It's funny how different the vibes of the upwardly mobile middle class are vs the downwardly mobile middle class. My friends with blue collar dads getting engineering degrees vs kids of doctors and lawyers studying humanities, stark divides in attitudes and optimism levels
there was all this fuss about "watermarking gpt outputs is impossible" last year. but they did it! i can recognize chatgpt text basically instantly. the writing style is insanely distinctive.
guys I work at openai and yesterday our WIP model (codename "shrike") proved that P=NP and found a unified theory of quantum gravity(!!). I can't give any proof but it was kinda spooky hehe. Altman told me not to release it because we don't want to scare the hoes (world leaders)
if you ever feel small, like your anon lowbie trolling has no effect
just remember this journalist poasted ass because a random with 70 followers called her bitter
@Britonomist
the avg leftist response I see to this stuff is "destroy finance industry" when really we should just make academia less miserable. the conditions for a pre tenure academic are so shitty that even if there were no quant jobs they still aren't going to stick around making 60k
you can probably get 80% of the benefit of a math degree in <20% of the classes. Real Analysis, Linear Algebra, Abstract Algebra, Mathematical Stats. That gets you most of the way there I'm pretty sure. Maybe throw in an applied course like Numerical Analysis or some Physics.
i really hate how i very deeply want to disagree with EY but basically all the arguments against his position are terrible.
getting close to grudgingly accepting his position despite the vibes
ML systems are just matrix algebra training data soup, and will only ever be as good as their training data. This is why AlphaZero is only as good as human players at Chess and Go :)
witnessed perhaps the highest iq counterfactual fail ever today
prof: assume that you have access only to knowledge of classical physics, how could we predict the result of this experiment?
student: we shall use the electrons spin sir
1/2
@jasoncrawford
cool paper but incredibly overhyped on twitter. it's essentially the insight that LLMs are a good mutation operator for program search with evolutionary algorithms. They don't "solve" anything, just push the sota on some combinatorics problems.
@dorkweeb
true, this was half joke. but also half serious. maybe a skill issue, but I used to be able to get it to imitate writing styles, sound natural etc. now its all in the same tone, and an *incredibly* distinctive tone at that.
in awe that this happened right in front of me. i feel like we should study this guys brain. or whatever professor managed to teach him basic quantum physics. something remarkable going on there
Got harassed/followed by a junkie on the way home from work today :(
Was preparing for it to come to blows but luckily a transit cop was nearby and helped deescalate. More motivation to lift heavy tonight I guess...
the value is going to be in training better foundation models, not doing random prompt engineering. that they're making this a "big thing" makes me think they're not able/willing to try and push the foundation model sota.
@rob_mcrobberson
"We need to make open source illegal so China can't use it.."
bruh they just need to send some hot spies to SF and chat up the engineers.
@Bdubs1776
the kind of people who become forklift guys has changed /a lot/ in the past 40 years. like i guarantee if that guy was a teen today he'd be going to college.
@rickyflowsinyou
@rob_mcstan
Haha. Hahaha. If only you knew how bad it gets. My boss at my first internship told me I'd be the last male intern they'd get for a while as HR was getting on their ass about the gender ratios in engineering
It’s the era I think. I noticed a sharp drop in men cold approaching me around 2016- I thought it was just bc I was getting older and was married, but I’ve met cute girls in their 20s who say men only approach them on the apps
@AW_Baker
@EigenGender
to learn how to do useful things LLMs cant do you have to learn the foundational skills that LLMs can now do. you have to be bad at something in order to eventually become good at it
@rickyflowsinyou
@rob_mcstan
This is by far the worst I've seen, but at every company being not male or not white/Asian was seen as a non trivial positive for a candidate
There are zero good reasons to be an atheist and about a million good reasons not to be
At this point, it appears to be an artifact of youth, religious trauma or poor reasoning capabilities
im thinking about code interpreter being released as a watershed event. if this is not a world changing gdp shifting product im not sure what exactly will be. every person with a script kiddie in their employ for $20/month
@madeofmistak3
I don't know you, but I'll make a simple request.
please do not compare yourself to (possibly) the smartest man who ever lived. it is not productive. There are probably double digit people throughout history who can compare themselves to JVN in a healthy manner.
Eliezer (correctly) perceived that aligning the coming ASI was the most important thing he could be doing, and then tried to do it analytically and got depressed when he realized that was impossible. Luckily for us, we can just do it numerically and it just kinda works.
how can i cultivate my timeline to be more ml and math and less politics / culture war stuff. feels like im kinda scraping for gems in the rough on twitter rn
Young Iranian couple sitting across from me in the cafe. Girl is excitedly showing the boy chatgpt. The boy is amazed, "how can it talk like that!".
He proceeds to dictate MMA questions to his gf for 15 minutes before I have to leave.
I've basically always believed scaling LLMs isnt enough, but honestly at each new step that belief needs to pay more rent. I'm much less certain now than I was in e.g 2019 that scaling LLMs isn't enough. At this point I think a priori arguments (i.e imitation learning can't be
Sam Altman (CEO of OpenAI): There are more breakthroughs required in order to get to AGI
Cambridge Student: "To get to AGI, can we just keep min maxing language models, or is there another breakthrough that we haven't really found yet to get to AGI?"
Sam Altman: "We need
@hollowearthterf
@chickenstrip007
@CartoonsHateHer
I don't think we can conclude that based on one reddit post from her perspective, these kinds of things are subconscious. If she's treating him worse she probably doesn't even notice it.
@annakhachiyan
The only thing here that's missing from the bronze age is the iPhone. Bronze age warfare is putting men to the sword, cities to the torch and enslaving whoever's left. (And this is why progress is good!)
@thechosenberg
this is a totally reasonable request for both genders to make of each other if oral sex is happening. hair in your mouth is very unpleasant, my gf and I both shave "down there" for each other.
the avg person who hates "hedge funds" and "wall st" has no idea what either of those entities actually do. it's literally 100% vibes / generalized hate for wealthy/successful people.
@genomerambler
between this and my morning walk to the beach i am really oscillating between "i love nature the universe is so beautiful" and "oh my god we must impose total human order and values on everything the world is full of horrors"
one of the fun parts of being the guy in the group of math majors who is "the coding guy" is you get to be in the equally entertaining / frustrating position of being blue explaining something to red.
that is, explaining something to someone who has a much higher IQ than you
in programming there is saying, “the camel has two humps”: if you plot who’s good at programming there are two humps. it’s essentially a normal distribution of general intelligence, divided by a yes/no on the special property of “just gets it”
a lot of courses after this just felt like learning Weird Tricks. but these ones are like, core to how i approach every problem in my life and how i think about basically everything.
@pmddomingos
I don't think this will age well. Theres a bit of a revolution in robotic control rn centered on using DRL to learn a policy in sim and zero shot transferring to a real robot. Needs a bit more time to mature but I wouldn't be surprised if it has a large commercial impact
@constans
he's getting dunked on this but wikipedias operating costs are like 2% of the Wikipedia foundations budget and they are deliberately misleading about this when asking for donations
@augmentaltech
this seems like some pretty amazing accessibility tech for people with neck down paralysis or other conditions that make using a mouse impossible/painful