@tszzl — page 25/103

2020-11-05 → 2020-12-01 · posts 12001–12500 of 51,350
i don't even care who wins i just love elections dont pretend this doesn't get the blood pumping
32 ♥ · 1 RT · x.com →
· ↳ reply to @jachaseyoung
@jachaseyoung never gonna happen his financial incentives would be to never accept the declining accuracy of public polling, because it basically kills his whole platform
2 ♥ · x.com →
@Randylad buddy i got CFB, porn, and election coverage open on 3 screens
3 ♥ · 1 RT · x.com →
i f*cking knew it
115 ♥ · 10 RT · x.com →
@ProfoundProfit universal voter ID system they manage it in countries a lot less developed than ours
8 ♥ · x.com →
@ProfoundProfit voter id would mean 1 person = 1 ballot with strict coupling you can recycle all IDs of dead people
2 ♥ · x.com →
@ProfoundProfit it incrementally adds more transparency i dont know if you can get to completely trustfree elections. probably requires some cryptographic shit
3 ♥ · x.com →
@ProfoundProfit i think we already do this aside from pt 1 and its obviously not ideal especially in pandemic times. they pair up one democrat and one republican to count votes in michigan precincts IIRC
1 ♥ · x.com →
· ↳ reply to @tszzl
‘a statement such as “The polls messed up,” is not just a statement about the polls, it’s a statement about how the polls are interpreted. More realistic modeling of the polls can correct for biases and add uncertainty for the biases we can’t correct.’
3 ♥ · x.com →
the only thing more embarrassing than being a big tech bugman is getting owned by them
18 ♥ · x.com →
· ↳ reply to @tszzl
is this yet another CIA op? i cant even tell
14 ♥ · x.com →
@Randylad dude had no idea the kind of horrors he was bringing into this world
"steelmanning" is a cognitohazard and should never be tried
19 ♥ · x.com →
· ↳ reply to @discourseloverr
@discourseloverr doesn’t mean that everyone is cut out for programming lack of blue collar jobs is a still unsolved problem
6 ♥ · x.com →
@browserdotsys it’s so hard bc the dudes a journalist... everything pisses him off
2 ♥ · x.com →
· ↳ reply to @tszzl
why don’t you just go ahead and apply that to every tweet
10 ♥ · x.com →
· ↳ reply to @TetraspaceWest
@TetraspaceWest we have to trust the prediction market runners to decide this correctly — for the kind of trustless society we are moving to its difficult
1 ♥ · x.com →
· ↳ reply to @suavelambi
@suavelambi ya, the tech companies are purely reactive. they just take the oath of least resistance wrt elections, instead of being proactive and setting standards
monarchy is alive and well inside the chaebols
20 ♥ · x.com →
i bet our nuclear arsenal works about as well as our pandemic response
156 ♥ · 11 RT · x.com →
this imperial mf will be your president in 2040
31 ♥ · x.com →
· ↳ reply to @luke_metro
@luke_metro haha I mean the generalized chaebol, including eg the family run NYtimes
2 ♥ · x.com →
@AOCummies do it and uncover the lovecraftian shit your family is hiding
4 ♥ · x.com →
what ppl say: I don’t see race! their revealed preference: ancestry testing is a billion dollar industry
66 ♥ · 3 RT · x.com →
· ↳ reply to @NemoBane
@NemoBane honestly most of the time I download good memes and forget where I got them from lol Sorry!
2 ♥ · x.com →
@ProfoundProfit completely true, and they'll start admitting it once joe biden is inaugurated
2 ♥ · x.com →
wtf lol
102 ♥ · x.com →
@gigafelon yeah i dunked on his dancing tweet like yesterday lmao didnt think hed have the time to be this petty
4 ♥ · x.com →
does anyone remember what it was like to get work done? not me
84 ♥ · 5 RT · x.com →
@browserdotsys so many eric garland bangers lost in time, like tears in the rain
2 ♥ · x.com →
· ↳ reply to @eigenrobot
@eigenrobot idr but i think you go to a demon's Special Torture Zone til the end of time so i will be answering no
@AOCummies why even live if you can't reply to this?
9 ♥ · 1 RT · x.com →
publication chances increase with both "coolness of the idea" and "improvement in objective metrics" why does it seem like there's an inverse relationship between the two?
