- youtu.be/9aOLwjUZLm0?t=1216 superconducting qubits are bad because it is harder to ensure that they are all the same
- youtu.be/9aOLwjUZLm0?t=1270 our wires are provided by lasers. Gives example of ytterbium, which has nice frequencies for practical laser choice. Ytterbium ends in 6s2 5d1, so they must remove the 5d1 electron? But then you are left with 2 electrons in 6s2, can you just change their spins at will without problem?
- youtu.be/9aOLwjUZLm0?t=1391 a single atom actually reflects 1% of the input laser, not bad!
- youtu.be/9aOLwjUZLm0?t=1475 a transition that they want to drive in Ytterbium has 355 nm, which is easy to generate TODO why.
- youtu.be/9aOLwjUZLm0?t=1520 mentions that 351 would be much harder, e.g. as used in inertially confied fusion, takes up a room
- youtu.be/9aOLwjUZLm0?t=1539 what they use: a pulsed laser. It is made primarily for photolithography, Coherent, Inc. makes 200 of them a year, so it is reliable stuff and easy to operate. At www.coherent.com/lasers/nanosecond/avia-nx we can see some of their 355 offers. archive.ph/wip/JKuHI shows a used system going for 4500 USD.
- youtu.be/9aOLwjUZLm0?t=1584 Cirac and Zoller proposed the idea of using entangled ions soon after they heard about Shor's algorithm in 1995
- youtu.be/9aOLwjUZLm0?t=1641 you use optical tweezers to move the pairs of ions you want to entangle. This means shining a laser on two ions at the same time. Their movement depends on their spin, which is already in a superposition. If both move up, their distance stats the same, so the Coulomb interaction is unchanged. But if they are different, then one goes up and the other down, distance increases due to the diagonal, and energy is lower.
- youtu.be/9aOLwjUZLm0?t=1939 S. Debnah 2016 Nature experiment with a pentagon. Well, it is not a pentagon, they are just in a linear chain, the pentagon is just to convey the full connectivity. Maybe also Satanism. Anyways. This point also mentions usage of an acousto-optic modulator to select which atoms we want to act on. On the other side, a simpler wide laser is used that hits all atoms (optical tweezers are literally like tweezers in the sense that you use two lasers). Later on mentions that the modulator is from Harris, later merged with L3, so: www.l3harris.com/all-capabilities/acousto-optic-solutions
- youtu.be/9aOLwjUZLm0?t=2119 Bernstein-Vazirani algorithm. This to illustrate better connectivity of their ion approach compared to an IBM quantum computer, which is a superconducting quantum computer
- youtu.be/9aOLwjUZLm0?t=2354 hidden shift algorithm
- youtu.be/9aOLwjUZLm0?t=2740 Zhang et al. Nature 2017 paper about a 53 ion system that calculates something that cannot be classically calculated. Not fully controllable though, so more of a continuous-variable quantum information operation.
- youtu.be/9aOLwjUZLm0?t=2923 usage of cooling to 4 K to get lower pressures on top of vacuum. Before this point all experiments were room temperature. Shows image of refrigerator labelled Janis cooler, presumably something like: qd-uki.co.uk/cryogenics/janis-recirculating-gas-coolers/
- youtu.be/9aOLwjUZLm0?t=2962 qubit vs gates plot by H. Neven
- youtu.be/9aOLwjUZLm0?t=3108 modular trapped ion quantum computer ideas. Mentions experiment with 2 separate systems with optical link. Miniaturization and their black box. Mentions again that their chip is from Sandia. Amazing how you pronounce that.
The people who work on this will go straight to heaven, no questions asked.
This job announcement from 2022 gives a good idea about their tech stack: web.archive.org/web/20220920114810/https://oxfordionics.bamboohr.com/jobs/view.php?id=32&source=aWQ9MTA%3D. Notably, they use ARTIQ.
@cirosantilli/_file/python/pytorch/python/pytorch/matmul.py Updated 2025-04-24 +Created 1970-01-01
Matrix multiplication example.
Fundamental since deep learning is mostly matrix multiplication.
Tends to be Ciro Santilli's first attempt for quick and dirty graphing: github.com/cirosantilli/gnuplot-cheat.
When it doesn't, you Google for an hours, and then you give up in frustration, and fall back to Matplotlib.
Couldn't handle exploration of large datasets though: Survey of open source interactive plotting software with a 10 million point scatter plot benchmark by Ciro Santilli
E.g. showing live data from a scientific instrument! TODO:
- superuser.com/questions/825588/what-is-the-easiest-way-of-visualizing-data-from-stdout-as-a-graph
- unix.stackexchange.com/questions/190337/how-can-i-make-a-graphical-plot-of-a-sequence-of-numbers-from-the-standard-input
- stackoverflow.com/questions/44470965/how-can-you-watch-gnuplot-realtime-data-plots-as-a-live-graph-with-automatic-up
- stackoverflow.com/questions/14074790/plotting-a-string-of-csv-data-in-realtime-using-linux
- stackoverflow.com/questions/11874767/how-do-i-plot-in-real-time-in-a-while-loop-using-matplotlib
The Google Story suggests that this practice existed in academia, where it was brought from. But I can't find external references to it easily:
At Google, the preference is for working in small teams of three, with individual employees expected to allot 20 percent of their time to exploring whatever ideas interest them most. The notion of "20 percent time" is borrowed from the academic world, where professors are given one day a week to pursue private interests.
React setups:
Introduction to Spintronics by Aurélien Manchon (2020) giant magnetoresistance section
. Source. Describes how giant magnetoresistance was used in magnetoresistive disk heads in the 90's providing a huge improvement in disk storage density over the pre-existing inductive sensors
More comments at: Video "Introduction to Spintronics by Aurélien Manchon (2020)".
Unlisted articles are being shown, click here to show only listed articles.