Matthew P. Grosvenor

An engineer of many sorts with a thing for high performance systems.

[Matthew P. Grosvenor @mpgros]

@heyalexhey That sounds like a terrible idea? Far from talent, far from resources, far from connections to grow and scale. If you last more than 12mo you’re going to be moving to Sydney (or San Francisco).

[Matthew P. Grosvenor @mpgros]

@usanewshq Both. That’s the whole point of relativity.

[Matthew P. Grosvenor @mpgros]

Remember, when you come across someone who “does it all”. Remember that they rarely do it all well.

[Matthew P. Grosvenor @mpgros]

@AbiGreen04 He’s incredible and very funny. But he has spent years (decades!) practicing that skill (like any artist).

[Matthew P. Grosvenor @mpgros]

@ZPostFacto This is a c++ problem. Not a C problem

[Matthew P. Grosvenor @mpgros]

@timurkuran It is a general trend that the more educated you are, the more you 1) understand and accept modern science. 2) care about your fellow humans. So you’ve mixed cause and effect. Universities aren’t blue because they hate reds, they’re blue because reds hate science.

[Matthew P. Grosvenor @mpgros]

@boyto_ @jamonholmgren Yes. Getting Claude and others NOT to do this is really hard. There’s some protocol parsing work in doing where I’ve had to fight it hard to use external references for test validation rather than “I declare the code I wrote is great”.

[Matthew P. Grosvenor @mpgros]

@boyto_ @jamonholmgren I’ve tried a few new things recently on this front. Mutation testing is interesting. Make mutations to my code. See if the tests catch it. Wild stuff. Great for improving test coverage.

[Matthew P. Grosvenor @mpgros]

@boyto_ @jamonholmgren I’ve taken to accepting this as ok, as long as 2 things remain true: 1, correctness checks pass. 2. No performance regressions. Number of tests and importance of the testing has escalated dramatically. I’ve even started with applying formal methods as another defence.

[Matthew P. Grosvenor @mpgros]

@gavinrbrown1 Being liberal is not the same as being sloppy. And sloppy is what we’ve got. Long lists of citations lumped together to “bulk out”‘ the paper. Authors who cherrypick a phrase out of context. Reviewers gating acceptance on their papers being cited. This all happens now.

[Matthew P. Grosvenor @mpgros]

@rbnmckenna86 Because so so many papers are bait and switch marketing exercises. If you look at the actual methodolgy and results you’d be shocked what makes it through. Complete nonsense in top tier venues. But if you haven’t checked how would you know?

[Matthew P. Grosvenor @mpgros]

@alemelk This is a strawman argument. The actual phrase is “ If a submission contains incontrovertible evidence that the authors did not check the results of LLM generation, this means we can’t trust anything in the paper.” Which is all three of those things.

[Matthew P. Grosvenor @mpgros]

@JimDMiller @predict_addict @tdietterich @arxiv “ If a submission contains incontrovertible evidence that the authors did not check the results of LLM generation, this means we can’t trust anything in the paper.” This is the standard. Stop straw manning this.

[Matthew P. Grosvenor @mpgros]

@scottnarmstrong Because so so many papers are bait and switch marketing exercises. If you look at the actual methodolgy and results you’d be shocked what makes it through. Complete nonsense in top tier venues. But if you haven’t checked recently how would you know?

[Matthew P. Grosvenor @mpgros]

lots strawman fallacy arguments nitpicking over edge cases in citation systems. Here’s the key phrase “if a submission contains incontrovertible evidence that the authors did not check the results of LLM generation, this means we can’t trust anything in the paper.” It’s clear.

[Matthew P. Grosvenor @mpgros]

@VladChituc Reading the full paper before citing it is so important. It’s absolutely wild what you find when you read more than just the abstract of a paper. So many papers are bait and switch and you’d never know unless you read them properly.

[Matthew P. Grosvenor @mpgros]

@eduleadership The public is finding out that “eminent” professors haven’t even read most of the papers their names are on.

[Matthew P. Grosvenor @mpgros]

@zdeborova 1) your argument seems to be “verification and fact checking is good for everyone else but not for me”? Am I hearing that right?
2) Nobody is penalising anyone for “minor mistakes”. “incontrovertible evidence that the authors did not check the results” is the standard.

[Matthew P. Grosvenor @mpgros]

@SamuelBiagetti They are exposing the publish or perish reality of academia. Exposing the open secret of academia, that senior academics have rarely even read the papers their names are attached to let alone checked the references. And yet have the most to loose by the policy being applied.

[Matthew P. Grosvenor @mpgros]

@mcxfrank “If a submission contains incontrovertible evidence that the authors did not check the results of LLM generation, this means we can’t trust anything in the paper” — this is not a debate about edge cases in referencing.