New @DataColada post providing evidence of fraud in the work of Harvard Professor Francesca Gino. http://datacolada.org/109 I can't say I am surprised by this one, but it is very good she was caught! This is part 1 out of 4.
Where’s the best online place to find a sublet in London? Looking for something for 3-4 weeks starting at the end of August (ideally a studio, but will also share with someone else if needed). AirBNB is so expensive 😥
claimed GPT4 can score 100% on MIT's EECS curriculum with the right prompting.
My friends and I were excited to read the analysis behind such a feat, but after digging deeper, what we found left us surprised and disappointed.
https://dub.sh/gptsucksatmit🧵
Exploring the MIT Mathematics and EECS Curriculum Using Large Language Models
Presents a comprehensive dataset of 4,550 questions and solutions from all MIT EECS courses required for obtaining a degree
https://arxiv.org/abs/2306.08997
AISF programme, then probably hiring a tutor to brush up on linear algebra (any other math topics I should start with?) and then tutoring for ML (DL I guess) specifically thereafter. Maybe a master’s at some point but idk if the opportunity cost is too high
Standard advice is that you should have a low bar for removing parts of your writing, but for me if I write a line/paragraph I like I try hard to keep it
This probably makes my writing longer, and in some sense worse, but is good training for thinking about the structure of it
'oh I have this line on page 3 I like, but it doesn't transition well, how do I rearrange things so it fits well?' is something I end up dealing with a lot, which is good training for a part of writing (and probably pushes me more into a certain type of style)
Well... it didn't.
The authors evaluation uses GPT 4 to score itself, and continues to prompt over and over until the correct answer is reached. This is analogous to someone with the answer sheet telling the student if they’ve gotten the answer right until they do.
(2/4)
Many exaggerate the amount of randomness and bias in society.
Yes, there is a substantial amount of that, but overall, people in more influential and important positions are on average more competent than the population average.
Thanks! I definitely don't want to force conclusions on anyone. I'm often struck by the fact that complaints about excessive spending are more frequently made in the longtermist space than in the short-termist space.
This is because the nature of the work imo. Broadly, much of global health is executing evidence-based projects where impact is more legible. Longtermism is a lot more exploratory, it's harder to measure the value of different projects, so they can (& should) get scrutinized
I hate this website because unless it's all uncontroversial shitposting all day every day you people start bashing your heads against the wall and getting angry at people ostensibly on the same side as you for no reason other than a significant lack in reading comprehension
The Japanese dwarf flying squirrel usually dwells in coniferous trees and possess a patagium, which is a skin membrane used in gliding from tree to tree.
[read more: https://buff.ly/3PbNe1C]
Hilarious thing is that below-grade space usually doesn't count against FAR in America
This is in fact the only form of mixed-use urbanism allowed--you can literally build an unlivable labyrinth underground as long as Boomers can't see it aboveground
For some reason my headphones will randomly change the left right audio levels, sometimes subtly (30 vs 32) and it's a great way to know my hearing is the same in both ears
[manually editing the neural net's parameters in Excel to build an aligned ASI]
ok... if I just tweak this weight here... maybe that'll do it
*ceiling splits in half, light pours into the room from the sky*
𝕐𝕆𝕌 𝔸ℝ𝔼 𝕋ℝ𝕐𝕀ℕ𝔾 𝕋𝕆 𝕊𝕆𝕃𝕍𝔼 𝕋ℍ𝔼 𝕎ℝ𝕆ℕ𝔾 ℙℝ𝕆𝔹—
I think many attribute overconfidence in one's beliefs to social desires - e.g. to project greater competence than one has.
But overconfidence is likely partly due to cognitive constraints.
It's easier to create a model of the world and just believe it (naive realism) than ->
has passed away. She was a delightful person and an important figure in Animal Studies. Do yourself a favor and check out her podcast Knowing Animals and her book Animals, Equality, and Democracy to see what made her so wonderful.
We are very sad to share that Siobhan O’Sullivan, the founder of Knowing Animals, passed away peacefully yesterday after a battle with cancer. Our thoughts are with her family and friends. Her influence will be felt in the global animal studies community for years to come.
Kant is so easily refuted: You should not lie because if universalized, lying wouldn't work. Bad consequences, in other words. It's just utilitarianism. Why has no one noticed this?
Good post by Dr Buolamwini. I agree strongly re: the opportunities for strategic cooperation between groups who don't see all aspects of the risks the same way; many concrete interventions that serve to make the world safer across these concerns. 1/2
ICYMI , this morning Tawana Petty spoke alongside Professors Yoshua Bengio (Mila) and Max Tegmark on @democracynow about AI threats. I have immense respect for Dr. Bengio who put his reputation on the line to defend early AI bias research from corporate attack. I do not view…Show more
Okay, two medical studies someone might do:
1) Does forecasting training make doctors better prognosticators?
Many doctors are poor prognosticators; this matters, because predictions of a patient's trajectory greatly affect the care doctors provide. Is this fixable?
2) How does physician payment reform affect gendered pay disparities?
