Uploaded February 2025 | Updated September 2026, 2 weeks ago
Building a large-scale #quantum computer requires effective error correction. Quantum error-correction codes achieve this by encoding logical information redundantly across physical qubits. A key challenge is accurately decoding noisy syndrome data to recover the correct logical information.
@googledeepmind's #AlphaQubit, a transformer-based neural network, learns to decode the surface code, outperforming traditional decoders on real-world data from Google’s Sycamore processor. This work highlights #machinelearning as a strong contender for decoding in quantum computers.
Thomas Edlich, Senior Research Engineer at Google DeepMind, presents this work to the BuzzRobot community.
Timestamps:
0:00 Introduction
0:35 What is quantum error correction?
3:03 Memory experiment
4:29 Quantum Error Correction (QEC) challenges: accuracy and speed
5:35 Machine learning for quantum error correction
8:20 AlphaQubit architecture
12:38 AlphaQubit training: pretraining and finetuning
14:04 Experimental results
17:00 Scaling AlphaQubit to larger code distances
19:24 Richer inputs for scaling AlphaQubit
20:20 Generalizing to many rounds
21:15 Calibrated outputs
22:48 Throughput
23:45 Summary and open challenges
24:58 Q&A
#ai #quantumcomputing #googledeepmind #machinelearning #errorcorrection #alphaqubit #google #computing #technology #research #neuralnetworks #coding #llms #llm #reinforcementlearning #techtalk #techtalks #aitalks #aitalk #science #python #pythonprogramming #programming
Social Links:
Newsletter: buzzrobot.substack.com
X: https://x.com/sopharicks
Slack: join.slack.com/t/buzzrobot/shared_invite/zt-2s067rv7n-guPIMGe62rbp9ncxdnOUfQ
Building a large-scale #quantum computer requires effective error correction. Quantum error-correction codes achieve this by encoding logical information redundantly across physical qubits. A key challenge is accurately decoding noisy syndrome data to recover the correct logical information.
@googledeepmind's #AlphaQubit, a transformer-based neural network, learns to decode the surface code, outperforming traditional decoders on real-world data from Google’s Sycamore processor. This work highlights #machinelearning as a strong contender for decoding in quantum computers.
Thomas Edlich, Senior Research Engineer at Google DeepMind, presents this work to the BuzzRobot community.
Timestamps:
0:00 Introduction
0:35 What is quantum error correction?
3:03 Memory experiment
4:29 Quantum Error Correction (QEC) challenges: accuracy and speed
5:35 Machine learning for quantum error correction
8:20 AlphaQubit architecture
12:38 AlphaQubit training: pretraining and finetuning
14:04 Experimental results
17:00 Scaling AlphaQubit to larger code distances
19:24 Richer inputs for scaling AlphaQubit
20:20 Generalizing to many rounds
21:15 Calibrated outputs
22:48 Throughput
23:45 Summary and open challenges
24:58 Q&A
#ai #quantumcomputing #googledeepmind #machinelearning #errorcorrection #alphaqubit #google #computing #technology #research #neuralnetworks #coding #llms #llm #reinforcementlearning #techtalk #techtalks #aitalks #aitalk #science #python #pythonprogramming #programming
Social Links:
Newsletter: buzzrobot.substack.com
X: https://x.com/sopharicks
Slack: join.slack.com/t/buzzrobot/shared_invite/zt-2s067rv7n-guPIMGe62rbp9ncxdnOUfQ
