Viewing profile — CompleteSkeptic

constrained decoding tends to make models dumber - this is why it's rarely used

I helped train some of the first "magic" models at OpenAI[1] and it was a wild ride. We were a pretty sane + skeptical team and we weren't totally convinced the models were as gene…

TypeSafe AI | Software Engineer | ONSITE (San Francisco) | Full-time TypeSafe AI ( https://typesafe.ai/ ) is a well-funded stealth mode company making a new class of AI models whic…

Is this the first time OpenAI has published comparisons to other labs? Seems so to me - see GPT-5.4[1] and 5.2[2] announcements. Might be an tacit admission of being behind. [1] ht…

There actually is some work ( https://arxiv.org/abs/2003.13630 ) claiming that FLOPS are a poor measure of real-world performance - with some of the more recent FLOP-efficient mode…

This isn't directly relevant to PP-YOLO, but I'm surprised roboflow is still promoting "YOLOv5" - despite that model not having an associated paper and it not being made by the aut…

An appropriate quote: "If you can't intelligently argue for both sides of an issue, you don't understand the issue well enough to argue for either." There are many people for whom …

I think he was referring to the fact that the request completes in finite time isn't a tight enough bound.

State-of-the-art results no longer require unsupervised pretraining with autoencoders or RBMs, but back when unsupervised pretraining was more popular, top researchers were rationa…

You may be interested in DBpedia, which gets its data from wikipedia but presents it in a structured format (for example, the page for Clojure: http://dbpedia.org/describe/?url=htt…

Yup! We were about to implement something like that, but we realized Reagent's model of having atoms store the state worked pretty well with javelin's FRP, and once you break down …

I wouldn't go so far as to say that Reagent's model is perfect. Definitely pretty good, but we've had some issues that were somewhat non-obvious. If you're worried about it being m…

If any of my post sounded as if I was complaining about the lack of syntactic sugar, I apologize as that was not at all my intention. The best practices and syntactic ugliness didn…