bglazer

22 ms·

241.
by bglazer 3y ago
Off topic, but ChatGPT was able to correctly explain the joke, which I think is pretty crazy
242.
by bglazer 3y ago
I’m a scientist, and I use LLM’s fairly frequently. The first and primary use is for coding. Copilot is quite helpful and automates writing a bunch of the tedious steps in data wrangling and complex plot generation. I have also been using p
243.
by bglazer 3y ago
Yes this is borderline incoherent
244.
by bglazer 3y ago
You should read the actual paper before dismissing it. Fig 6 and the surrounding discussion shows that their system is able to iteratively optimize a reaction by adding steps and changing reaction conditions. Based on how poorly the model d
245.
by bglazer 3y ago
I’m not necessarily disagreeing, but I remember word2vec being a big deal when it was introduced. In particular the “king-man=queen” example was mind blowing to everyone I talked to and it was widely referenced as illustrating the power of
246.
by bglazer 3y ago
Why did we stop?
247.
by bglazer 3y ago
I would be pretty surprised if Deepmind could automatically identify drugs that pass phase 3 trials in the next five years. First, there’s a banal point that many trials take years to read out, so any prospective study would have to be begi
248.
by bglazer 3y ago
I’m a graduate student doing computational biology. I wonder sometimes what cryo-EM will look like in the future. Do you think it will ever be possible to get a cryo-EM image of a volume of tissue and get a count/location of all the pr
249.
by bglazer 3y ago
I think this essay is partly a simple longing for the past, which is understandable (if a bit boring), and partly a quite imprecise rant against shitty incentives worsening our experience with technology. Theres much to be said (and has alr
250.
by bglazer 3y ago
NP-complete just means that it’s very difficult to verify that any solution is the best possible solution. It’s very easy to compare any two attempts at a solution. That’s what the researchers did, they saw that the knapsacks from the drugg
251.
by bglazer 3y ago
I get the impression this model is meant for very localized, short term prediction. Like whether or not its going to be raining in your neighborhood in the next hour.
252.
by bglazer 3y ago
Total bullshit. I looked at the corresponding paper and it contains virtually no actual mathematics, mostly just incoherent pseudo-philosophical rambling, including this insightful observation: “Many different persons looking at our physica
253.
by bglazer 3y ago
Honestly, I don’t think we have metaphors or figurative language that really captures how biological systems work. That said, I think the most faithful description is something like a dynamical system with incredibly structured and robust e
254.
by bglazer 3y ago
Overall five year survival rates for cancer have risen from about 50% in 1975 to about 75% currently. That includes several extremely difficult subtypes of cancer, which means that many types of cancer have improved much more than the overa
255.
by bglazer 3y ago
I'm glad to hear that, I hope you stick to those principles if the police or military come calling. They have deep pockets and your investors will want their 100x.
256.
by bglazer 3y ago
Do you have any ethical concerns about the use of your technology for surveillance of civilian populations? That’s the obvious use case for this technology, right? Cheaper and more easily retargetable than satellites. Longer duration, quiet
257.
by bglazer 3y ago
You’re describing human communication as “nothing special”, just a string of patterns that have no significance beyond their low entropy. Frankly that’s solipsistic, bordering on pure nihilism. When I read another person’s writing or talk t
258.
by bglazer 3y ago
Even if there weren’t stylistic and logical problems with the story, I don’t think I’d be terribly interested in reading it. Fundamentally, literature is about communication with other people, living in another person’s mental world or unde
259.
by bglazer 3y ago
I’m sure there are plenty of places that would be happy to have people with experience running life or death operations involving the most complex infrastructure imaginable. The fact is that defense contractors are not employing them for th
260.
by bglazer 3y ago
It’s an appealing vision. Distributing the early review process across time and the whole community would solve a lot of problems with the current model. I do suspect that this would strengthen the winner-takes-all dynamics in academic publ
261.
by bglazer 3y ago
“We’ll use it to find people in building rubble” really begs the question of why the building is rubble in the first place
262.
by bglazer 3y ago
I quit after Musk personally intervened to reinstate a user that posted CSAM. The offending material remained online, undetected by automated systems for many many hours, receiving hundreds of thousands of impressions. This strongly indicat
263.
by bglazer 3y ago
This also ignores the effect of the microbiome, which metabolizes many drug compounds. Also how pre-existing conditions effect drug metabolism
264.
by bglazer 3y ago
If you “trust nobody” and you aren’t immediately and cripplingly paralyzed by the task of rigorously verifying every piece of information you rely on to live, then you’re not being serious. How do you trust that your car, electrical wiring,
265.
by bglazer 3y ago
What do you trust instead? Like, how do form a conception of reality based on events you don’t directly observe?
266.
by bglazer 3y ago
Yes, if you put the “reject all”, button behind a “customize cookie settings” click, you’re a bad, anti-social person.
267.
by bglazer 3y ago
The obvious followup question to this line of questioning is whether your proposed mechanism is robust enough to cause a 62% year over year increase. I don’t doubt that the spike protein could damage islet cells, frankly I’m not nearly expe
268.
by bglazer 3y ago
He won the fields medal, which is the effectively the Nobel prize for math, and has been cited nearly 100k times. Academically, these are extremely high level achievements. He’s contributed across very diverse fields of math, so it’s diffic
269.
by bglazer 3y ago
I downvoted this because you’re making a interesting claim but it’s frankly irresponsible to not even try to source it.
270.
by bglazer 3y ago
Thanks for summarizing this work so well. You mention TEM of the glutamatergic vesicles in the astrocytes. Can TEM give any information about the vesicles contents? Or can it be inferred from the imaging somehow? Just curious how one would

More ›