. It took 50 years to go from punch card systems to iPhones: if Nevin’s Law is true we’ll see quantum systems increase in a fraction of that time. What a character!

jchw 7y ago

Quantum supremacy is actually a more specific concept than you might be thinking: https://en.m.wikipedia.org/wiki/Quantum_supremacy https://en.m.wikipedia.org/wiki/Quantum_supremacy

hedora 7y ago

There is such a thing as a double exponential rate: https://en.m.wikipedia.org/wiki/Double_exponential_function https://en.m.wikipedia.org/wiki/Double_exponential_function If their claims are correct, using the Wikipedia example, if they are as fast as a cell phone in year one, then they should be faster than something like a million of the biggest super computer in year two. In year three, the incremental improvement will be incomprehensibly faster. So, they are making some pretty strong claims. Their current claims are definitely stronger than the ones that led to the AI winter. (edit: and, if it could do physical simulation poorly in year one, by year four or five, it would be simulating an incalculable number of universes starting from the Big Bang, at much faster than real-time. I guess they could then recursively contain comparable quantum computers, which would progress even faster than the host machine?)

perl4ever 7y ago
As shown in the graph in your link, the double exponential has a sharper "elbow" and therefore it seems to me that if you start at the same y-value, then the double exponential will lag behind the single one for a period of time, which in practice might be any length.

scottlocklin 7y ago

Some boob at google's QC group overfit 3 data points, two of which were questionable, and called it a "law" named after himself. I look forward to people attempting to justify why this is still a "law" when there is still no quantum supremacy 10 years from now...

lowdose 7y ago
>Nevin’s Law is currently more of an affectionate term for a rule coined by Google‘s Hartmut Nevin. At the company’s Spring Quantum Symposium this May, Nevin made the claim that quantum systems are increasing in performance at a doubly-exponential rate. This means, rather than doubling in performance with successive iterations as was the case with classical computers and Moore’s Law, quantum technology is increasing in performance at a much more dramatic rate. It took 50 years to go from punch card systems to iPhones: if Nevin’s Law is true we’ll see quantum systems increase in a fraction of that time. What a character!