9 ♥ · x.com →
· ↳ reply to @tszzl
feel like a lot of brainpower in ML academia is spent making an idea or method sound novel after getting good results
16 ♥ · x.com →
· ↳ reply to @tszzl
ML in industry isn't immune to this lol -- "i tuned the hyperparameters" doesn't make for good marketing you have to pitch it using whatever buzzwords the higher ups are talking about lately. they pretend to be metrics driven but theres always more to it
9 ♥ · x.com →
all my homies hate the object level
43 ♥ · 3 RT · x.com →
if you haven’t either become president or abandoned politics by age 25 wtf r u doing
35 ♥ · x.com →
every dude ages 15-25 is in a group chat called "the degenerates"
92 ♥ · 1 RT · x.com →
This event takes place on the ancestral lands of the Smilodon, sadly driven to extinction by the "natives",
6 ♥ · x.com →
· ↳ reply to @tszzl
@eigenrobot although we probably need to distribute several orders of magnitude more corona vaccines than the ebola ones
5 ♥ · x.com →
· ↳ reply to @_djpn
@_djpn I hear aoc cummies But according to the man it’s AO cummies
2 ♥ · x.com →
@ellegist my first reaction was Old Testament rage at the unholy idiots this quiz really is something else
2 ♥ · x.com →
· ↳ reply to @aniiyengar
@aniiyengar lmao based alcohol probably increases the reproduction rate of a population
11 ♥ · x.com →
@halvorz yeah i think the better question is, given an ez virus that yields to known vaccine technologies, how do we get to vaccine launch in months instead of years
4 ♥ · x.com →
@halvorz What are the most rigid barriers for the phase III trials? For example, I would wager we could make improvements in the speed of recruiting trial participants
2 ♥ · x.com →
absolute hellworld
44 ♥ · 1 RT · x.com →
seeing the inherent value in space exploration is a male trait
52 ♥ · x.com →
@ProfoundProfit the last cope of a ProfoundProfit with no rebuttal! sad!
3 ♥ · x.com →
so whats the over/under that pfizer delayed the results til after the election
19 ♥ · x.com →
all modern debates are ended by whoever says "lol cope" in the most innovative and destructive way possible
103 ♥ · 3 RT · x.com →
· ↳ reply to @aquariusacquah
@aquariusacquah i remember at the start some medium article said we'd have a system where we'd go on and off lockdowns when the ICU beds reached a certain % of capacity and i was like lol these tech autists have no clue what theyre talking about
11 ♥ · x.com →
how to speed up vaccine development? i mean hats off to pfizer but ~1 year is still too long
18 ♥ · x.com →
· ↳ reply to @tszzl
setting aside a pool of testing machines for vaccine trials only is probably one low hanging fruit
7 ♥ · x.com →
· ↳ reply to @NemoBane
@NemoBane lol nah i meant ~1 yr from sequencing of the virus genome
1 ♥ · x.com →
· ↳ reply to @alexmschatz
@alexmschatz ok sure but it's objectively easier to recruit and monitor 100 people for an HCT rather than the ~50,000 required for the phase III trial
· ↳ reply to @archiveOfAwe
@_vivalapanda they have no qualms with that part but likely dont have the rest of the biopharma apparatus
2 ♥ · x.com →
· ↳ reply to @tszzl
not to detract from the success of the pfizer vaccine -- it's amazing that it works with 90% efficacy and after record breaking dev time -- even 60% would've been a success the market moves are generally warranted
6 ♥ · x.com →
· ↳ reply to @tszzl
@alexmschatz even with the current vaccine candidate, safety is not guaranteed. it will require 12 month, 24 month checkups etc so far the patients go through a vaccine 1, vaccine 2, and a checkup a month later for a total of about 50 days lag time but the vaccine trial took 100+ days
· ↳ reply to @tszzl
fwiw betfair is pricing this possibility at <5%
9 ♥ · x.com →
· ↳ reply to @tszzl
nate is clearly selectively looking at predictit, which favors long odds due to fee structure and pretending he found something interesting
9 ♥ · x.com →
Bay Area Drewligan Meetup
11 ♥ · x.com →
@browserdotsys @moultano otoh also makes me lol seeing NIPS submissions grow exponentially but the acceptance rate stays at exactly 21% very arbitrary stuff
1 ♥ · x.com →
elon musk is the only saving grace of the reddit ideology. they would have literally nothing to show without him posting rick and morty 420 cringe every now and then
17 ♥ · x.com →
@_AlastairX_ DL is not that great or interesting imo reads like a technical manual i second Norvig though also a sucker for Griffiths “Nature abhors a change in flux”
@AOCummies the whole ensemble with the twitter tag is just hilarious
10 ♥ · x.com →
rn im blackpilled on both theory and empiricism. send spiritual PhilSci help
6 ♥ · x.com →
so here's the bit: i weaponize math and science to arrive at all the same conclusions as my ancestors
22 ♥ · 2 RT · x.com →
the TL ranking algo fucks you if you deacc for a week lol i imagine it relies heavily on profile_views_7day
12 ♥ · x.com →
can't believe there's a whole journal that's just called Science. seems a bit on the nose
18 ♥ · x.com →
QT'ing my enemies with a simple "cope nerd. read infinite jest" and then logging off
27 ♥ · 2 RT · x.com →
Maggie Thatcher vs the British royals upjumped bourgeoisie vs landed aristocracy
11 ♥ · 1 RT · x.com →
· ↳ reply to @tszzl
tfw you will never reach the playful self satisfaction of the royals
3 ♥ · x.com →
Fleets are good akshually
8 ♥ · x.com →
anyone else still habitually make up a fake birthday on website dropdowns despite actually being older than 18
58 ♥ · 1 RT · x.com →
· ↳ reply to @tszzl
ok tbh I didn’t even think of the data privacy angle lmao. I just don’t remember that I’m not underage anymore
