𝔊𝔴𝔢𝔯𝔫@gwernMay 31GPT-3 is terrifying because it's a tiny model compared to what's possible, trained in the dumbest way possible on a single impoverished modality on tiny data, yet the first version already manifests crazy runtime meta-learning—and the scaling curves 𝘴𝘵𝘪𝘭𝘭 are not bending! 😮 twitter.com/hardmaru/statu…
142,940
7,389
5.2%
View Tweet activity
𝔊𝔴𝔢𝔯𝔫@gwernMay 4Q: are there any good nonfiction (or fiction) speculations about what a society of human clones might be like? All the depictions are really dumb (Brave New World), small-scale (a few clones), or assume the radically different scenario of mind-copying (Kiln People, Age of Em).
𝔊𝔴𝔢𝔯𝔫@gwernMay 2You can legally do it reliably and painlessly for animals, but only legally do it unreliably and extremely painfully for humans.
What activity is this?"
(And yes, I did check the US bestiality/necrophilia laws, that's how it is.)
49,212
576
1.2%
View Tweet activity
𝔊𝔴𝔢𝔯𝔫@gwernMay 7Here's another idea, also related to GB work, funnily enough: can we use contrastive losses, inspired by SimCLR, to replace the adversarial part of GANs entirely?
D gets SimCLR, and G uses _z_ noising to generate positive samples, and that's it: github.com/tensorfork/ten…
44,380
83
0.2%
View Tweet activity
𝔊𝔴𝔢𝔯𝔫@gwernMay 2Some feel ethically required to do it, while others feel required not to.
It is ethical to do it to an animal, but not to your grandparents.
It's acceptable to do it for yourself or family, but not strangers; however, it's acceptable for strangers to do it for strangers/you....
𝔊𝔴𝔢𝔯𝔫@gwernMay 2A riddle from chat today:
"You can do this activity at the beginning or end of a horse race but not the middle.
This activity is acceptable to do in public but unacceptable in your bedroom.
It is illegal to have sex with an animal before this activity, but legal afterwards...
𝔊𝔴𝔢𝔯𝔫@gwernMay 1"What did I expect to see 𝘣𝘦𝘧𝘰𝘳𝘦 I saw X?"
(And yes, you always expected 𝘴𝘰𝘮𝘦𝘵𝘩𝘪𝘯𝘨: because of meta-analyses, replication rates, evolution, efficient markets, equilibriums, Fermi estimates, decision-making, physical constraints, everything is correlated...)
𝔊𝔴𝔢𝔯𝔫@gwernMay 161982, EA: "Can A Computer Make You Cry?"
2020, Sonantic sonantic.io: Can you make a computer cry?
(Also I kinda hate that I can stare at a blurry monitor & instantly know he's ssh'ed into a Ubuntu cloud instance running Tensorboard on a TF S2S spectrogram NN.)
𝔊𝔴𝔢𝔯𝔫@gwernMay 19New datapoint: arxiv.org/abs/2005.06840 Similar to Mozilla, adblocker-installation as quasi-experiment (so probably overestimate):
"...Users who adopt ad blockers subsequently consume 20% more news articles corresponding to 10% more categories. The effect persists over time..."
26,012
91
0.3%
View Tweet activity
Get your Tweets in front of more people.
Use Tweet Activity to track how your Tweets are doing.
![](data:image/svg+xml,<svg xmlns="http://www.w3.org/2000/svg" width="128" height="128"><rect fill-opacity="0"/></svg>)
