Opens profile photo
Follow
Dan Elton
@moreisdifferent
Problems are solvable - so let's solve some! Im at v
Science & TechnologyVibecamp moreisdifferent.substack.comJoined October 2010

Dan Elton’s Tweets

One of the great lessons people learn in physics is the importance of defining terms precisely and using them consistently. I wonder if we taught more physics in High School (like other countries do) if people wouldn't constantly intermix the terms "sex" and "gender".
1
3
Show this thread
I've talked about this before on here.. I think this is a common assumption about deep learning that is probably untrue in most cases ("if the model is trained long enough it will find the true generating function for the data"). There's something called the Rashomon effect --… Show more
Quote Tweet
Replying to @di_goldene_pave and @moreisdifferent
Having no technical background in this field, it seems plausible to me that a sufficiently large model trained for long enough could just fully learn the underlying pde.
1
5
Consider applying if you: "Currently organize ACX/LW/rationality meetups or other rationalist community events anywhere east of the Mississippi river. Would like to start an ACX/LW/rationality community in a city where none exists, or where the community has been small or… Show more
Show this thread
Tempted to paint my nails now just to see all the various reactions from Twitter people. I have a bottle of dark green matte nail polish somewhere I bought a few years ago but then never worked up the courage to wear.
Quote Tweet
Replying to @moreisdifferent
Indeed. Also, I got gel nails done for a wedding recently and they’re awesome.
2
PSA: applications are open for the Longevity Biotech Fellowship It starts with a 3 day IRL retreat, which is followed by a 12 week online program. longbiofellowship.org
Quote Tweet
1/ 🧬 Applications are OPEN for LBF2! 🧬 Join a community that refuses to go gently into the night - we 🧱 build, 💰 fund and 📢 advocate for tech to maximize healthy lifespan. Apply now at longbiofellowship.org
Show this thread
Embedded video
0:44
5K views
4
This is real bad, folks. These models are not replicating the real-world distribution (which is already problematic), they pushing the worst stereotypes to an extreme degree. I knew this was a big problem but now realizing its even worse than I thought.
Quote Tweet
For example — while 34% of US judges are women, only 3% of the images generated for the keyword “judge” were perceived women. For fast-food workers, the model generated people with darker skin 70% of the time, even though 70% of fast-food workers in the US are White. 🧵 4/13
Show this thread
Embedded video
0:10
9.9K views
3
13
Yeah... the whole idea of training LLMs on "synthetic data" from other LLMs to advance AI capabilities doesn't even pass a basic "sniff test" in my mind. It's the very definition of "pulling yourself up by your own bootstraps". Turns out it may actually be harmful!
Quote Tweet
In work led by @iliaishacked we ask what happens as we train new generative models on data that is in part generated by previous models. We show that generative models lose information about the true distribution, with the model collapsing to the mean representation of data
Show this thread
Image
3
9
Some of the fraudulent papers were in Nature and Science. For instance: nature.com/articles/natur This sort of fraud is incredibly destructive and damaging. In this case, there were multiple unsuccessful attempts to replicate it, leading to a massive waste of limited resources.
3
Show this thread
The president of , Marc Tessier-Lavigne, was lead author on multiple papers w obvious image manipulation. It doesn't matter if he knew about the fraud or not. As a co-author, it was part of his responsibility to catch it, and he should be punished for not catching it.
1
3
Show this thread
Excellent article by : "Who Is Responsible For Research Fraud Anyway?" You would think that everyone on a paper should be held reponsible when there is fraud inside. But this isn't how things work in science right now! Lame & disappointing..
1
1
Show this thread
A couple of weeks ago I wore a skirt for the first time, to a contra dance event. I just want to say that skirts are really comfortable and men can wear them too. Unfortunately it's kind of hard to find skirts with more masculine/boy color schemes/appearance/vibe that aren't… Show more
Image
14
37
Show this thread
To elaborate more, there are definitely conflicting results here.. but the popular narrative that on average people continually become happier as they get older seems pretty suspect. It seems overall people’s happiness & reported life satisfaction declines during mid-life, then… Show more
4
10
Show this thread
#QuoteOfTheDay “Do we search for stasis—a regulated, engineered world? Or do we embrace dynamism—a world of constant creation, discovery, and competition? Do we value stability and control, or evolution and learning” - Virginia Postrel in "The Future and Its Enemies"
Image
1
18
Show this thread
Given the continual attacks on Fauci (mostly from the right), there has been a counter-narrative that Fauci was a defender of science who stood up to Trump and stated the truths nobody around him wanted to hear. If you believe that, I recommend reading this article:… Show more
3
I am holding onto my $ASTR stock! I still believe in this company's team and their strategy (mass assembly of simple, cheap, disposable rockets).
Quote Tweet
Quick summary of what we unveiled at @Astra Spacetech Day 2023. Watch the full presentation here: youtube.com/watch?v=OyN37-
Embedded video
1:33
4.3K views
1
1
Show this thread
People seem to have a hard time accepting that some breeds are more aggressive, despite massive amounts of data showing this is so (attached is data on dog bites in New York City) and ignoring the long history of barbaric practices used to select and breed these dogs.
Image
1
4
Show this thread
This is one of those issues people have trouble accepting--intrinsically dangerous & aggressive dog breeds, in particular pit bulls. They were bred to be that way! Pit bulls should be banned in the US-- they are already banned in the UK. I was bit by a dog when a kid-- not fun!
Quote Tweet
Last year 10 people died to dogs in the UK - the highest in 4 decades. This year there's already been another 5 deaths. Many of the dead are children. One breed is behind of nearly all these attacks - we need to stop pretending that that doesn't matter. lawrencenewport.co.uk/p/why-are-so-m
Show this thread
1
10
Show this thread
(2/7) Concept erasure is an important tool for fairness, allowing us to prevent features like race or gender from being used by classifiers when inappropriate, as well as interpretability, letting us study the causal impact of features on a model's behavior.
Image
5
54
Show this thread
#HealthRamble I know 2-3 months ago I said I had recovered from #LongCOVID .. actually I'm not entirely sure if that's accurate. Basically I have been completely crashed 1-2 days a week the past several months, but more or less OK the other days. Last week I had some really good… Show more
3
6