Joyful complex work
https://cutlefish.substack.com/p/tbm-3952-joyful-complex-work https://twitter.com/grimalkina/status/1561035008554348547 Doing research with developer teams, something that really strikes me is how much people look for ways to make complex problems easy rather than make it easy to work on complex problems What I mean is, sometimes it’s really useful to admit we just are trying to accomplish...
How GPT-3 helps me write short stories
https://storiesby.ai/p/ai-is-my-co-writer https://www.sudowrite.com/ What if we can play between these two extremes, and explore the goofy and creative side of AI’s emerging capabilities for writing fiction and generating art? As an AI engineer/researcher and creative writer, I’m fascinated by the ways that incorporating AI into the creative process can change how...
How to measure cohort retention
https://www.lennysnewsletter.com/p/measuring-cohort-retention Step 1. Define “active” Most companies use logins or app opens as main events for the “active users” definition. But given that I always aim for the cleanest and most precise data reporting, I’d recommend using the main user action as the activity event, for example: Pros to this...
Gender-diverse teams produce more novel and higher-impact scientific ideas
https://www.pnas.org/doi/10.1073/pnas.2200841119 Science’s changing demographics raise new questions about research team diversity and research outcomes. We study mixed-gender research teams, examining 6.6 million papers published across the medical sciences since 2000 and establishing several core findings. First, the fraction of publications by mixed-gender teams has grown rapidly, yet mixed-gender teams continue...
PEER: A collaborative language model
https://ar5iv.labs.arxiv.org/html/2208.11663 Text from current language models can be useful as a rough draft, but that leaves the polishing to human writers. A language model learned how to generate and respond to editorial directions. What’s new: Timo Schick and colleagues at Meta proposed Plan, Edit, Explain, and Repeat (PEER), a text...
Engineering levels at Honeycomb: Avoiding the scope trap
https://www.honeycomb.io/blog/engineering-levels-at-honeycomb/ It has been seven years since Rent the Runway posted their engineering ladder, kicking off a veritable trend of engineering teams open sourcing their ladders. Interestingly, nearly all of them seem to have coalesced around “area of scope” as a useful proxy for level. While talking about scope, one...
Talking with strangers is surprisingly informative
https://www.pnas.org/doi/10.1073/pnas.2206992119 A meaningful amount of people’s knowledge comes from their conversations with others. The amount people expect to learn predicts their interest in having a conversation (pretests 1 and 2), suggesting that the presumed information value of conversations guides decisions of whom to talk with. The results of seven experiments,...
Atlas: Few-shot learning with retrieval augmented language models
https://arxiv.org/abs/2208.03299 https://github.com/facebookresearch/atlas Perhaps a good approach for fact checking? A large language model uses its huge complement of parameters to memorize information contained in its pretraining and fine-tuning datasets. It wouldn’t need to memorize so much — and thus wouldn’t need so many parameters — if it had access to...
Preventing soft skill decay among early-career women in STEM during COVID-19: Evidence from a longitudinal intervention
https://www.pnas.org/doi/10.1073/pnas.2123105119 As the workforce shifts to being predominantly hybrid and remote, how can companies help employees—particularly early-career women in science, technology, engineering, and mathematics (STEM) fields—develop greater confidence in their soft skills, shown to improve organizational retention? We evaluate the effects of an online longitudinal intervention to develop soft skills...
TextWorldExpress: Simulating Text Games at One Million Steps Per Second
https://ar5iv.labs.arxiv.org/html/2208.01174 https://github.com/cognitiveailab/TextWorldExpress SayCan 같이 로봇 가르치는데 이런게 필요한게 아닐까? Text-based games offer a challenging test bed to evaluate virtual agents at language understanding, multi-step problem-solving, and common-sense reasoning. However, speed is a major limitation of current text-based games, capping at 300 steps per second, mainly due to the use of...
About
a collection of excerpts from interesting research, blogs, events, etc...