Quantum info, Magic the Gathering jokes, and whatever else. | Postdoc @ UWaterloo IQC, Perimeter Institute.
Joined July 2009
- Tweets 1,252
- Following 388
- Followers 260
- Likes 1,137
171 Photos and videos
Jun 10
arxiv is about to lose a lot of its clean structural spines and get a lot of equations labelled (β ) and (β)
1
115
Jun 10
This is from a job advert in the UK. Without further context, what kind of job do you think this is for?
1
2
163
Jun 10
If I saw this without context, I would think it was a kindergarten teacher. This equates to about $77k USD, which is obviously on the high end, but totally exists in the US in some very high-cost-of-living areas or at private schools.
1
107
Jun 10
But it's actually... a math professor! With a PhD! "Grade K" is one of the steps in their gradation of experience. https://www.bristol(dot)ac(dot)uk/jobs/find/details/?jobId=388703 (X blocks this link!?)
82
Jun 4
Excited to be giving a keynote speech at SAIR's Science x AI Summit at the end of the month.πI'll talking about physics and Lean!
sair.foundation/events/scienβ¦
1
14
356
May 29
Sometimes Claude will refer to subagents as child processes, or for short, children. When this propagates through notes, sometimes it misunderstands itself. Which leads to it writing summaries of mathematical analysis like this:
1
78
May 19
The first stage of SAIR's "Math Distillation Challenge" completed. Happy that I got 8th place! Right below, *ahem*, someone named "magmaballs".π
Stage 2 will involve generating Lean proofs and disproofs. Excited to see will come out of it!
competition.sair.foundation/β¦
8
217
May 17
"All the world's a stage,
And all the men and women are merely players;
They have their exits and their entrances,
And one man in his time plays many parts."
Until you appreciate this, you will not correctly interpret mechanistic interpretability research on LLMs.
1
1
77
May 15
*Log on and see the agent did overnight*
Me: "Wow this is a neat optimization, sounds smart but I don't understand why it works. Can you explain it?"
AI: "Your prompt explicitly instructed me on this optimization, bc ..."
Oh, it wasn't genius AI, *I* was smart! And forgetful. :)
2
81
May 14
Who thought that Hall of the Mountain King was a good song for waiting on hold with Urgent Care!?
this is not what ANYONE needs
1
59
May 6
There are two kinds of tweets on my feed:
* Naomi Osaka is CLRS Algorithms
* Naomi Osaka is Elesh Norn
May 4
Love seeing Naomi Osaka honor the CLRS Algorithms textbook at this year's Met Gala
149
May 5
Claude Opus 4.7: Arguably a genius in every domain
Also Claude Opus: "3 isn't a valid hex char" ??
1
85
Apr 25
I ask Claude some follow up questions on a thread.
"These are sharper objections than they might look. Let me take them seriously."
Is this incredible negging or what? If a student in a class asks a question and the professor says it's "a sharper question than it sounds"... π
100
Apr 23
The Lean language (@leanprover) has utilities for verifying software, and AI is adept at using it. But can AI prove correctness for a *foreign architecture* with *no existing API*? It turns out, yes!
@HarmonicMath's Aristotle wrote a z80 emulator:
github.com/Timeroot/Z80Emu (1/n)
1
6
58
48,323
Apr 23
It's just that no sane human would bother trying to verify this. But this works!
I think this spells a bright hope for software verification in the future.
There are some open questions: could an AI abstract these tools to make a reusable kit for more programs?
1
2
367
Apr 23
In other words, could it move 'up'? The codebase as written doesn't scale well to larger programs. The fact that it decompose is a strong sign, but could an agent (say) define its own Hoare logic for the assembly programs?
I didn't try, but I'd be optimistic! (/end)
1
4
334