16 ♥ · 1 RT · x.com →
fleet is an instant success of a feature but nobody wants to admit it
30 ♥ · 1 RT · x.com →
@ProfoundProfit every few years you go and cull all the low performing features, problem solved
3 ♥ · x.com →
@ProfoundProfit @Everyone darpa is interestingly a very small and idiosyncratic institution (<$4b/year) that refuses repeated attempts to increase its funding bc it knows it cannot scale
6 ♥ · x.com →
@ProfoundProfit @Everyone the internet is a series of innovations made by gov agencies, academics, and vast amounts of big business, exactly what's needed for pharma as well its also not that controversial to say that the productive output of government research has fell off a cliff since the 60's
3 ♥ · x.com →
of all sad words of tongue or pen the saddest are these
23 ♥ · 1 RT · x.com →
· ↳ reply to @tszzl
“roon was right again”
9 ♥ · 1 RT · x.com →
· ↳ reply to @postjawline
@billpshort if you do this repeatedly, the models will start escalating your claims to higher scrutiny
9 ♥ · x.com →
ok here’s the bit. ill hate every last thing about my industry, regularly downplay its genuine scientific achievements, but still work in it bc hey haha it’s good tweet material
22 ♥ · x.com →
· ↳ reply to @tszzl
i remember staying up all night as a freshman synced to Korea time to watch Lee Sedol v AlphaGo. I consider it a formative experience
14 ♥ · x.com →
· ↳ reply to @tszzl
@vontallboy hyper parameter tuning is for the undergrads 😎
1 ♥ · x.com →
· ↳ reply to @alth0u
@alth0u this is the move almost as good as teaching Indian kids MCAT
2 ♥ · x.com →
· ↳ reply to @eigenrobot
@eigenrobot Ya agreed, it’s just the best steelman I can think of israelis gotta help us at this point
3 ♥ · x.com →
just make an ai smart enough to solve the alignment problem. there, you can disband miri now
15 ♥ · x.com →
this should be illegal
82 ♥ · 1 RT · x.com →
· ↳ reply to @NeoLibBen
@NeoLibBen swe @ FB&Goog can make this much out of college but you’ll find that it’s exactly as tedious as accounting
17 ♥ · 1 RT · x.com →
· ↳ reply to @jungseneca
@jungseneca @NeoLibBen no clue honestly maybe smaller companies have lower executive load (eg coordination and planning overhead) also most of my working life has been during covid so i may be overselling the tedium
5 ♥ · x.com →
· ↳ reply to @alth0u
@alth0u despite all my rage the high status of doctors is lindy and their public image will long outlast me
3 ♥ · x.com →
@LeefVan @NeoLibBen this is total comp figure including stocks and bonuses and most of them pay the same everywhere in the US senior eng (~5 years experience) can make 200 base
3 ♥ · x.com →
· ↳ reply to @tszzl
@LeefVan @NeoLibBen living in Seattle or Austin works out great bc you can get the same big tech salaries at low cost of living
2 ♥ · x.com →
imagine trying to systematize ur beliefs. Lmao couldn’t be me
13 ♥ · x.com →
Tesla is an AI company. the rest is irrelevant
11 ♥ · 1 RT · x.com →
@AOCummies roon is dead. long live turmeric monster!
16 ♥ · x.com →
· ↳ reply to @tszzl
this twete is designed to enrage the most possible people
12 ♥ · 1 RT · x.com →
· ↳ reply to @alth0u
@alth0u ooh I didn’t think about this maybe the boutique human driver niche will remain if they do their branding well enough look at the proles being driven around by robots
4 ♥ · 1 RT · x.com →
· ↳ reply to @tszzl
@alth0u I wonder if we can add this to the culture war Only blue tribe idiots would take a driverless car!
3 ♥ · 1 RT · x.com →
some personal news: I am joining Vox
70 ♥ · x.com →
· ↳ reply to @NateTheSpy
@NateTheSpy waymo is taking a very conservative approach technologically, reliant on deep mapping and LIDAR, and I think they’re getting muscled out by Tesla’s sheer data volume
2 ♥ · x.com →
· ↳ reply to @NateTheSpy
@NateTheSpy I mean I’m not sure what exact statistic but it’s clearly not the one to look at. Think of all the Californians owning Teslas and turning on autopilot. It’s gotta be at the minimum 10s of millions of miles in 2019 alone
2 ♥ · x.com →
· ↳ reply to @NateTheSpy
@NateTheSpy imo this has nothing to do with the marketing. Tesla has been a luxury vehicle until recently. Bought by well to do urban yuppies. they’re well aware that the car can’t drive itself; it’s just a natural inclination to stop paying attention when the AI system is performing so well
3 ♥ · x.com →
if you want the EU to be taken seriously at all move Macron to Rome rn so he can start the real “Jupiterian Presidency”
11 ♥ · x.com →
when you have to retain a large following of Marvel polytheists with robinhood accounts
28 ♥ · 1 RT · x.com →
· ↳ reply to @tszzl
nooo im not neurotypical don’t cancel my Thiel fellowship pls
34 ♥ · x.com →
resume of everyone who worked on fleets: “Built and scaled zero to one groundbreaking social features”
29 ♥ · x.com →
· ↳ reply to @averykimball
@averykimball Chesterton’s fence delivers a directional morsel of wisdom: the average mutation is net negative
7 ♥ · 1 RT · x.com →
· ↳ reply to @tszzl
@averykimball this is an important prior to have when proposing grand transformations
5 ♥ · x.com →
I’m wasted who up
36 ♥ · 1 RT · x.com →
· ↳ reply to @archiveOfAwe
@_vivalapanda yes but it’s more about encouraging competent governance via objective metrics (eg “beating the Soviets”) birth rates are a secular global trend due to cheap high tech birth control, women’s changed role in society etc
2 ♥ · x.com →
how am i getting full bars inside Yosemite. lame
61 ♥ · 1 RT · x.com →
· ↳ reply to @tszzl
why did they build this life size replica of my mac background ... seems like a waste ...
