norvid_studies (@norvid_studies) 's Twitter Profile
norvid_studies

@norvid_studies

charts & graphs // ecology, macrohistory, evolution, complex adaptive systems, concepts & models

norvidshoshin.wordpress.com/blog/

ID: 977198947159638016

linkhttps://www.goodreads.com/review/list/21739657-ne calendar_today23-03-2018 15:02:39

93,93K Tweet

3,3K Followers

960 Following

Low fat sweaty cat 🐬/acc (@silicon_alien) 's Twitter Profile Photo

was recently bitten by horseflies. pain was suprisingly intense but shortlived with no itch. it made me wonder why horseflies and mosquitoes despite having the same lifestyle and ancestry are morphologically so different.

was recently bitten by horseflies. pain was suprisingly intense but shortlived with no itch. it made me wonder why horseflies and mosquitoes despite having the same lifestyle and ancestry are morphologically so different.
doomslide (@doomslide) 's Twitter Profile Photo

There are many. The most mundane overlooked bottleneck is probably context length. Hundred of billions of dollars deep in the tech tree and we still can't reliably stretch capabilities beyond ~100K tokens without substantial perf degradation. Lesson in that(?).

Low fat sweaty cat 🐬/acc (@silicon_alien) 's Twitter Profile Photo

norvid_studies the use of cat like and dog like are mainly historical. i would almost never pay attention to such namings, e.i. not all bristleworms have bristles not all carnivora are carnivores. also dog-like morphologies show up among feliformes and cat-like ones among carniformes

Greg Burnham (@greghburnham) 's Twitter Profile Photo

Rookie wh Jasper Dekoninck Lots of good answers already. I'll just add: P6 this year is very hard. A coach for the US IMO team rated it as one of the three hardest IMO problems since 2000. So, "why LLMs fail" might be overdetermined. 1/ web.evanchen.cc/upload/MOHS-ha…

Greg Burnham (@greghburnham) 's Twitter Profile Photo

Rookie wh Jasper Dekoninck It's also the kind of problem we've seen all AI systems struggle with: combinatorics with a geometric flavor. These problems are often a bit off-the-beaten-path. On this one you can use some known theorems on some steps, but the overall approach requires a bit more creativity. 2/

Low fat sweaty cat 🐬/acc (@silicon_alien) 's Twitter Profile Photo

norvid_studies very difficult to say, but arthropods have a printer-like style of limb generation during embryogenesis, its easy for them adjust the number of segments and among dozens of legs some are free to mutate since there is functional redundancy

Low fat sweaty cat 🐬/acc (@silicon_alien) 's Twitter Profile Photo

norvid_studies vertebrates also have this printer style development, but for the spinal cloumn and there is actual heterogeneity in the number of vertebrae animals have. if you want to read about it the keyword is somitic clock

Jacob Witten (@jacobwitten1) 's Twitter Profile Photo

norvid_studies Affinity maturation, multiple exposures to diverse antigen can select for broadly neutralizing antibodies People saying V(D)J recombination are incorrect, affinity maturation is largely driven by somatic hypermutation of already-VDJ-recombined B cells

Jacob Witten (@jacobwitten1) 's Twitter Profile Photo

norvid_studies This doesn't usually work, usually you get a polyclonal response against a bunch of different epitopes (this is driven by VDJ recombination) not one super-antibody. This is why people have tried and failed for decades to elicit broadly neutralizing anti-HIV antibodies w/a vaccine