arjmandi
Joined 10 years ago
Karma
86
Submissions
37
Member since
Mar 28
Mohsen Arjmandi
https://freddiespirit.com
mohsen.arjmandi[at]gmail
View on Hacker NewsRecent submissions
Comment · 11 days ago
I wrapped a human-in-the-loop RSI loop around ASSAY (my open-source reasoning harness) and ran it on the 25 public ARC-AGI-3 games with Haiku 5.5. Assay also introduces 5 levels of abstraction for reasoning and has equivalent plugs for each layer. This design was used to guide the analyst model in this experiment to recommend fixes at the layer where the reasoning broke, and the harness runs again.
Comment · 4 months ago
LLM inference performance is governed by three competing bottlenecks: compute time, memory bandwidth, and communication latency. In this post, we've covered what allows full hardware utilization and key constraints.
Comment · 3 years ago
Most of the comments were about either piracy is harmful or not. Another point here is that, for me, it's natural that the EU rejects or neglects a comprehensive report because it contradicts previous findings, especially in cases like piracy that have innate harmful characteristics.
Comment · 4 years ago
He had a talk three days ago, explaining his thesis where he remarked:
“When the paper was posted online just a few days ago, many people who don’t focus on mathematics didn’t understand it, thinking that it was the Landau-Siegel zeros conjecture solved, and some even thought that it proved the Riemann Hypothesis is wrong. Actually, I don’t have this ability. I only partially solve the Riemann hypothesis within a certain range. If I say I overturned Riemann Hypothesis, few people would believe it.”[1]
Maybe he loves what he's doing and that's the root of being stubborn and "genuinely oblivious to ordinary material feedback". Although love or passion can be overrated or too general to describe his attitude toward problem-solving, I think people can't be just stubborn, there's a drive that holds them to a higher standard.
[1]https://pandaily.com/mathematician-yitang-zhang-confirms-par...
Comment · 6 years ago
https://t.co/EuRuVazuKk
Erik Hoel in this paper offers an audacious hypothesis: Our brain, during the its evolution, has developed dreams as a way to solve over-fitting
Since we’re learning from a limited samples of data in the real world, chances of overfitting (I call it judgement) goes higher. In ML we inject randomness and noise to avoid overfitting. Hoel theory can explain why our dreams are so sparse & hallucinatory
Comment · 7 years ago
That's the dream, all we need right now is support from the tech community to come to an agreement for opening new windows for innocent people who have contributed a lot to science, tech, art and culture of the entire world.
They can make a change, a real difference