35 ♥ · 1 RT · x.com →
@borebrush there are more people in the park than in sf I swear
2 ♥ · 1 RT · x.com →
absolute best time to have a horrifying but localized facial abnormality
8 ♥ · 1 RT · x.com →
· ↳ reply to @eigenrobot
@eigenrobot i think we’re silently doing this with Anduril and virtual border walls and whatnot not sure what the state of this is
3 ♥ · x.com →
still thinking about how Yuval Harari basically said he meditates in isolation for 100 days a year so he could come up with better takes for the other part
45 ♥ · 1 RT · x.com →
sapiens is lowkey not that bad. I will be deleting this before tech twitter kills me
132 ♥ · x.com →
· ↳ reply to @tszzl
we’ve reached the counter counter signaling stage
44 ♥ · x.com →
· ↳ reply to @harveykrishna_
@harveykrishna_ beloved in like 2015-2016, hated after that until now. It will be making a comeback after my revolution
33 ♥ · x.com →
· ↳ reply to @postjawline
@billpshort first ~200 pages concern the prehistory of man and the agricultural revolution, i’d say 4.5-5. really well done stuff, gripping narrative rest is like 3.5, it’s mostly just Harari’s crack theories about history, some obvious, some radical and refreshing, some super wrong
10 ♥ · x.com →
· ↳ reply to @RaccoonYeshua
@RaccoonYeshua it was a super popular book in like 2015 after it went on Bill gates list. Got some heavyweight recommendations from zuck, Obama etc Basically purports itself as a universal history of mankind. It’s ideas became outdated almost as soon as it hit the shelves —
7 ♥ · 1 RT · x.com →
· ↳ reply to @tszzl
@RaccoonYeshua But imo the undertaking was interesting and valuable. Probably overrated revenue wise but gets too much negative heat as well
9 ♥ · 1 RT · x.com →
· ↳ reply to @RaccoonYeshua
@RaccoonYeshua haha yeah it assumes the starting point as the somewhat dead ideology of the obama clinton neoliberalism
· ↳ reply to @roddur_dasgupta
@roddur_dasgupta @billpshort i mean a lot of it is sort of neo-buddhist philosophizing re the human mind and its motivations sort of like what fukuyama attempts to do with End of History but not as skilled
9 ♥ · 1 RT · x.com →
· ↳ reply to @apoorvasriniva
@apoorvasriniva @RaccoonYeshua yeah, i appreciate the ambition of Big History / Hegelian Universal History and I don’t think “it lacks nuance” is a good enough criticism. better one would be that it’s not daring enough or the broad strokes are wrong
3 ♥ · x.com →
the hysteria was again unwarranted
50 ♥ · 3 RT · x.com →
· ↳ reply to @tszzl
@bayes_baes he was always going to make a shit ton of noise but then leave quietly. it’s exactly his style — it’s about the distinction between media artifice and reality
10 ♥ · x.com →
· ↳ reply to @bayes_baes
@bayes_baes ah yeah if he had a real but narrow path to victory, it would be a different story. but same would be true of any president the difference with this one is that he’ll keep tweeting about voter fraud even while he’s stepping out of office
1 ♥ · x.com →
@oldvillagesage @bayes_baes specifically the hysteria about coups, militias, fascist takeovers was unjustified. can’t say anything about long term damage to norms and the information ecosystem.
