HN user
Scientist/programmer.
A number of startups are working in verifiable domains where they can provide realistic data. This is an interesting thread from one of those startups: https://x.com/khoomeik/status/1973056771515138175
Here's a discussion with Isomorphic Labs (Google DeepMind spinoff) on this line of thinking: https://www.youtube.com/watch?v=XpIMuCeEtSk
Aside from the better versions of what AI is visibly doing now (software dev, human language translation, video gen, etc), many of the AI bears are dismissing the potential impact of hooking AI up with automated experimentation so it's able to generate new types of data to train itself. The impact on drug discovery, material science, and other domains are likely to be very significant. The Nobel Prize in Chemistry for AlphaFold is just a glimpse of this future.
Nice collection. I'd suggest adding an unsubscribe option to your initial email, particularly if people reflexively login via Google, etc.
Very nice. I'd suggest adding another deluxe bundle for non-Guitarists without the guitar theory. I'd pay extra for the ear training + the base package.
Packt is hit & miss but some have decent authors and good reviews.
"Machine Learning with PyTorch and Scikit-Learn" by Sebastian Raschka et al.
"In-Memory Analytics with Apache Arrow" by Matthew Topol.
65 is the new 45 :) Sounds like there's a perfectly logical reason to do the 3rd startup: you're passionate about it. Best of luck!
He's a professor at Caltech and did Nobel laureate-level work. He might be less productive than his prime and still be competitive with other professors, particularly since he brings deep understanding of the history of approaches in his field.
Then maybe you shouldn’t generalize your one viewpoint to say that early Ruby/Rails was not a worthy community for professionals or of use to people who practice jointly valued skills like entrepreneurship.
It’s also odd to see your long rant of whimsical vs professional when some of the most well-known companies were built with that community of so-called non-professionals. We have a difference in opinion on what constitutes “professional”.
I think you're missing my point. The early Ruby and Rails community I remember was a collection of very smart and explorative programmers who wanted to build cool stuff with this interesting language. People were trying out DSLs -- sure they could've used LISP -- but Ruby's metaprogramming was inviting and was a reason for the succinct Rails syntax which was a selling point compared to say Java's cumbersome approach.
The speed of trying stuff out (even if it wasn't super efficient) why startups used it. So it was a community of highly productive people sharing their love of building new things. That's my memory of that time period.
I would much rather participate in a community of professionals who've organized themselves around sufficiently overlapping shared intents
I thoroughly enjoyed my involvement with early (US) Ruby and Rails folks from the first Rails conf to _why's unusual entertainment to Matz's calm and humble demeanor. People bounced ideas off each other and just enjoyed coding up interesting things. Dave Thomas and the Pragmatic Programmer group wrote what many of us used, not so much _why's guide which was still a fun read. I moderated a Ruby panel at the old Odeo HQ just before they pivoted. I didn't know the group gathering at that Ruby SF meeting would include not only Twitter but Github founders as well. At the time, tweets seemed pretty absurd to some of us but guess what happens when you try out ideas in a community that was into exploration?
Yes, you are correct that I should’ve said symptomatic infection (or COVID-19) because they were only trying to power the study for what they needed to clear EUA and not measuring ability to block transmission. The second paper I linked made the same mistake in its table header.
Agreed. His first position on lockdowns is a GBD talking point that is strongly opposed by Sweden vs Norway through mid-2021 (the end of major vaccine rollout) when they had different positions on lockdown but shared many other characteristics. Sweden had 10x more deaths during that time period, about 1200 extra deaths per million people. It also completely ignores the morbidity introduced by all the extra cases while focusing on non-COVID morbidity.
That vaccines never did anything to prevent transmission is a pretty ridiculous contention for the initial wave.
That vaccines posed more of a danger to most of the population compared to COVID-19 itself is also ridiculous. The US had the largest vaccination drive in its history in early 2021, peaking at 3.5 million jabs per day in April. Excess death dropped dramatically throughout the rollout and kept falling past mid-2021.
Of course it was tested for preventing infection. It was tested for that as well as efficacy against severe disease: "Efficacy against laboratory-confirmed COVID-19 (with an onset of >= 7 days after receipt of the second dose) and against severe COVID-19.. was assessed among the participants 12 years of age or older." https://www.nejm.org/doi/10.1056/NEJMoa2110345
You seem to be conflating the vaccine having efficacy against infection and whether a vaccinated individual who still catches COVID-19 (since protection against infection is obviously not 100%) can transmit it. The efficacy against original COVID-19 strain infection was high (91.3% for the Pfizer vaccine).
Your first link speaks to the situation of a breakthrough infection being further transmitted. Since it wasn't clear HOW MUCH it curbed infection (under non-ideal/trial conditions), quarantines and caution were reasonable.
