The trajectory is clear
New on our Frontier Red Team blog: using Claude to formally verify software at large scale.
Formal verification can prove that code is free of certain bugs, but historically it has only been feasible for small codebases (or required a lot of expert time).
With AI models, however, this is beginning to change. In collaboration with @theoremlabs, we used Claude to prove security properties in over one hundred real-world open-source libraries over the span of a few weeks.
Read our research in full here: anthropic.com/research/forma…
6
17
112
tzap can now generate optimized Pauli-based computation (PBC) circuits. Got some fun new PL ideas in the process…more soon: github.com/qqq-wisc/tzap
it can also visualize PBC circuits:
it can also visualize PBC circuits:
0
0
7
Our promo video for the College of AI is basically @EthanCecchetti designing a type system in the dark :)
4
1
33
Nice to see SlopCodeBench the #2 topic on HN right now, right under Dario's statement. Go @GOrlanski!
repo: github.com/SprocketLab/sl…
paper: arxiv.org/abs/2603.24755
repo: github.com/SprocketLab/sl…
paper: arxiv.org/abs/2603.24755
1
3
17
As part of SIGPLAN blog, Adrian Sampson and I started an interview series where we talk to PL luminaries. Our first one is a super fun conversation with Ranjit Jhala (@RanjitJhala). 1h26m of goodness. youtube.com/watch?v=goUZcz…
4
19
71
One of the main reasons quantum computing can be scary for many CS students is the notation.
E.g., what's the probability of measuring 0 for the ith qubit? The standard construction is horrendous, but it's really just a simple filter/map/reduce:
E.g., what's the probability of measuring 0 for the ith qubit? The standard construction is horrendous, but it's really just a simple filter/map/reduce:
2
3
25
i recently gave a personal talk about "uncertainty and my scientific journey" at PLMW. I covered the different forms of uncertainty that I've encountered through grad school, tenure-track, and post-tenure.
youtube: youtube.com/watch?v=kYZiPZ…
youtube: youtube.com/watch?v=kYZiPZ…
1
4
59
First, @awsTO. He completely changed my way of thinking about research. Before working with AWS, I was afraid of venturing in directions I didn't know much about. Now I embrace the unknown.
0
0
11
with Loris (@lorisdanto) leaving, it's the end of an era in Madison. Here are 5 things I learned from him:
3
7
99
a german magazine published a cover-page article calling my larger family (the Barghouthis) “shadow warriors who spread hatred for Israel around the world”
this is dangerously racist, and no different than the antisemitic tropes about the Rothschilds and the elders of Zion.
this is dangerously racist, and no different than the antisemitic tropes about the Rothschilds and the elders of Zion.
3
2
25
I'm coediting the SIGPLAN blog with Adrian Sampson.
New post by @benzorn and @emeryberger makes the argument that "AI Software Should be More Like Plain Old Software"
blog.sigplan.org/2024/04/23/ai-…
New post by @benzorn and @emeryberger makes the argument that "AI Software Should be More Like Plain Old Software"
blog.sigplan.org/2024/04/23/ai-…
0
1
21
I can’t get over this provision in the spending bill Biden just signed. This is the equivalent of “sure you can tell the teacher, but we’ll beat you up after school”
0
0
4
Happy to announce a program-synthesis breakthrough, led by Amanda Xu and to appear at @PLDI
tl;dr -- in 1 minute, we can synthesize a quantum-circuit optimizer that outperforms the best expertly written ones
paper: arxiv.org/abs/2211.09691
tl;dr -- in 1 minute, we can synthesize a quantum-circuit optimizer that outperforms the best expertly written ones
paper: arxiv.org/abs/2211.09691
2
8
62
whenever i hear someone make the argument that LLMs are not 'creative' or that they just 'predict the next word', I remember these lyrics by Sixto Rodriguez:
0
0
3
I find that asking it to *not* give an explanation works well, because it tries to walk through a standard algorithm and then loses track midway through :)
This is my anti-chain of thought approach :)
This is my anti-chain of thought approach :)
2
1
3
Probably the first birthday cake made with the help of generative AI (@OpenAI dalle)
My daughter riding a unicorn — she loved it
My daughter riding a unicorn — she loved it
4
0
24
Finally, the only way I could get it to do foldr correctly is to explicitly point out its mistake. And it gets it!
1
0
1
The teacher doesn't despair.
Below it gets foldl right but not foldr. It messes up double negation.
Below it gets foldl right but not foldr. It messes up double negation.
1
0
1
Finally, it gives me a correct example (minus is not associative). But it gets the evaluation wrong :)
1
0
2
Trying hard to get chatGPT to give me an example where foldl and foldr return different results ... Will it succeed?
First try: clearly the evaluation of foldl is wrong ... 🧵
First try: clearly the evaluation of foldl is wrong ... 🧵
1
0
5
My students, Abtin Molavi & Amanda Xu, have been building a flexible quantum compiler with as little code as possible, using SAT/SMT and synthesis. Here's our first paper on qubit mapping/routing to appear at MICRO 22.
paper: pages.cs.wisc.edu/~aws/papers/mi…
code: github.com/qqq-wisc/satmap
paper: pages.cs.wisc.edu/~aws/papers/mi…
code: github.com/qqq-wisc/satmap
1
1
43
An armed pineapple with eyes riding a horse through a post apocalyptic landscape, photorealistic
0
0
1
prompt: "programming languages professors at UW-Madison dancing on a frozen lake"
generated by @midjourney
generated by @midjourney
2
0
10
David starting his defense … if he withstands our offense he’ll be my 4th PhD student. Let’s see
2
0
37
Israel steals Palestinian land daily to build homes for Jewish settlers and Jewish only roads. The landscape is an archipelago as it is littered with illegal settlements and dehumanizing military checkpoints. Can’t move between two towns without a machine gun pointed at your face
1
1
17
