omnivore. also at https://xitter.cf/t.co/3Q36jNfaxj

nyc
Joined June 2014
I have always believed that art is a driving force behind society's evolution. At @betaworks, we’ve long championed not only coders but also artists, designers, and creatives of all kinds. Our demo nights are a celebration of this synergy - a space where art, design, and code stand as equals, shaping the future together. This Thursday, we invite you to join us for an evening of discovery and inspiration. Meet the visionary artists and engineers who are redefining what’s possible, pushing boundaries, and helping society move forward. Together, let’s explore the ideas that will shape tomorrow. We'll be welcoming some new names to the betaworks stage like @jediahkatz @DayOfLane @juanbfiguera, Octavio Figueroa and welcoming back @willrob4 @yiliu_shenburke - luma in comments!
5
4
1
45
4,141
recsys products are starting to feel a little stodgy and inflexible in the era of genAI discover weekly needs a “temperature” setting for the nerds
1
1
505
There is this widely-held belief that current models are magic, instead of systems: 1. Millions of people generating training data 2. People aligning models through supervised RHLF / constitutional ai 3. Human filters for model outputs 4. People building human-centric UIs
8
68
4,976
This is your periodic reminder that user interfaces are important, and text is a good lowest common denominator, not the endgame. The world and our senses have a lot more to offer.
15
27
3
374
57,902
A cognitive bias that makes large language models seem even more impressive than they are is that our brains, which are far more impressive than LLMs, decode their ungrounded outputs into grounded representations, and this happens reflexively without us noticing.
5
12
1,438
bill retweeted
Sam Altman
@sama11 Dec 2022
ChatGPT is incredibly limited, but good enough at some things to create a misleading impression of greatness. it's a mistake to be relying on it for anything important right now. it’s a preview of progress; we have lots of work to do on robustness and truthfulness.
762
2,999
744
25,393
1/ The *last* thing we need right now is more untested speculation about how to fix social media. We have *research* that can help us evaluate @elonmusk’s proposals to transform Twitter, and many of these studies might inspire him to throw some cold water on his plans 🧵
36
431
89
1,545
“Cities are not loud. Cars are loud.” -Paris edition
71
1,897
264
14,275
I noticed that DistilBERT loves movies filmed in India, but not in Iraq, so I plotted the result for each country: the resulting map is scary. #aibias
54
313
66
1,777
In software development learning is not a big part of the job. It is the job.
10
344
16
1,334
who's with me 🖐️😏 #langtwt #refold
3
24
2
150
When you’re a wordcel with shape rotator tendencies.
3
31
Who feels the power? #langtwt #refold
20
2
123
I have just discovered that several of my books are available for free download at hu1lib.org. Please help yourself and share with others!!!! Just go to the website and type in Krashen.
10
49
6
189
ich und die zwei leute die alle meine posts liken
6
161
9
2,093
proposal for an internet research lab: - study how people learn on the web - build new crawling, search, recommendation algos - browser prototypes/extensions - live streams of going down rabbit holes end goal: ecosystem of tools to help ppl learn/create/chill on the web better
15
6
3
149
schubert was in his bag with this
Schubert died on this day, 193 years ago, so very young. But he lives on for ever and ever through music like this. How lucky we are!! String quintet during @DelftMusicFest with Esther Hoppe, Jennifer Stumm, Ivan Karizna and Jakob Koranyi. #eternity #music
3
15
2
102
This reminds me that INTEREST was a core value of Kató Lomb's methodology for learning languages. She was one of the hyperpolyglots studied by Kr*shen during his development of Comprehensible Input theory. But although she loved reading, she was very clear input was not the key.
100%. Which is why I am against putting the pure consumption itself at the center of any methodology without any discrimination. Consumption is just one tool, not a core value. Something else like interest would be a lot more suitable.
1
1
4
The world if NeurIPS papers were better datasets instead of better models
4
35
4
445
One of my most controversial software opinions is that your sleep quality and stress level matter far, far more than the languages you use or the practices you follow. Nothing else comes close: not type systems, not TDD, not formal methods, not ANYTHING. Allow me to explain why.
140
4,687
786
11,461