Here's a chart that showed the original test results across a number of vaccines. The efficacy vs infection dropped quite a bit with variants though the efficacy vs severe disease holds up better:
https://jbiomedsci.biomedcentral.com/articles/10.1186/s12929...
I’d love “informed voters” though from what I see of social media, it does anything but inform through its creation of cliques and upvoting based on popularity, not veracity or even reasoned argument.
One example is the switch to paid blue checkmarks on Twitter. Prior to the switch, there’d be some semblance (not great) of debate on comments. Now, all the musk fans and RW-oriented subscribers completely dominate so all you see in response to a Dem or “RINO” tweet is reflexive comments calling the OP lies, incorrect whataboutism, conspiracy theory of the day, etc, with more reasoned fact checking buried way down.
This is particularly pronounced with any tweet about the Trump indictment with little discussion of the substance of the indictment. The crackpot hypothesis/meme of Paul Pelosi’s attack being a gay love spat is another example. The participants didn’t even want to read the police report before moving on to another conspiracy theory about the report’s generation. Reasoned discussion would be great. It’s just not happening.
Translation: I was wrong but I'll pretend I'm right.
Extend the timeline out as much as you want. Sweden still has a lot more cumulative dead per capita. And you are simply lying that Norway had increased fatalities and so just shifted it in time. The gap between Sweden and Norway slightly increased by May 2023.
Spare me your opinions and standard talking points after seeing data contradicting Sweden was some shining star of stellar policies due to GBD principles. A lot of our "leaders" like DeSantis forgot that pretending your expert in things you know nothing about and finding the most partisan hack doc you can for Surgeon General doesn't mean you're making correct decisions.
We could also talk about morbidity but that would make your weak argument even worse. After all, quality of life is SO important during the limited lockdown times but let's not measure significant decreased qualify of life for millions of Americans from Long COVID. And lets all hope there isn't significant subclinical issues like increased thromboembolic activity (frequently seen post-COVID) that will rear its head later.
Sweden does stand out compared to their Nordic neighbors. They had ~1200 more deaths per million than Norway & Finland in the first year of pandemic. We're so inured to COVID death once we hit a million that some multiple of 9/11 doesn't stand out, eh?
https://tinyurl.com/sweden-vs-nordic-covid
https://tinyurl.com/sweden-vs-nordic-cumulative
Glad to hear we've moved from me being "badly misinformed" (now that you took a second to look at the actual data) to now having a "myopic fixation on covid".
Speaking of myopic, please do talk to the families of the ~1200 additional deaths per million people in Sweden and tell them their tragedy was "only visible if you looked at it with a microscope."
Sweden had a 10x death rate for the first year of the pandemic vs its direct neighbors. That's the data published by Sweden, Norway, and Finland. You can ignore it if you want or just claim it's "highly politically motivated and extremely biased" because it's against what you want to believe. All Nordic countries did relatively well on death rate, that's why GBD have seized upon Sweden as a success story, but it was obviously worse than Norway and Finland.
When I talk with GBD proponents it always boils down to this: they won't believe the data and make up a variety of reasons why not to trust it, then they turn around and trust far worse analyses like the article comparing Sweden and non-Nordic countries with much worse obesity and far different systems (like the article comparing Sweden vs other EU).
Feel free to construct a competing comparison showing how the 10x death rate of Sweden is just an artifact and not a failure of public health. Or are you going to claim they're just old people so no big deal?
I think Jay B was better than most of the GBD supporters because he seemed to focus more on the targeted prevention. The problem was he didn't counter the downplaying of SARS-CoV-2 by people using his own seroprevalence paper (which had problems), or at least I never saw him push back on Atlas and others. So his criticism about public health issues is completely devoid of his own part in the story. They went heavy with part 1 (COVID isn't that bad for non-elderly) and were surprised that downplaying the impact didn't shift resources to the elderly. If people using his data were saying the overall IFR was like the flu, why should the govt go way into debt protecting the elderly?
"the 3 Covid vaccines all failed to show any actual evidence of protection or immunity"
You must be explicitly ignoring a ton of papers and data showing the opposite. The death rate of unvaxxed vs vaxxed is incredibly higher and CDC has data for it.
"I have come to the conclusion the vaccines saved virtually no one and has put a similar number into an early death (e g. The clot shot)."
In early 2021, the US had the largest vaccination campaign ever peaking at about 3.5 million shots per day in mid-April. Do you know what happened to the death rate (excess mortality) in 2021 through June? It dropped substantially. Isn't it odd to you that mass vaccination using a "clot shot" somehow drops the death rate substantially? It's rhetorical because to form the opinions you have, you must go out of your way to cherry pick papers and ignore data (or have conspiracy theories that data contradicting your opinion is fake).