· ↳ reply to @trevposts
@trevorjtweets @jrobertsesquire i am disputing the idea that there was a ever even a serious attempt to pull off an information warfare based coup this is all just postmodern — trump cannot admit defeat or it ruins his brand
Manufacturing Dissent
23 ♥ · x.com →
· ↳ reply to @alth0u
@alth0u wonder why they don't move to the general space of logic puzzles, which would be way harder to memorize even a small % of
1 ♥ · x.com →
· ↳ reply to @alth0u
@alth0u @gigafelon on the high end this does not seem like a plus when you pay 300k for an engineer you probably want them to demonstrate high independence and direction
3 ♥ · x.com →
@halvorz tesla is an AI company and the rest are not — the value of a car that turns into a robotaxi is several times higher than COGS and will make their margins stratospheric, and not in the same league as the other auto makers. i think they are undervalued as yet
12 ♥ · 1 RT · x.com →
· ↳ reply to @tszzl
@JustJoshinYou13 @halvorz it was basically a bundle of vaporware they bought at a ridiculous valuation at the peak of the AV startup bubble to convince investors they weren’t missing the future — clearly it didn’t work
2 ♥ · x.com →
@forbca @halvorz oh i am absolutely competing with them by owning a lot of $TSLA lol more and more all the time
2 ♥ · x.com →
· ↳ reply to @tszzl
@idontwa86202030 @JustJoshinYou13 @halvorz imo waymo has failed in a surprising myriad of ways — google is yet again proving its inability to see new R&D ventures to completion. They are taking a conservative approach with deep mapping and LIDAR that I think have clearly been proven mistakes at this point
@halvorz haha explain why this is just the state of the science and engineering as i see it
1 ♥ · x.com →
· ↳ reply to @CkLorentzen
@CkLorentzen @halvorz probably. there may be a significant uphill legal battle against entrenched interests. I don’t know how to value those but I’m taking a guess that it won’t be that bad if they’ve proven a solid track record / demonstrably good metrics
@halvorz 2021 is ridiculous — I don’t agree with the exact year, just the directikn of the argument that they’re 5+ years ahead of competitors
1 ♥ · x.com →
· ↳ reply to @idontwa86202030
@idontwa86202030 perhaps but the mapping introduces a fundamental bandwidth / connectivity gap. can’t drive anywhere without a stable connection, without a serious degree of previous mapping. I agree the sensor suite is the bigger issue
· ↳ reply to @tszzl
@idontwa86202030 Especially since software can be shown to reach LIDAR levels of depth perception with just binocular Siamese neural nets
1 ♥ · x.com →
· ↳ reply to @CkLorentzen
@CkLorentzen @halvorz actually im expecting the public to be extremely laissez faire about it. Like Uber launching in a new city and the public consumers becoming its greatest defenders against the taxi interests and their cronies in government
· ↳ reply to @tszzl
@CkLorentzen @halvorz never underestimate the degree to which people just want a cheap ride but Tesla has been saying for a while that it needs to prove 10x than humans better before launching anyway
· ↳ reply to @idontwa86202030
@idontwa86202030 local just wouldn’t scale — these are extremely rich maps. So I assume server side with generous local caching
1 ♥ · x.com →
· ↳ reply to @sgodofsk
@sgodofsk @halvorz @petergodofsky @Rationalbot I can’t see half the comments in this chain but li cell batteries are way overhyped for energy storage if Tesla is primarily an energy company then they’re fucked — it would never justify a valuation this high
1 ♥ · x.com →
@Cullen_OK lol it seems her mothers family is from Manargudi though
1 ♥ · x.com →
$2.50/ride mile * 50% electric robotaxi margin * 100,000 cab miles per year (around avg for new york taxis) = $125k / year discount at 5% over 10 years the NPV of a Tesla model 3 with robotaxi abilities is around $1m at a cost of production of $20k after r&d -- 50x unit margin
23 ♥ · 1 RT · x.com →
· ↳ reply to @tszzl
100,000 miles is for human driver and seems low since the robot never needs to sleep. 50% margin is also probably low you won't buy $1m car to drive around in and then send to the taxi fleet for in the spare time tesla will sell them like capital assets to holding companies
10 ♥ · 1 RT · x.com →
· ↳ reply to @techsasbro
@ne0agent1c yeah $2.50/mile price point is not maintainable once a certain scale is reached
2 ♥ · 1 RT · x.com →
· ↳ reply to @Hellachans
@Hellachans not at the beginning anyways. they are simply too valuable for the average consumer to own for personal use. maybe after the margins slim quite a bit
2 ♥ · 1 RT · x.com →
· ↳ reply to @jeff82874662
@jeff82874662 @Hellachans how many cars will it take to satisfy the entire taxicab demand for the first world? and keep in mind that this demand expands as price per mile goes down but yeah at some point an equilibrium is reached and buying for personal use becomes viable again
2 ♥ · x.com →
· ↳ reply to @JeremyDanielFox
@JeremyDanielFox edge conditions are exactly why Tesla's data volume driven approach is most likely to work. the questions are of efficiency and not if the DL approach will work at all. which is why i'm pretty confident they'll get there at some point soon
3 ♥ · 1 RT · x.com →
· ↳ reply to @techsasbro
@ne0agent1c true good that there are alternatives, but what about when Amazon self publishing takes you off the kindle store?
3 ♥ · 1 RT · x.com →
· ↳ reply to @elidourado
@elidourado @calebwatney what about deployment / production? shortening the trial length may not make any different if the parallel production ramp takes equally long
7 ♥ · x.com →
pretty skeptical of these claims most IoT devices worth their weight in shit will fulfill their basic function with no connectivity https://t.co/zMYklM7SRk
25 ♥ · x.com →
· ↳ reply to @tszzl
the ny subway was NOT offline lol. Just the MTA website
9 ♥ · 1 RT · x.com →
· ↳ reply to @tszzl
the Roombas are still working! they just get slightly dumber and require hard boundaries to know where to vaccum to replace the software boundaries
8 ♥ · 1 RT · x.com →
How did they manage to break us-east-1 DURING thanksgiving code freeze that’s wild
46 ♥ · x.com →
@halvorz @jeff82874662 @SkepticalAlpaca assuming you could shorten the trials with HCT — is production still the bottleneck? are there any apparent ways we can reduce the time to produce a million doses of a novel mRNA therapy? How standardized is the production from one mRNA therapy to another?
