52% are incorrect or 52% contain inaccuracies?
What does 'incorrect' even mean when referring to an LLM output?
Posts
Filter by tag
hallucinationsocial-search-requestPerplexity-AIcomparing-resultsgenerative-searchdissertationprompt engineeringaiextensionsjagged-frontierseries/extended-frontierworking-draftsearch-auditsmicro interactions in searchYou.comspeculative-designGoogle-BardCGTbetter-searchcompetition-in-searchchoice-in-searchgenerative-web-searchknowledge-cutoffsalternative-search-enginesevaluating-results-metasearch-quality-complaintsstanding-queriessearch-engine-optimizationrepository-of-examplesAnthropic-ClaudeCFP-RFPRAGPhinddoubtingadvertising-in-searchseo-for-social-goodOpenAI-ChatGPTsituated-practicepublic-interest-technologyspeculative-classroom-exercisesTavilysearch-user-interfacesgenerative-search-APIfresh-searchautocomplete-in-web-searchfalse-premiseTikTokAndievaluating-search-enginesarticulation+234 more
"We also asked participants about their ChatGPT expertise"
Excerpt from a study on programming expertise showing participants self-reported as mostly proficient programmers but only competent in ChatGPT usage, revealing a skill gap between traditional coding and AI tool proficiency.
Heteromation astride 'automation v. augmentation'
Brief post positioning heteromation concept alongside the automation versus augmentation debate, referencing Ekbia's work.
How do people perceive and perform-with tool outputs?
To better understand and address tool use, we need to understand not 'accuracy' but interaction.
Rethinking Rethinking Search: Making Domain Experts out of Dilettantes
Analysis of Metzler et al.'s 'Rethinking Search' paper and its framing of LLMs as dilettantes, reframing the concept to consider how human dilettantes might interact with or as domain experts in search practice.
a toothpick, a bowl of pudding, a full glass of water, and a marshmallow
ChatGPT-4 seems able to now adequately address this specific stacking challenge?
What if general-purpose web search were more like searching for directions?
Thinking about what if general-purpose web search were more like searching for directions...
"Google the Gatekeeper: How Search Components Affect Clicks and Attention"
Are prompts—& queries—not Lipschitz?
A tweet from @zacharylipton: Prompts are not Lipschitz. There are no “small” changes to prompts...
Data Voids and the Google This Ploy
Thinking about comparing and contrasting 'prompt injection' and the 'Google This Ploy' (Caulfield, 2019)
But how do we ground our relations to hallucination?
Thinking about hallucination with Klosterman, Leahu, Munk et al., Rettberg, and Powles & Nissenbaum.
Keyword search is dead?
"However, we often consult Google about topics we might not know enough about..."
"For if a reader goes to a book searching for new ideas..."
OWASP Top 10 for Large Language Model Applications
Here is the 'OWASP Top 10 for Large Language Model Applications'. Overreliance is relevant to my research. (I’ve generally used the term “automation bias”, though perhaps a more direct term like overr...
hints to jump to search
Logan Kilpatrick just announced search in OpenAI's developer docs so I glanced at it and saw the hint to jump to the search bar (their DocSearch-Button-Keys): ⌘ & K. I appreciate such micro interactio...
My academic search kit
Collection of tweets describing academic search workflow using Twitter, Semantic Scholar, local whoosh index, and various citation tools while exploring alternatives to Google Scholar.
academic search tools
This is an incomplete listing of a few academic search tools. See also: - my paper w/ Jake Goldenfein (Google Scholar – Platforming the scholarly economy (goldenfein2022platforming)) - my scholar prof...
"A good search engine advises its users..."
they answered the question
This is partially about prompt engineering and partially about what a good essay or search does. More than answer a question, perhaps? (this is engaged with in the essay, though not to my liking). Gri...