There's the GBD proposal, and then there's the actual media tours by the GBD proponents, particularly on Fox, where COVID-19 was downplayed. Some of them famously wrote columns how there'd be maybe 10,000 fatalities and it was like the flu. Is it surprising that downplaying the virus also worked against spending huge amounts of money to protect the elderly? Like govt renting hotels to isolate elderly living with families? Or providing deliveries so they could stay at home?
Compare Sweden vs "their neighbors" as I said above. They had much worse COVID death rate from their Nordic neighbors (Norway, Denmark, Finland) through initial COVID and the vaccine rollout to summer 2021.
Go to https://ourworldindata.org/ and check Sweden vs Norway, Finland through the vaccine rollout.
https://tinyurl.com/sweden-vs-nordic-covid
https://tinyurl.com/sweden-vs-nordic-cumulative
Completely correct? GBD advocated letting SARS-CoV-2 rip through the general pop BEFORE vaccines while "targeting" protection of the elderly, which of course no country managed to do, particularly their poster-child Sweden. Now GBD fans pretend Sweden was a success story when any observation of the COVID impact on Sweden vs their neighbors through the vaccine rollout shows a completely different story. (Sweden had ~10x death rate of bordering Norway for first year of pandemic.)
Also GBD pretends mortality is the only worthy metric. There's such a thing as morbidity as the millions of people with Long COVID would like to discuss.
But Bhattacharya and friends with strong connection to Hoover Inst have a particularly nice revisionist view of their early work:
* Go on Fox and other media to downplay SARS-CoV-2 before we even know what's going on.
* Write seroprevalence papers with suspect specificity (changed between v1 and v2 of pre-print due to outrage from stat experts) that don't describe how the visual tests were interpreted, why they differ from the CZI BioHub paper's measurement of the Premiere test, or why a pathologist running ELISA tests to verify wanted off the paper.
* Stand by while friends like Scott Atlas make a media tour saying the seroprevalence paper suggests as many as 10,000 dead from COVID. Thus a fair portion of our population decides COVID is just the flu, and some of the GBD supporters still think it's "just a cold" despite the incredible evidence we have that it's much much worse (for example, see Iwasaki's work) and there's the little matter of over 1 million Americans dead.
* Then write the GBD before we have the vaccine or have any real public effort to try to protect the elderly.
On the bright side, Bhattacharya has been a proponent of the vax and many people (including me) agreed that targeted protection of elderly was very important. It's just that the revisionist commentary from his supporters is they were the only ones saying it. Completely false.
Agree. They should’ve allowed for more discussion before sending invitations and more effort should've been made to come to a mutually beneficial solution. But going from an invited keynote (great honor) to an invited talk (honor) is not what I would call a deliberate attempt to "disgrace one of the experts in my field." It’s a single track conference where the speaker would be heard by all attendees and could’ve championed his ideas to the larger Rust community.
What does "last minute" mean here? Isn't RustConf in mid-September and it's a single track conference so speakers are heard by all participants? It would be a good way to get your work and opinions out to the broader community.
Clearly the process was botched and more effort should've been made to come to a mutually beneficial solution, but going from an invited keynote (great honor) to an invited talk (honor) is not what I would call a deliberate attempt to "disgrace one of the experts in my field."
I thought this was a good overview of the idea Triton can circumvent the CUDA moat: https://www.semianalysis.com/p/nvidiaopenaitritonpytorch
It also looks like they added MLIR backend to Triton though I wonder if Mojo has advantages since it was designed with MLIR in mind? https://github.com/openai/triton/pull/1004
Maybe you are purposely ignoring many data points of how badly things are going. Musk instituted blue checkmarks that are purchased by a subset of customers with particular views. If you question this, look at the responses to Biden tweets and see the positions of anti- vs pro-Biden comments.
Because of this, Twitter is ridiculously partisan on my feed. I’ve limited my consumption of it to very small subject matters and will likely jump ship totally to BlueSky with some Mastodon thrown in. I know a lot of people in science who have completely abandoned Twitter.
If you don't need range scans, have you looked at WiscKey approaches? https://www.usenix.org/system/files/conference/fast16/fast16...
There are a number of implementations including Badger (used in dgraph) and a variant that's "RocksDB for large-value use cases" (https://rocksdb.org/blog/2021/05/26/integrated-blob-db.html)
It removes all default browser styling for hyperlinks. It wasn't clear from the docs that it would do that so I'm glad the fix came up: https://stackoverflow.com/questions/63883580/tailwind-css-ho...