1 ♥ · x.com →
· ↳ reply to @gbrl_dick
@gbrl_dick not sure if its a misunderstanding so much as its a cope starting another company that makes life for urban yuppies 0.5% easier is just more likely to make it to the end lol
1 ♥ · x.com →
@halvorz @jeff82874662 @SkepticalAlpaca oh wow i knew Moderna was new but not this new -- i had assumed they had a few successful drugs with this approach already also cackling at stock ticker $MRNA
3 ♥ · x.com →
· ↳ reply to @gbrl_dick
@gbrl_dick yeah, a lot of people that were previously in startup world are quitting and coming to FAANG from what i can tell my manager quit his company and came to work at FAANG like six months ago, dejectedly telling everyone that startups suck now
4 ♥ · x.com →
· ↳ reply to @tszzl
@gbrl_dick there isn't much energy in the ecosystem since there aren't many new technologies optimistically it's a short lull before the storm of whichever new platform comes next after mobile
2 ♥ · x.com →
there are as many people claiming that the internet is creating division and factionalism as there are claiming that it's creating a homogenizing monoculture what means
38 ♥ · x.com →
· ↳ reply to @jachaseyoung
@jachaseyoung very odd claim haha i doubt Ted Chiang is selling THAT much product to the tiny audience of would-be short story writers
5 ♥ · x.com →
severe shortages of Vision everywhere
32 ♥ · 1 RT · x.com →
· ↳ reply to @archiveOfAwe
@_vivalapanda on the other hand interest rates are record lows which suggests long term thinking at the ecological level
6 ♥ · x.com →
@browserdotsys yeah for some reason saw lots of people online claiming the subway itself was nonfunctional lol there's no way any of the controls software is internet connected
3 ♥ · x.com →
· ↳ reply to @samgdf
@samgdf yes this is just as much a self reflection haha
1 ♥ · x.com →
· ↳ reply to @AlecStapp
@AlecStapp how’s that looking though lol SF management is very well paid for a city widely accepted as one of the worst run
9 ♥ · x.com →
· ↳ reply to @tszzl
another tell for this is that the people who spend the most time thinking about politics are strangely enough often undecided voters. the more they analyze the more they're buried under the sheer complexity
16 ♥ · 1 RT · x.com →
why democracy? the decision heuristics people use to democratically choose leaders are no better than random; neither the elites nor the public have any insight into the complex systems they wish to govern
51 ♥ · 2 RT · x.com →
· ↳ reply to @tszzl
pair this with the increased inclination to vote against incumbents when economic conditions are bad, and you have a random system that kicks out ruling parties stochastically, weighting toward leaders who oversee bad times and that's pretty much it
19 ♥ · x.com →
· ↳ reply to @tszzl
simple -> democracy randomly cycles out the tired elites with a new order every X years i suspect this accounts for most of the efficacy of the democratic algorithm. RNG, a source of noise. rapid flux in leadership decreases the probability of entrenched corruption
24 ♥ · x.com →
· ↳ reply to @tszzl
there need not be any relation between the tribal aesthetics of a would-be leader and their ability. policy discussions elude everyone, including the wonks. at best, a policy proposal is an educated guess. at worst, specious nonsense. so why democracy? why does it work better?
12 ♥ · x.com →
· ↳ reply to @jonshea
@jonshea @gaitanalyst yeah I’ve never actually seen it I kind of doubt there’s any good space opera tv that I’ve missed
· ↳ reply to @wesyang
@wesyang we can measure the impact with click through digital ads
6 ♥ · x.com →
which financial instruments let you bet on the real estate of a specific city or metro area?
22 ♥ · 1 RT · x.com →
normalize turning on the shower to check water pressure when touring a new apt
88 ♥ · 2 RT · x.com →
· ↳ reply to @tszzl
car:test drive::new place:test shower
11 ♥ · 1 RT · x.com →
being anti IoT is not contrarian at all. i will be purchasing smart trash cans and whatnot
43 ♥ · 2 RT · x.com →
· ↳ reply to @ChrisCroy
@ChrisCroy lmao but that's the thing everyone seems to miss -> as long as there's always a hardware fallback, IoT is a pure net positive (aside from infosec concerns)
7 ♥ · x.com →
· ↳ reply to @tszzl
@ChrisCroy that may be a big asterisk for some people but i seem to just not care about data privacy at all after all i overshare on the internet to 2500 people with a real face avi
7 ♥ · x.com →
@browserdotsys lmao no but i imagine the firmware is written by offshore contractors for pennies
5 ♥ · x.com →
· ↳ reply to @alth0u
@alth0u paper analyzes the self play / simulation training argument vs overfitting argues that the absurdist/fabulist nature of dreams makes them poor training data for making real world predictions may play a different role in children tho
1 ♥ · x.com →
· ↳ reply to @tszzl
hmm this may also explain a bit of the "seeing a problem with fresh eyes" phenomenon
3 ♥ · x.com →
low real interest rates imply a forward thinking / long term focus in the economy but it doesn't feel like it culturally?
