mchinen

7 ms·

61.
by mchinen 2y ago
Most of the work on objective quality metrics (e.g. PESQ, POLQA, ViSQOL, DNS-MOS, NISQA) focus on speech because of telecommunications demands, but some of these have an audio mode. But there are some new promising audio ones that are ML b
62.
by mchinen 2y ago
I'm also ex-Google working in the space on the funding and company formation side. Would love to chat.
63.
by mchinen 2y ago
I enjoyed this. I wished a bit that the website's "Increased Lifespan (source)" also claimed a 50% lifespan increase from the linked SA blog and associative study that showed a 50% likelihood of survival increase over a given
64.
by mchinen 2y ago
I also think the bottleneck isn't access to the papers today and data access and silos are more important. But I also think the GP's claim and yours are not incompatible. I wonder how much survivorship bias this has since it only
65.
by mchinen 2y ago
I'm assuming the mushroom in question is agaricus biosporus in white button form. Interesting to hear the 3 star Michelin chef say there is a huge difference between paris mushrooms and industrial mushrooms of the same species due to
66.
by mchinen 2y ago
Nice to read this. Curious how CORDIC performs against cubic or other polynomial interpolation with a small table. I was taught that resource constrained synthesizers sometimes used cubic interpolation, and this was presumably when CORDIC
67.
by mchinen 2y ago
I am trying to understand how accurate the docking predictions are. Looking at the PoseBusters paper [1] they mention, they say they are 50% more accurate than traditional methods. DiffDock, which is the best DL based systems gets 30-70% de
68.
by mchinen 2y ago
So this means there is only one agent with multiple accounts running the same that has a bug to use the same seed every time (possibly seeded on a hash of the content if we are generous)? edit: or greedy sampling only?
69.
by mchinen 3y ago
https://vitamindforall.org/letter.html is a letter that goes into why >= 4000 IU daily is recommended based on blood serum levels with more literature review.
70.
by mchinen 3y ago
This is fascinating to watch. I understand that they get the proteins from PDB/ChimeraX, but how much manual process is involved to map and place the individual proteins? The paper says it gets the protein locations from CryoET tomogr
71.
by mchinen 3y ago
It makes sense. I should say I'd estimate 70-90% of the class didn't want to work their own ideas, but would be happy to work at a big company forever, and this was the mindset going into the CS program. A fair number of people
72.
by mchinen 3y ago
It's interesting to hear your experience at this point in time. Mine was later but noticeably different. I started coding in the 90's as a teen releasing shareware, with my first gig at Adobe in 2001, where they paid me $14/
73.
by mchinen 3y ago
From the 5% reduction it seems (1) was less likely. To bobfunk, the response needs more empathy and explanation around the obvious frustration around why there is no slider for cost limitation. As it is, it feels like the minimum viable cor
74.
by mchinen 3y ago
There is a sort of nihilism that emerges when you look deeper at counterculture and it's conformity. Neo recognizing with horror the other neos is a great example. (In that case it only worked out for him because of hollywood's
75.
by mchinen 3y ago
Interesting, I have had a different experience with Gemma 7B in being able to be generally useful for basic questions after a couple prompts. It frequently gets stuck and starts outputting grammatically incorrect sentences and answers unrel
76.
by mchinen 3y ago
Reposting comment from last time since I'm still curious: This is really fun to see. I love toy experiments like this. I see that each plot is always using the same initialization of weights, which presumably makes it possible to have
77.
Trying Out Stable Cascade Local Image Generation(michaelchinen.com)
2 points by mchinen 3y ago|0 comments
78.
by mchinen 3y ago
Given that you can pick llama or mistral in the NVIDIA interface, I'm curious if this is built around ollama or reimplementing something similar. The file and URL retrieval is a nice addition in any case.
79.
by mchinen 3y ago
This is really fun to see. I love toy experiments like this. I see that each plot is always using the same initialization of weights, which presumably makes it possible to have more smoothness between each pixel. I also would guess it&#x
80.
by mchinen 7y ago
Seems reasonable, from what is known about the capacity of speech, although extra cool if there is really some universal number we converge on. Kuyk and Kleijn found the upper bound of the speech channel to be 100 bits per second, in agree