It's worth noting that this is "compute-bound optimal", i.e., given fixed compute, the optimal choice is 20:1.
Under Chinchilla model the larger model always performs better than the small one if trained on the same amount of data. I'm not sure if it is true empirically, and probably 1-10B is a good guess for how large the model trained on 80B tokens should be.
Similarly, the small models continue to improve beyond 20:1 ratio, and current models are trained on much more data. You could train a better performing model using the same compute, but it would be larger which is not always desirable.
Quantum volume is a good metric but that's kind of one-dimensional take. Almost any interesting circuit doesn't requires all-to-all connectivity and superconducting QC are bad at all-to-all connected circuit so we can have interesting NISQ experiments without particularly large QV
It is not a one dimensional take... it is a stress test of qubit gate fidelity [across all qubits involved in the circuit], state prep and measurement , lifetime (coherence), memory errors, etc.
Now I agree that there are other great stress tests of quantum computer systems... but most of the industry agreed that quantum volume was a great metric several years ago. As many companies systems have been unable to hit decent QV, companies have pivoted away from QV to other metrics... many of them are half baloney.
> fidelity [across all qubits involved in the circuit]
I don't see a scenario in which the fidelity of 2QG between two far away qubits matter. Stress tests should be somehow related to the real tasks the system is intended to solve.
In case of quantum computers, the tasks are either NISQ circuits or fault-tolerant computation, and in both cases you can run them just fine without applying 2QG between far-away qubits that translate in large amount of swaps.
If you're interested in applying Haar-random unitaries, then surely QV is an amazing metric, and then systems with all-to-all connectivity is your best shot (coincidentally, Quantiniuum keeps publishing their quantum volume results). It's just not that interesting of a task.
It is not very clear from the text and from what I can say there is no "verifiability" concept in the papers they link.
I think what they are trying to do is to contrast these to previous quantum advantage experiments in the following sense.
The previous experiments involve sampling from some distribution, which is believed to be classically hard. However, it is a non-trivial question whether you succeed or fail in this task. Having perfect sampler from the same distribution won't allow you to easily verify the samples.
On the other hand these experiments involve measuring some observable, i.e., the output is just a number and you could compare it to the value obtained in a different way (one a different or same computer or even some analog experimental system).
Note that these observables are expectation values of the samples, but in the previous experiments since the circuits are random, all the expectation values are very close to zero and it is impossible to actually resolve them from the experiment.
Disclaimer: this is my speculation about what they mean because they didn't explain it anywhere from what I can see.
I've played one game of ransom. This is fun but I have some comments/suggestions.
1. Letters are sometimes barely shuffled
2. Sometimes same clue with different redacted words come one after another so you know which words are redacted
3. Sometimes (foreign names for example) the answer is not redacted due to accents. Same for similar words (complexity uncensored for "complex system"). Sometimes picture or video contains answer.
4. I guess you use some popularity metric for articles? I got Greece, ancient Greece and archaic Greece, though the topic is allegedly physics? Maybe filtering a bit more would be better
5. Before I started game I didn't know how long it is supposed to last. Apparently it's indefinite, and I lost on purpose to verify it.
6. Some feeling of progression with harder tasks would also feel nice.
It's highly non-trivial claim that macroscopic system can have quantized energy levels and exhibit measurable quantum effects. You can't just solve Shroedinger equation of 10^24 particles to show that.
Trip to Mars (~its closest point) is much longer than time we had. The chances the next one will be close enough to Earth and low enough speed so it can be matched is astronomically low and such interceptor would be quite expensive
Under Chinchilla model the larger model always performs better than the small one if trained on the same amount of data. I'm not sure if it is true empirically, and probably 1-10B is a good guess for how large the model trained on 80B tokens should be.
Similarly, the small models continue to improve beyond 20:1 ratio, and current models are trained on much more data. You could train a better performing model using the same compute, but it would be larger which is not always desirable.
reply