Shreya Shankar
Computer scientist working on LLMs in data systems; computer science PhD at UC Berkeley, creator of DocETL, co-teacher of an AI evals course with Hamel Husain, and incoming assistant professor at Carnegie Mellon.
Shreya Shankar did not write this page.
We collected these quotes from things they published elsewhere, and every quote links to where it was said. They have no account here and have not endorsed this site. Quotes are word for word; the short line under each one is our own restatement, not their wording. Their own site. Is this you? Claim it or ask us to remove it. Or tell us what is wrong here.
Filter All topics · Newest first
-
Our reading
AI has overused the rhetorical tools meant to aid comprehension until readers stop noticing them.
Their wordsAI has overused the tools designed to aid human comprehension to the point where I’ve stopped noticing them.
↗On the Consumption of AI-Generated Content at Scalesh-reya.com 1st of 3 in this piece
-
Their wordsFor complex tasks, in the pre-LLM era, generating output was expensive but verification was comparatively cheap. You put in the work to produce something—a document, a piece of code, an analysis—and checking it was tractable.
↗On the Consumption of AI-Generated Content at Scalesh-reya.com 2nd of 3 in this piece
-
Our reading
If readers cannot comprehend complex ideas or notice errors, they become easier to manipulate.
Their wordsFirst, if consumers can’t comprehend complex ideas or notice errors, they’re easier to manipulate.
↗On the Consumption of AI-Generated Content at Scalesh-reya.com 3rd of 3 in this piece
- 6 months earlier
-
Their wordsEmpty summary sentences feel conclusive, but say nothing.
↗Writing in the Age of LLMssh-reya.com 1st of 3 in this piece
-
Their wordsLLMs often overuse bullet points, especially nested ones. Lists help when items are parallel and independent, but when ideas are connected or need context, a paragraph is usually better.
↗Writing in the Age of LLMssh-reya.com 2nd of 3 in this piece
-
Our reading
LLMs cannot tell what knowledge the reader already has, so a human writer has to fill the gap.
Their wordsIn summary, LLMs can’t reliably distinguish what’s assumed knowledge and what needs explanation, so they often gloss over the hard parts, and a human writer has to fill that gap.
↗Writing in the Age of LLMssh-reya.com 3rd of 3 in this piece
- 12 months earlier
-
Their wordsI think, in ML and AI engineering, that teams often expect too high of accuracy or alignment with their expectations from an AI application right after it’s launched, and often don’t build out the infrastructure to continually inspect data, incorporate new tests, and improve the end-to-end system. Then, over time, they only get more and more disappointed.
↗Short Musings on AI Engineering and "Failed AI Projects"sh-reya.com