dekhn

18 ms·

421.
by dekhn 10mo ago
I've been inside the conference- I used to do due diligence and discovery for Google Ventures and they gave me a ticket one year. The "talks" were eminently forgettable ("we put this in X people and Y died") and th
422.
by dekhn 10mo ago
23&Me data is simply not accurate enough to make reasonable predictions about health outcomes. I have a whole genome and nothing Google has built has been able to do anything useful with it, medically speaking. I could use DeepVariant
423.
by dekhn 10mo ago
I expect any large-scale learning system to use huge amounts of memorization and to exploit that information at inference time.
424.
by dekhn 10mo ago
typically around one micron (which is about the same as one pixel with my objective and camera). I am not trying to take min steps. I'd rather have smooth motion, with fairly high accuracy. The current system I am building is mainly
425.
by dekhn 10mo ago
I don't particularly like the openflexure approach. Instead, I build stages similar to standard XY stages: https://www.asiimaging.com/products/stages/xy-inverted-stage... which uses standard motion component
426.
by dekhn 10mo ago
But what do you down with the Thorlabs Lab Snacks? I thought that was the main reason grad students ordered things from them.
427.
by dekhn 10mo ago
I haven't read the post yet because my work blocks access to the domain. I agree there is a huge middle ground- for example, I make hobby microscopes at home, and much of my work has been making accurate and precise 2D/3D stages.
428.
by dekhn 10mo ago
With products from a vendor like Thor Labs, you're getting a lot of quality and knowledge built into the system. Mechanical engineers, electrical engineers, optical engineers... all of which means an edu kit like that will train a stu
429.
by dekhn 10mo ago
I think this makes more sense: "When God said: "Let there be light!", Jeff Dean said "LGTM, Approve"
430.
by dekhn 10mo ago
At the time I first got involved, Google Health was still a thing but it was clear it was not going to be successful. I felt that Google's ML (even early on, they had tons of ML, just most of it wasn't known externally) was going
431.
by dekhn 10mo ago
Peter worked on search, yes. I think he mainly focused on quality. See his comment on "Managing Gigabytes" ( https://www.amazon.com/review/R33MIMY7A7C2H8/ref=cm_cr_srp_d... ) which was an early reference
432.
by dekhn 10mo ago
The one thing I noticed when I worked near Jeff and Sanjay and talked to them over coffee is that Jeff is the smart one, but Sanjay is the wise one. Jeff always had an idea how to make something a bit faster using a clever trick, but Sanjay
433.
by dekhn 10mo ago
I really want to see a hybrid miata. the miata was a fun car but really lacked acceleration. A good new design would make it both fun to drive, and obscenely good at acceleration.
434.
by dekhn 10mo ago
Jeff is so good at AI, he invented much of the infrastructure to implement LLMs, mostly to increase the productivity of other programmers.
435.
by dekhn 10mo ago
Yes, but the downside is even highly accomplished engineers feel unworthy. For example, when I started in 2008, they said everybody should make a "Google Resume" (ongoing list of all the stuff you did at Google) and linked to Jeff
436.
by dekhn 10mo ago
Yes, Jeff told me this one personally. Knuth came in right before the talk started and the room was full. I think somebody later gave up a seat for him.
437.
by dekhn 10mo ago
Garbage article.
438.
by dekhn 10mo ago
The onus is on you to test this. Come up with a physical experiment (not math) that physicists can understand, in which your theory predicts the outcome accurately, while existing theories cannot.
439.
by dekhn 10mo ago
This article is basically cargo cult biotech.
440.
by dekhn 10mo ago
The predispositions you describe are on the order of 1% which would not explain parents having (say) 5 all boys.
441.
by dekhn 10mo ago
The DMV is not a private company with enormous amounts of fraud/scam. Anyway, it's what I was told when I joined Google Ads a long time ago and it seems consistent with their philosophy and behavior.
442.
by dekhn 10mo ago
Of course. All successful large-scale political movements have underlying narratives that only loosely correlate with reality.
443.
by dekhn 10mo ago
Are you saying that children's sex is heritable? I don't think there is any strong evidence for that (just weak evidence of tiny effects). Or are you just saying that for any 50/50 distribution, there will be some outliers
444.
by dekhn 10mo ago
Google won't tell you this because they believe it would reveal information to scammers.
445.
by dekhn 10mo ago
It looks like he got the problem resolved by providing a spreadsheet with copyright violations and his original sources. THink of Google like how Cantrill described Ellison: You need to think of Larry Ellison the way you think of a lawnmowe
446.
by dekhn 10mo ago
hire a lawyer, they know how to communicate with these teams. Trying to deal with Google Customer service when you are a technologically savvy and logical individual won't get far: GOogle's default attitude towards anybody who con
447.
by dekhn 10mo ago
Why do you care about this so much?
448.
by dekhn 10mo ago
More details required. Like the other comment says, this can be done with a beam-expanding telescope (which can be as simple as two lenses whose focal length ratios determine the magnification). https://www.newport.com/n&#x
449.
by dekhn 10mo ago
Ah, I only noticed the "shell startup" bit. Yes, after 2-3 I agree you'd start to notice if you were really fast. I suppose at that point I'd just have Gemini rewrite the prompt-building commands in Rust (it's quit
450.
by dekhn 10mo ago
That doc predates 2012 significantly. From what I've been able to glean, it was basically created in the first few years Jeff worked at Google, on indexing and serving for the original search engine. For example, the comparison of cac

More ›