Angehefteter Tweet
Is a cryptographically relevant quantum computer (CRQC) possible?
I don't know. Nobody knows.
We have extensive evidence that building a CRQC is exceptionally difficult. That decades of research and breakthroughs have yielded only small improvements in the critical metrics of coherence time and error rates. That billions of invested dollars have failed to make meaningful progress in real world quantum computing power.
The current crop of quantum optimism reflects the specific properties of the time we're living in. Many people, myself included, have been surprised by the rapid ascent of AI from useless to useful over an incredibly short time. We should have seen it coming. I especially should have seen it coming because the evidence was literally in my face every time I used FSD beta in my car.
We're living in a time where the moon is suddenly back within reach, where the impossible seems impossible, and where things are changing at an ever faster pace. So it's natural to feel like quantum computing too could be just around the corner.
Why is quantum any different than AI?
AI improvements move primarily at the speed of software. While new hardware dramatically improves the rate of training, the fundamental algorithmic improvements and innovations happen at the software level which makes the iteration speed an order of magnitude or so faster. Compare the rate of development at xAI vs SpaceX - two companies with the same leadership and engineering philosophy but one working in the physical world primarily and the other in the software world primarily. SpaceX is exceptionally fast for a physical world company and yet they they're nearly 2 and a half decades in and haven't solved their original primary engineering problem - full and rapid reusability. They're close and may well have it nailed just before their 25 year anniversary. Meanwhile xAI is not even 3 years old and has delivered monumental progress on their primary goal of understanding the world through AI.
Quantum is nearly the exact opposite of AI. The algorithms and theory behind quantum computing have been fundamentally understood for over 30 years, but our ability to build and operate the necessary devices has failed to bring them into the real world. Physical progress on quantum devices is much slower than progress on rockets. While SpaceX can fabricate a new rocket in months, taking an innovation in quantum device fabrication from theory to practice is a multi-year endeavor involving developing entirely new methods of manipulating atoms or subatomic particles, new fabrication techniques for optical systems, or similar. So if AI takes 5 years and space takes 25 years, perhaps quantum takes 125? And that would assume that the same kind of exceptional leadership applied to xAI and SpaceX were brought to bear on quantum computing. It isn't.
Academics do not operate Elon style. Academics literally cannot tell whether an innovation is going to be commercially viable and that lack of constraint prevents them from making the kind of rapid and dramatic progress that Elon's ventures make. The massive funding flowing into quantum research has failed to produce a commercially useful product because nobody has created a long term mass market reason to do it which would attract the class of engineers that Elon attracts. Without this pro-human commercial motivation the likelihood of quantum computing devices becoming useful is further reduced. So the timeframe expands further, from 125 years to 5+ centuries. This is another reason that many missed AI. While it was academic, it made no progress. Elon realized that AI could make mass market pro-human products and started working on it both at Tesla and OpenAI and later at xAI, and that shift from acadamia to product is a big part of how we got here with AI. Quantum computing has no such commercial application. Stealing bitcoin and secrets? Meh.
How will we know when this changes?
With quantum computing is being developed by someone like @bscholl , @elonmusk , or @sdamico to satisfy concrete consumer needs with a specific vision for how it will be applied; OR a quantum computer is demonstrated in the real world that scales over successive generations for mulitple years. Until either or of these conditions are met, there is no reason to take any more action in response to quantum computing than we do in response to the flying spaghetti monster, or the perfectly wrapped Twinkie floating in space on the far side of the sun.
English
