I built an open-source AutoML tool that searches 17 quantum ML classifiers (PennyLane + Optuna) and benchmarks them against classical baselines

hiddentalent • 1d ago Quantum ML tooling is thin because scientifically there is no quantum advantage to any known ML application. It's just people smashing together buzzwords in the hopes that it'll create internet cred. Learn rigor. Do real science. Stop smashing buzzwords together. Able-Let-5930 • 1d ago can you tell me which aspect of quantum algorithms really has scope. because everywhere people say that there is no verifiable advantage, for like any practical algorithm. even shors is far away, and we need a lot more qubits. and by that time we would have upped our encryption protocols anyways. hiddentalent • 1d ago Yes, what you're saying is the scientific consensus. I can't tell you otherwise, or I'd be a billionaire. more reply more reply Continue this thread Continue this thread Continue this thread Able-Let-5930 • 1d ago can you tell me which aspect of quantum algorithms really has scope. because everywhere people say that there is no verifiable advantage, for like any practical algorithm. even shors is far away, and we need a lot more qubits. and by that time we would have upped our encryption protocols anyways. hiddentalent • 1d ago Yes, what you're saying is the scientific consensus. I can't tell you otherwise, or I'd be a billionaire. more reply more reply Continue this thread Continue this thread hiddentalent • 1d ago Yes, what you're saying is the scientific consensus. I can't tell you otherwise, or I'd be a billionaire. more reply more reply Continue this thread Kinexity • 1d ago Open source? More like source never opened. You didn't build shit, Claude did. You should know better as a PhD student. Able-Let-5930 • 1d ago I mean doesnt everyone use AI to write code? it just optimising your work load. ShacoinaBox • 10h ago i don't but tbf i primarily write stuff in langs that LLMs are, currently, p objectively (n understandably) terrible at. lil bit of boilerplate n trivial stuff works ok, otherwise it's jus giving u something completely broken, using something outdated (less common of an issue than a year ago tho) or giving a poor implementation. like dafny: not a lot to train on, not a lot of examples to use, best possible analog is ada which still works differently n is lexically+semantically quite different. it's understandable that it blows at it currently n idc cus i like writing it by hand lol n fwiw im not, nor would i ever, work in industry. obv if I did, id prob be using LLMs a ton. Kinexity • 1d ago Why do I even have to explain the nuance here? Have you spent at least 10 seconds thinking where the problem is before commenting this? Same shit have been said again and again for the last 3 years. There is no problem per se with using AI for coding. The problem is people just going "Claude, take the wheel" and sharing unchecked AI slop. more replies more replies Continue this thread Continue this thread Continue this thread Able-Let-5930 • 1d ago I mean doesnt everyone use AI to write code? it just optimising your work load. ShacoinaBox • 10h ago i don't but tbf i primarily write stuff in langs that LLMs are, currently, p objectively (n understandably) terrible at. lil bit of boilerplate n trivial stuff works ok, otherwise it's jus giving u something completely broken, using something outdated (less common of an issue than a year ago tho) or giving a poor implementation. like dafny: not a lot to train on, not a lot of examples to use, best possible analog is ada which still works differently n is lexically+semantically quite different. it's understandable that it blows at it currently n idc cus i like writing it by hand lol n fwiw im not, nor would i ever, work in industry. obv if I did, id prob be using LLMs a ton. Kinexity • 1d ago Why do I even have to explain the nuance here? Have you spent at least 10 seconds thinking where the problem is before commenting this? Same shit have been said again and again for the last 3 years. There is no problem per se with using AI for coding. The problem is people just going "Claude, take the wheel" and sharing unchecked AI slop. more replies more replies Continue this thread Continue this thread ShacoinaBox • 10h ago i don't but tbf i primarily write stuff in langs that LLMs are, currently, p objectively (n understandably) terrible at. lil bit of boilerplate n trivial stuff works ok, otherwise it's jus giving u something completely broken, using something outdated (less common of an issue than a year ago tho) or giving a poor implementation. like dafny: not a lot to train on, not a lot of examples to use, best possible analog is ada which still works differently n is lexically+semantically quite different. it's understandable that it blows at it currently n idc cus i like writing it by hand lol n fwiw im not, nor would i ever, work in industry. obv if I did, id prob be using LLMs a ton. Kinexity • 1d ago Why do I even have to explain the nuance here? Have you spent at least 10 seconds thinking where the problem is before commenting this? Same shit have been said again and again for the last 3 years. There is no problem per se with using AI for coding. The problem is people just going "Claude, take the wheel" and sharing unchecked AI slop. more replies more replies Continue this thread ShacoinaBox • 10h ago you laugh you lose: look op's reddit bio lol Direct_Pop_4278 • 1d ago Curious whether conditional search spaces across 17 circuit families is the right abstraction — or if QML AutoML should instead fix architecture and only tune angles/depth?