5 ms·
General purpose QC isn't going to happen all that soon, and may always be faced with scaling issues. But we will see QC in a very specific application before t
by kromem 3y ago
General purpose QC isn't going to happen all that soon, and may always be faced with scaling issues.
But we will see QC in a very specific application before the decade is out.
It's kind of a perfect marriage between ML and QC. If you want stochastic outputs and aren't going to know what's going on at each step in the network anyways, the issues of error correction around measurement aren't nearly as crippling as they are for general purpose QC.
While initially the work will be converting from classical to optoelectronic hardware for matching current operations, such as MIT's work this year, once we've seen greater availability of optoelectronic hardware I suspect we'll see algorithms for ML developed that would only work in photonic networks and fully exploit the quantum properties therein.
- B1FF_PSUVM 3y ago> If you want stochastic outputs and aren't going to know what's going on at each step in the network anyways You have a way with words, I'm already more at ease with the perspective of a schizophrenic AGI ...
- kromem 3y agoIt's not really any different from the current state of models, it's just that moving the same constraints to a different hardware foundation opens up significant performance gains.
- eigenket 3y agoI am a researcher in quantum information and this seems highly unlikely to me. Quantum computers, as far as I can tell, get significant improvements over classical computers only for quite specific highly mathematically structured problems like the discrete log problem or simulating other quantum systems. I don't think it's likely that there will be substantial "quantum adantage" in the machine learning/ai area.
- itsacomment 3y agoDon't greyed out comments mean someone is downvoting you? What's going on in this thread, this is the second comment of yours that's factually far better than the preceding one but greyed out. I agree with you, quantum ai has only ever been people shouting "quantum" at black box neural nets, as if that had ever produced a viable algorithm yet...
- sporadicallyjoe 3y agoThat's correct. Comments become grey when they receive multiple downvotes.
- programjames 3y agoI'm not a researcher here, but I'm pretty sure Bayesian neural networks are realizable with quantum computers, which is a substantial "quantum advantage".
- programjames 3y ago"While initially the work will be converting from classical to optoelectronic hardware for matching current operations, such as MIT's work this year," I'm pretty sure you're talking about mean Dirk Englund's photonic matrix multiplication. Also, you should do better at explaining what you mean, because if I wasn't already very familiar with this I would have no clue what you're talking about.
- kromem 3y agoNo, that was 2022. I'm talking about this one: https://dl.acm.org/doi/10.1145/3603269.3604821 https://dl.acm.org/doi/10.1145/3603269.3604821
- programjames 3y agoLook at the author list... Dirk Englund. What do you mean by "no"? Do you mean, not that paper (that neither of us linked because there are probably several on matrix multiplication) but instead this other one that also has optics and ml and Dirk? I'm seeing a pattern of you understanding what you write, but it being impossible to decipher on my end.
- kromem 3y agoYou said "I'm pretty sure you mean Dirk Englund's photonic matrix multiplication." No, I don't mean his photonic matrix multiplication from 2022. The issue seems more to be with your over specificity leading to ambiguity than anything else here.