19 ♥ · 1 RT · x.com →
ultimately there is a hard limit on how much RETVRN is possible due to fact that rival countries will beat your military technology
57 ♥ · 4 RT · x.com →
· ↳ reply to @tszzl
you could argue that google and facebook are bad for the culture, causing social degradation etc, but you'd also have to contend with the fact their output in computer vision and such will directly contribute to future military tech
29 ♥ · 2 RT · x.com →
· ↳ reply to @0xrohan
@RohanXKapur haha this wasnt a subtweet dawg im just counter signaling all the IoT hate i saw after us-east-1 went down to be honest i dont own any IoT stuff and i even threw out my smart lightbulb -- more effort than it's worth
1 ♥ · x.com →
· ↳ reply to @orthonormalist
@orthonormalist maybe so, but in general a technology oriented society is prerequisite for military + economic dominance
1 ♥ · x.com →
@pookleblinky forget resource pooling. We could have game consoles running giant neural agents for in game AIs lol. no more shitty hard coded boss fights, no more boring procedurally generated worlds — each level radiantly handcrafted by an NN
@pookleblinky it’s true, but training bottlenecks on interlink bandwidth, especially for reinforcement learning. you’d get v tiny utilization as most time is spent waiting for data. there are probably federated learning algos that solve some of these problems but not confident
1 RT · x.com →
· ↳ reply to @tszzl
@pookleblinky as in most cases, you pay a heavy cost for decentralization in the form of efficiency/performance. This sort of distributed learning would be ideal for problems you literally cannot trust to governments or companies
plastic bags cost $.25 a pop in San Francisco lmao why does california hate the poors
30 ♥ · 2 RT · x.com →
· ↳ reply to @rchitectopteryx
@rchitectopteryx lol im not price sensitive enough to care and ill happily rip through a ton of plastic bags this only hurts the poor plastic bags end up safely in landfills anyway
1 ♥ · x.com →
two numbers: A: the threshold immigration capacity that Americans are comfortable with — any more and they revolt B: the number of immigrants that the US actually needs to backfill flagging fertility rates and maintain positive pop growth If A < B we’re already fucked
44 ♥ · 2 RT · x.com →
· ↳ reply to @777_sage
@KiIIAIICops as the prime age population : elder ratio skews towards elders, our already stressed entitlement system will fold
5 ♥ · 1 RT · x.com →
· ↳ reply to @tszzl
@KiIIAIICops not to mention that unceasing economic growth is the engine of our civilization and all of our financial assets etc depend on it
6 ♥ · 1 RT · x.com →
· ↳ reply to @tszzl
the cia only wishes they could do inception shit like this cc @gbrl_dick
11 ♥ · x.com →
· ↳ reply to @LTF_01
@LTF_01 one of the quietest most boring suburbs in the world lol
8 ♥ · x.com →
· ↳ reply to @alth0u
@alth0u actual Carthaginian figs legislative play awesome
1 ♥ · x.com →
wait so you mean not everyone grows up spending 6+ hours on the internet per day in maplestory chat rooms since age 11 thereby fine tuning their brains for online communications only
153 ♥ · 7 RT · x.com →
· ↳ reply to @acczibit
@acczibit in the comics the plane incident literally causes 9/11 so yeah spot on
@The_Geraldoid yeah i don’t really see it but I guess it vaguely looks like the meat is full of maggots / triggers disgust instinct
1 ♥ · x.com →
· ↳ reply to @R1Jack
@R1Jack yeah agreed it will only ever be a short term fix
twitter is a prison and the timeline is my guard Read Foucault
32 ♥ · x.com →
· ↳ reply to @_metallurgist
@metametrician @mechanicalmonk1 there’s some spooky shit I’ve noticed though if you visit someone’s house for the first time it’ll correlate everyone under the same external IP address and recommend you to each other not mystifying from a technical standpoint but the effect is still magical
5 ♥ · x.com →
· ↳ reply to @aniiyengar
@aniiyengar at a truly big company they won’t ask you this. they’ll move onto the whiteboard questions lol startups want to see some level of enthusiasm for the product tho
6 ♥ · x.com →
· ↳ reply to @sonyasupposedly
@sonyasupposedly @metametrician @mechanicalmonk1 actually kind of doubt this is true Twitter/Facebook’s most valuable asset is the trove of user data — if anyone had it, anyone could target ads as well as them. As such they take great care that only they get to extract the ML rents from it
6 ♥ · x.com →
· ↳ reply to @tszzl
the micro lending here is probably under appreciated
15 ♥ · 1 RT · x.com →
· ↳ reply to @aquariusacquah
@aquariusacquah i doubt there’s much opportunity in the US yeah primarily other markets — could see this taking off in India (paytm is not all that functional)
4 ♥ · x.com →
· ↳ reply to @aquariusacquah
@aquariusacquah tbh they’re not all that eager to let American companies in either but fb already has a giant footprint in India lol this may accelerate the merging of Aadhaar card identity with FB identity and I’m not sure if that’s good
2 ♥ · x.com →
if i like every tweet in your thread except the top one please note im just trying to avoid twitter snitching on me faving ur beautiful but problematic tweets
33 ♥ · x.com →
· ↳ reply to @Tyler_The_Wise