There are many efforts to change how doctors are paid (eg, less fee-for-service); some seem like they could reduce disparities (eg, value-based payments), while others could do the opposite. What has happened?
Why restrict the funds? I don't think the state has legitimate reason to meddle in people's consumption in this case. It also introduces unnecessary bureaucracy.
Wow: all 750,000 Germans turning 18 this year get 200 euro to spend on any cultural good: concerts, theater, movies, musical instruments, books, records etc. It'll cost 100 million euro and is given in recognition of what young people lost during pandemic.
https://dw.com/en/free-money-germany-launches-200-culture-ticket-for-18-year-olds/a-65887898…
Could anyone recommend good resources on a) why alignment is important and b) how you can start working on it for a technical, AI academic audience? Thanks a lot
100,000 Syrian refugee kids enrolled in school in Jordan in just one year in 2013 - thats 7% of the total school population.
Despite this, there was zero negative impact on learning for Jordanian kids
new by
A new Black Mirror season is out!! 🥳
Can't wait to see what s-risks they explore this time 😄
Just finished the first episode and it's incredible.
I will no longer be answering emails for the next 48 hours 😛
Update/[minor SPOILERS]: only the first episode was good according to me. Most weren't even sci-fi! They were just thrillers or fantasy.
If you're only in it for the thoughtful exploration of tech, only watch episode 1 and Beyond the Sea
I didn't finish the last episode…Show more
Daniel Ellsberg: "From the time the Berlin Wall came down, [inaudible] were in Warsaw, selling them on the need for F-22s..."against who?", as the Russians reasonably asked."
Yes, they were fearful of Russia, and it's pretty clear now that newly liberated Soviet colonies' fears…Show more
Every year, Swedish tabloids publish lists of people with the highest income
"13,000 names - search where you live"
Helps criminals and infringes privacy
Weird that privacy advocates don't focus on this very real privacy infringement, instead prioritising anti-crime measures
“I do not view AI existential risks in the same way that has been framed by Dr. Tegmark and others, and at the same time see room for strategic cooperation.”
We need more of this. Coalition partners don’t need to agree on all points to achieve mutually-desirable arrangements.
ICYMI , this morning Tawana Petty spoke alongside Professors Yoshua Bengio (Mila) and Max Tegmark on @democracynow about AI threats. I have immense respect for Dr. Bengio who put his reputation on the line to defend early AI bias research from corporate attack. I do not view…Show more
Many people I know (>100) have felt bullied and silenced about AI extinction risk, for many years, by being treated as crazy or irrational. Many of them were relative experts who knew AI would present an extinction risk to humanity, but said little or nothing in public or even to…Show more
Reminder that most people think (wrongly) that uncertainty within climate models means we should take climate change less seriously ("well, it's uncertain so who knows"), not more seriously ("shit, it could be worse than we expect"). Seems related to some dismissal of AI X-risk.
Amazed someone thought of this
"Novel ways of broadly sampling the human population in a non-invasive method would allow for large-scale surveillance at a reduced cost. Here we evaluate the collection of naturally bloodfed mosquitoes to test for human anti-SARS-CoV-2 antibodies"
Tracking SARS-CoV-2 seropositivity in rural communities using blood-fed mosquitoes https://medrxiv.org/cgi/content/short/2023.06.13.23291267v1…#medRxiv
Yes, if you put aside prejudice and class interest and see this with fresh eyes, it is pretty odd. (The quote is by Euan Blair, Tony Blair's son, talking about education.)
"I began my career in investment banking at Morgan Stanley. My degree in ancient history and my master's in international relations were in no way useful to me for this job I was supposed to know how to do, and yet they were, for some reason, deemed a prerequisite." twitter.com/TheNewsAgents/…
I also think possible motivated reasoning is worth tracking. While some recent examples have strained credulity, I'd personally like to see legal liability for harms (present and future) caused by models getting more attention in current discussions.
I’ve yet to meet a tech bro in 17 years in the valley. I grew up on football hockey and lacrosse teams. Bro dense environments. Tech is probably the least bro’y industry in all of capitalism. Dudes, many. Bros, nope. Have you seen commercial real estate? Wall st traders?
- 10 years ago, it was indeed the case that essentially no AI (or at least ML) researchers of note took AI x-risk seriously as a global priority.
- This is emphatically no longer the case.
- Nobody should get away with saying "experts aren't concerned" anymore; it's fake news.
"I'd only assume good faith of AI extinction concerns by someone both VERY respected in academia AND who had worked at the cutting edge of industry BUT ALSO had quit industry AND who was explicit about extinction"
You mean like Geoffrey Hinton?
"What? No, no – he doesn't count"
man... the list of bad jokes that workaholics writers weren't allowed to use is so painful, it's like a time capsule of the most annoying millennial humor from the early/mid 2010s
I've been trying to write a good Tweet(/thread) about why objections that AI x-risk is "not scientific enough" are misguided and prove too much.
There's a lot to say, and it deserves a full blog post, but I'll try and just rapid fire a few arguments: