levbishop.com
Quantum factotum @ IBM.
https://arxiv.org/a/bishop_l_1.html | https://research.ibm.com/people/lev-bishop | https://github.com/levbishop
83 posts
287 followers
225 following
Regular Contributor
Active Commenter
comment in response to
post
YouTuber Electroboom has a playlist for his international travels where he takes a closer look at the quality of electrical systems wherever he visits. Super interesting if you’re into that kind of thing youtube.com/playlist?lis...
comment in response to
post
In case you missed it @mcsweeneys.net had a great “marriage proposal in office jargon” www.mcsweeneys.net/articles/a-m...
comment in response to
post
Dropping E-Bombs
comment in response to
post
Here's a pdf havantcivicsociety.uk/wp-content/u...
comment in response to
post
I've found "Foils" for powerpoint slides is a good shibboleth to spot a true-blue IBMer.
comment in response to
post
I think IBM jargon is less SV, more Business English (circle back, deep-dive), plus a little local idiosyncracy (RA, Client Zero). There's "IBM Jargon and General Computing Dictionary" you can track history through the editions up to 10th (1990).
comment in response to
post
Maybe Valley Speak by Kopp and Ganz (I haven't read myself)
a.co/d/6ATwKs7
comment in response to
post
The hackaday retrotechtacular category is a great way to discover many films like this hackaday.com/tag/retrotec...
Also Dan Gelbart is making modern educational videos with a similar vibe youtu.be/kLgPW2672s4?... his stuff is a goldmine.
comment in response to
post
It’s a great list I’ve read 90% of them and agree they’re all-timers, including Lena for sure
comment in response to
post
Not as good as a graph. Better than going empty handed.
comment in response to
post
This is basically finding the best sequence of machine instructions to implement an arbitrary computation on 2 qubits. It's been identified as an important problem already in 1994, seriously attacked 2003 or so, major progress in the last couple years. This paper puts some final touches on it.
comment in response to
post
OK, like classical computer programmers don't code directly to the machine code as executed by the CPU, but in a more expressive language, letting a compiler translate. Similar compilers are needed for quantum computers. We give an optimal implementation of one aspect of this: SU(4) synthesis.
comment in response to
post
For the benefit of @dulwichquantum.bsky.social who only read papers in meme form, here's a low-effort meme for my paper with @evm9.dev "Two-Qubit Gate Synthesis via Linear Programming for Heterogeneous Instruction Sets" arxiv.org/abs/2505.00543
comment in response to
post
This work was led by Evan McKinney @evm9.dev during an internship at IBM Quantum last Summer. You should consider interning with us - you get to do fun stuff like this! http
comment in response to
post
I tentatively claim this finally "solves" the 20-plus year-old 2-qubit compilation problem as first attacked in ~2004 eg arxiv.org/abs/quant-ph... (The harder version of the problem, where there is also cost associated to the 1-qubit gateset, remains open).
comment in response to
post
Putting this all together this is almost as performant as the methods specialized to specific gatesets, but general to arbitrary gatesets, even including certain classes of parametric gates and approximate compilation. We call this method GULPS (Global Unitary Linear Programming Synthesis).
comment in response to
post
Given this trajectory we have a set of nonlinear problems to find the SU(2) parameters corresponding to each step. These problems are independent, can be solved in parallel and are small enough to be amenable to nonlinear optimization (especially since we have a guarantee that a solution exists).
comment in response to
post
In the present work, we avoid the repeated polytope projection by restructuring the heterogeneous compilation problem as a linear program. Modern LP solvers are marvels of engineering fast and robust. The LP solution gives an optimal trajectory through the Weyl chamber for arbitrary gatesets.
comment in response to
post
The other main approach has been to find special-case compilations, specialized to hardware-relevant gatesets. This includes our previous work for XX-class compilation arxiv.org/abs/2111.02535 as well as eg Alibaba's work on sqrt(iSWAP) arxiv.org/abs/2105.06074.
comment in response to
post
For concrete 2-qubit compilation, one approach is numerical optimization, NuOP being a good example arxiv.org/abs/2106.15490 but these methods are still quite slow and due to the shape of the optimization landscape they can often get stuck in local minima and generate suboptimal compilations.
comment in response to
post
It is also relatively expensive, requiring manipulation of large structures of bignum rational expressions. Together, these factors have relegated this analysis as mostly useful for offline QPU gateset comparison/optimization. For example our tool github.com/qiskit-commu... supports such use cases.
comment in response to
post
As was shown by Peterson et al. in arxiv.org/abs/1904.10541 by repeatedly projecting this 9-dimensional polytope one can determine the accessible SU(4) from a given sequence. However, this method non-constructive, giving no information about the required 1-qubit gates to synthesize the target.
comment in response to
post
Really, like… the concept of expertise is lib coded at this point. The 2025 American conservative view of knowledge appears to be that there are instinctual polymath common sense All Things Knowers and egghead specialists who pretend to know things but can easily be surpassed by Donald, Elon, etc.
comment in response to
post
Amy Gleason?
comment in response to
post
See “levels of friction” for a different spin on this open.substack.com/pub/thezvi/p...
comment in response to
post
I’m not sure if the headline is aiming for clickbait or they assume their readers wouldn’t understand what Rutgers is. (Either case is bad…)
comment in response to
post
I will support your Kickstarter for this and subscribe to your newsletter. It should be named “The Machine Stops”.
comment in response to
post
comment in response to
post
They’ll have to sing the “I’m very sorry” song
comment in response to
post
You really should.
comment in response to
post
I like NMAs take youtu.be/NWZcePcgQlc?...
comment in response to
post
My “following” feed can be hectic, depending on the latest happenings, but it’s from people I chose to follow and I can unfollow if too much.
For easily catch up on science “paper skygest” bsky.app/profile/did:... shows only posts about articles (still from people I follow) - much calmer timeline.
comment in response to
post
Frontline?
comment in response to
post
Thanks Jaz. Any chance to get the repo explorer working again?