@Tyler_The_Wise agree with the general sentiment but is there any specific case of the FDA eg waiting too long to approve the phase III trial or something? it doesn’t look like it
1 ♥ · x.com →
· ↳ reply to @aquariusacquah
@aquariusacquah yeah I’m guilty of this too tbf post hoc policy buzzword solutions are cheap and make you sound smart
8 ♥ · x.com →
· ↳ reply to @tszzl
@Willyintheworld @aquariusacquah also in this case HCT may have shortened the approval timeline by a few months but wouldn’t have changed the bottleneck of production ramp at all
1 ♥ · x.com →
@browserdotsys ok call me crazy but I think imagenet dying is net good for computer vision I don’t want to see any more 99.2 -> 99.4% improvements, the whole dataset got Goodhart’d
4 ♥ · x.com →
are there ML algorithms for jury screening yet
19 ♥ · 1 RT · x.com →
· ↳ reply to @ChrisCroy
@ChrisCroy this is true but at a surface level lot of naive 1:1 mappings between human and AI labor but have not rethought entire vertical stacks
3 ♥ · x.com →
@guywhits yeah high key loved it high key sucked
2 ♥ · x.com →
@halvorz deepmind has a problem where they’ll do something genuinely amazing but then oversell it anyway for marketing
4 ♥ · x.com →
· ↳ reply to @wesleyytian
@wesleyytian only the stuff that escapes elegant encapsulations: where induction is more powerful than deduction
2 ♥ · x.com →
· ↳ reply to @alth0u
@alth0u yup, repurpose self attention transformer to do graph learning and profit none of these are new methods
2 ♥ · x.com →
· ↳ reply to @WillManidis
@WillManidis @alth0u haha maybe so if you prune down the space to "applied ML" but i think there are lots of theory advances left in Deep RL, novel architectures etc some of these will become clearer on the next compounding of compute power
1 ♥ · x.com →
· ↳ reply to @tszzl
@WillManidis @alth0u i continue to believe that the interplay between cognitive neuroscience and deep learning will be one of the great flywheels of 21st century science and i'm not sure that we're close to the end
5 ♥ · x.com →
· ↳ reply to @WillManidis
@WillManidis @alth0u agreed tbh i got a whole neurips paper out of this insight and doing some engineering that no one else would do
1 ♥ · x.com →
· ↳ reply to @eigenrobot
@eigenrobot at the last i wish there were 50 state level FDAs with varying levels of risk tolerance + different approaches
39 ♥ · 1 RT · x.com →
· ↳ reply to @tszzl
@eigenrobot least* probably can't eliminate the clinical trial process entirely but it should function more like a multi armed bandit continuous growth of the test arm as long as the results seem better than the control arm
57 ♥ · 1 RT · x.com →
@halvorz the fact that not even China used them makes me sus
9 ♥ · x.com →
· ↳ reply to @tszzl
@halvorz also I’m not sure but “current methods” may be doing some heavy lifting by here
1 ♥ · x.com →
@browserdotsys @halvorz it is strange because their approach is conceptually very simple, and in fact it's pretty much what all the academic researchers were doing as well
3 ♥ · x.com →
· ↳ reply to @tszzl
@browserdotsys @halvorz they started with some fancy methods where a GAN would dream up fragment and another net would assemble together but afterwards they simplified to just predicting inter-residue distances (at a finer scale than most), and it was enough to win CASP13 by a long margin
· ↳ reply to @tszzl
@browserdotsys @halvorz CASP14 seems to just have added some well known graph-based attention architectures so i guess it's literally just superior engineering?
1 ♥ · x.com →
· ↳ reply to @alth0u
@alth0u @browserdotsys @halvorz i don't think learning TF is the bottleneck; there are plenty of really zany neural nets that have been thrown at this by academia
1 ♥ · x.com →
· ↳ reply to @tszzl
@browserdotsys @halvorz i'm also wondering does this enable the reverse AlphaFold? come up with an RNA sequence that will roughly match a desired shape
@halvorz @browserdotsys but we don't need to consider all 20^100 options right? just like we don't have to consider all 3^100 conformers when folding the protein if alphafold can inference shapes in novel parts of the protein space then it should help the reverse problem as well via trial and error
2 ♥ · x.com →
· ↳ reply to @tszzl
@halvorz @browserdotsys that may not have made too much sense but my point is that you can use alphafold as a sort of reward function to a reverse model if alphafold performs well enough out of distribution
2 ♥ · x.com →
@browserdotsys @halvorz imo this is the main reason why "protein folding is solved" is a bit too grandiose protein folding with co evolutionary data may be somewhat solved but may not help when trying to construct novel proteins
3 ♥ · x.com →
@halvorz seems to be burying the lede — how does protein structure not help with understanding viable drug targets? better, cheaper QSAR simulation, docking simulation, etc
1 ♥ · x.com →
· ↳ reply to @tszzl
@halvorz I’m sure Derek knows better than me but as an example, on the other side of the binding equation, conformer search is v important in searching for viable small molecules to bind to proteins
1 ♥ · x.com →
the most mysterious thing in all this is, which software startup is drowning in hot babes wtf my man is worrying about overpopulation on mars
71 ♥ · 1 RT · x.com →