Ben Recht
@beenwrekt.bsky.social
There's nothing wrong with bitter things. bsky.app/profile/been...
Since it's 5 PM here, let me follow up this post with my top 10 negroni variations.
Revisiting Sutton’s Bitter Lesson in the wake of GPT-5.
December 1, 2025 at 5:32 PM
There's nothing wrong with bitter things. bsky.app/profile/been...
The problem is that ACs can't do their job effectively at the scale where the AI conferences operate.
December 1, 2025 at 2:51 PM
The problem is that ACs can't do their job effectively at the scale where the AI conferences operate.
FWIW, my secret to positive academic community has been small workshops. I have not found a solution that scales. Though I'm not allowed to suggest this as a computer scientist, it might not scale!
December 1, 2025 at 1:53 PM
FWIW, my secret to positive academic community has been small workshops. I have not found a solution that scales. Though I'm not allowed to suggest this as a computer scientist, it might not scale!
As Christian warned you, I will now hold a grudge against you forever because you posted this negative comment under your real name.
December 1, 2025 at 1:48 PM
As Christian warned you, I will now hold a grudge against you forever because you posted this negative comment under your real name.
LOL, a mere 156.25 passer rating.
November 30, 2025 at 9:56 PM
LOL, a mere 156.25 passer rating.
Certainly an ebb and flow here. Also, nuclear weapons are a great example of how science is often best *confirmed* by engineering. There is probably no more convincing confirmation of the special theory of relativity.
November 30, 2025 at 4:07 PM
Certainly an ebb and flow here. Also, nuclear weapons are a great example of how science is often best *confirmed* by engineering. There is probably no more convincing confirmation of the special theory of relativity.
Neither of you has advanced an argument for why anonymity prevents the thing you are trying to stop.
November 30, 2025 at 3:53 PM
Neither of you has advanced an argument for why anonymity prevents the thing you are trying to stop.
I've found DR super disappointing, especially given the hype. I have never found that the results get you much beyond what you would get if you add "cite your sources" to a standard prompt. But, as always, this might be a skill issue.
November 30, 2025 at 3:52 PM
I've found DR super disappointing, especially given the hype. I have never found that the results get you much beyond what you would get if you add "cite your sources" to a standard prompt. But, as always, this might be a skill issue.
I don't understand your metaphor at all. In what way is vaccination like anonymity in peer review?
November 30, 2025 at 3:50 PM
I don't understand your metaphor at all. In what way is vaccination like anonymity in peer review?
In what way is anonymity in peer review like firefighting? It's as if firefighters were arsonists.
November 30, 2025 at 3:49 PM
In what way is anonymity in peer review like firefighting? It's as if firefighters were arsonists.
Perhaps, but I wonder whether these forces were ever truly kept at bay. One could view all post-Enlightenment science as a post-hoc rationalization of capitalist technology.
November 30, 2025 at 3:47 PM
Perhaps, but I wonder whether these forces were ever truly kept at bay. One could view all post-Enlightenment science as a post-hoc rationalization of capitalist technology.
That's amazing. There's no accountability for bad behavior?
November 30, 2025 at 3:34 PM
That's amazing. There's no accountability for bad behavior?
(2) and (3) are happening whether or not there is anonymity. cacm.acm.org/opinion/coll...
Collusion Rings Threaten the Integrity of Computer Science Research – Communications of the ACM
cacm.acm.org
November 30, 2025 at 3:32 PM
(2) and (3) are happening whether or not there is anonymity. cacm.acm.org/opinion/coll...
Which paper mill journals have no peer review?
November 30, 2025 at 3:29 PM
Which paper mill journals have no peer review?
Come on, man. There is zero evidence that peer review catches errors.
I'm saying that the arguments for blind peer review are all about how the less powerful must be empowered to critique the powerful. I find this argument highly specious.
I'm saying that the arguments for blind peer review are all about how the less powerful must be empowered to critique the powerful. I find this argument highly specious.
November 30, 2025 at 3:22 PM
Come on, man. There is zero evidence that peer review catches errors.
I'm saying that the arguments for blind peer review are all about how the less powerful must be empowered to critique the powerful. I find this argument highly specious.
I'm saying that the arguments for blind peer review are all about how the less powerful must be empowered to critique the powerful. I find this argument highly specious.
This is the same argument.
November 30, 2025 at 3:12 PM
This is the same argument.
I don't understand your argument here. Why are collusion rings more probable in an open review than in a blind review?
November 30, 2025 at 3:11 PM
I don't understand your argument here. Why are collusion rings more probable in an open review than in a blind review?
Is it better to tie yourself to a coarse incorrect model physics with many degrees of freedom, or to embrace pure inductive prediction with a more flexible method for interpolation? 4/4
November 30, 2025 at 3:10 PM
Is it better to tie yourself to a coarse incorrect model physics with many degrees of freedom, or to embrace pure inductive prediction with a more flexible method for interpolation? 4/4
So you tie yourself to an incorrect model with many degrees of freedom, and feel like your results are interpretable. But you fit many parameters to your data already, thus embracing an illusion of interpretation. 3/x
November 30, 2025 at 3:10 PM
So you tie yourself to an incorrect model with many degrees of freedom, and feel like your results are interpretable. But you fit many parameters to your data already, thus embracing an illusion of interpretation. 3/x
I've seen plenty of applications in practice where you can write down a beautiful physics model and find you have a huge set of model parameters you need to tune to get the simulation correct. But these parameters are no different than the weights of a neural network. 2/x
November 30, 2025 at 3:08 PM
I've seen plenty of applications in practice where you can write down a beautiful physics model and find you have a huge set of model parameters you need to tune to get the simulation correct. But these parameters are no different than the weights of a neural network. 2/x
Physics oversells its inability to supervene reality! The laws of physics are amazing, but they work in surprisingly narrow contexts. (I'm mostly paraphrasing Nancy Cartwright here.) 1/x
November 30, 2025 at 3:08 PM
Physics oversells its inability to supervene reality! The laws of physics are amazing, but they work in surprisingly narrow contexts. (I'm mostly paraphrasing Nancy Cartwright here.) 1/x
This take is far too sanguine about the fidelity, interpretability, and applicability of physics simulators.
November 30, 2025 at 2:54 PM
This take is far too sanguine about the fidelity, interpretability, and applicability of physics simulators.
Yes, I am solidly on that bandwagon!
November 29, 2025 at 3:04 PM
Yes, I am solidly on that bandwagon!