"The best case scenario is to exist long enough to get a billion dollar exit"
Sheesh, seems hardly worth it.
3,134 karma · joined June 16, 2015
Sheesh, seems hardly worth it.
>> * Fast hired engineers, engineering managers, product managers, and executives directly from Big Tech. Many of the software engineers joined from Meta, Google, Uber, Amazon, Apple, Microsoft, and other well-known companies. Many joining had competing offers both from Big Tech and other high-growth startups.
So it's not like millionaires with blurring are going to be getting in line. This seems like it's for the totally blind.
Perceive also:
https://secondsight.com/discover-argus/
https://www.marketwatch.com/investing/stock/eyes/financials?...
Who is the better/safer bet to fund: Kariko or Weissman (MD PhD fellowship at the NIH under Anthony Fauci, full professor at UPenn)?
I grew up not far from Galesburg, in a town roughly the same size. Many of the rural subdivision roads outside of downtown are paved with chipseal [1] rather than asphalt concrete or portland cement. While chipseal is certainly not as nice as concrete to drive on, it is much cheaper to maintain. I remember the road in front of my house getting re-treated every few years. Meanwhile the paved blvd connecting all these subdivisions hasn't been maintenanced since the 1980s, and is turning to rubble. So I personally think the problem (at least for small-ish midwestern towns) is the bias of state DOTs towards creating new infrastructure (paving new roadways / bridges) over maintaining the infrastructure they've already built (there's no ribbon-cutting ceremony when you're just filling potholes). Of course, this mentality only exacerbates the problem.
[1] https://en.wikipedia.org/wiki/Road_surface#Bituminous_surfac...
Also if theory is dead why are the scientists they interview in this piece obsessed with generating new theory...
https://cushman.host.dartmouth.edu/books/CommonSenseFM/Chap5...
Here's the trick - if you stimulate a group of neurons, their spontaneous activity will be (briefly) different from an unstimulated group. You can use this fact to make it seem like the NN is learning to play pong.
It would be like claiming a network of cats can learn to play pong. I put cats in two different boxes. To tell the cats whether the ball is above or below the paddle I will spray water into one of the boxes. Then I will listen for hissing from both boxes. Based on where I hear the most hissing is the direction i move the paddle.
He reads, "Step-1: prior to translating the note, we are going to train your neural networks on a Chinese corpus consisting of 410 billion tokens from Common Crawl, 19 billion tokens from WebText, 67 billion tokens from Books, and 3 billion tokens from Wikipedia. Training in 5, 4, 3..."
He shouts, "wait but if you do that wo shuō zhōngwén."
https://en.wikipedia.org/wiki/Spermatogonial_stem_cell
That our germline cells are immortal, yet we are not, suggests that we (our bodies) are merely a replication vessel. Humans don't live to be 150 because (according to natural selection pressures) it's easier to just make a new version of you (or several) than to waste DNA space on efforts to keep 100-year-olds alive. The genome only serves the host through child-bearing years (plus a few extra bc grandma may still be useful to the survival of her grandchildren; what happens to grandma after 90 years old the germline cares not.
It seems like Facebook is really trying to fully rebrand as Meta; even the Facebook 'about' page displays Meta instead of Facebook: https://about.facebook.com
"I would encourage the equal reporting of both positive and null findings. Performance would not in any way be based on positive findings, only that the experiments were carefully run."
Would you have considered the LHC experiments "failures" had we not found evidence of the Higgs Boson? I certainly wouldn't have. What matters most is that important questions backed by sound theoretical reasoning are addressed carefully, collaboratively, openly, and with plenty of long-term support. If we do that, we get genuine answers - such answers represent important information whether or not they support our original hypotheses.
Currently there is way too much trivial shit being published in the ever-expanding number of scientific journals.
Our current model for career advancement in academia is partially to blame for this. Landing a tenure track job requires having a prolific publication record. And once you've landed that coveted ladder-rank position, the pressure to publish only heats up, with tenure on the line. And then even after you've secured tenure, career advancement still depends heavily on the ol' publication record. Pre-tenure the pressure to publish in high-impact journals is immense; it's the kind of pressure that drives otherwise honest people to consider partaking in fraud (and unfortunately those who stick to their morals often lose out to those who fabricate results to some degree). Post-tenure the pressure changes from securing high-impact papers to just getting on whatever papers you can. Review boards for associate professors would more readily give a promotion to someone with 20 meaningless papers over the last three years than someone with 2 papers in CNS journals over the same timeframe, even though a single paper in Cell/Nature/Science is typically more impactful than 100 papers in Frontiers or other similarly dogshit journals. So post-tenure pay raises are based on getting as many words in print as possible, using the least amount of effort to do so.
I almost forgot where I was going with this... right, so, in my opinion we need to switch from a publication-based mindset to a discovery-based mindset. We (the public) provide the NIH with $30 billion dollars per year, with the idea that such an investment will lead to medical breakthroughs, discoveries, and other innovations that can concretely improve our health and wellbeing. However, so much of that is wasted on the idiosyncrasies of career advancement inside the ivory tower.
If I were the director of the NIH my sole purpose would be to end this nonsense. And my first order of action would be no longer accepting grant applications from individual PIs. I would only entertain grant applications from a small force of scientists (4-8 lab equivalents) with thorough and cogent plans for making breakthroughs on cancer or heart disease etc. I would change the minimum R01 funding amount from <$500k to >$10 million dollars. I would change the grant renewal timelines from every year to every 5 years but require yearly progress updates to ensure the proposed experiment were soundly conducted. I would encourage the equal reporting of both positive and null findings. Performance would not in any way be based on positive findings, only that the experiments were carefully run. I would require that all raw data be deposited into publicly accessible repos (not at the end of the study, not yearly, but whenever data is generated it should be made accessible asap). I would encourage that research groups bother with drafting/submitting interim manuscripts (interim meaning prior to completion of the full 5-10 year study) only if they find something important, otherwise just provide a comprehensive writeup at the end of the study. This final writeup would not be submitted to a 3rd party journal. It would be posted directly on the NIH website I'd have created for such results reporting. Naturally this would also be publicly accessible. That would be a start...
edit
@ 08:50 CET and it's up to 30%. She's in the clear.
https://www.nih.gov/health-information/nih-clinical-research...