@karinanguyen_ Profile picture

Karina Nguyen

@karinanguyen_

AI research & eng @OpenAI, prev. @AnthropicAI, @nytimes, @square, @dropbox + visual forensics for the Pulitzer Prize investigations

Similar User
lmarena.ai (formerly lmsys.org) photo

@lmarena_ai

Anthropic photo

@AnthropicAI

Latent.Space photo

@latentspacepod

Abhi Venigalla photo

@ml_hardware

Div Garg photo

@DivGarg_

Dust photo

@dust4ai

Sanchit Gandhi photo

@sanchitgandhi99

Eric Zhang photo

@ekzhang1

Percy Liang photo

@percyliang

Georgi Gerganov photo

@ggerganov

Yi Tay photo

@YiTayML

Barret Zoph photo

@barret_zoph

Hyung Won Chung photo

@hwchung27

Yao Fu photo

@Francis_YAO_

Ethan Perez photo

@EthanJPerez

Pinned

For the first time we are fundamentally changing how humans can collaborate with ChatGPT since it launched two years ago. We’re introducing canvas, a new interface for working with ChatGPT on writing and coding projects that go beyond simple chat. Product and model features:…


it is not an AGI if it a) can't write award winning novels b) doesn't talk like a god


Dreams come true! Thank you so much. Back in 2014 in Ukrainian high school, I learned a lot from Coursera, including English and programming. So grateful for this opportunity to do this 💛

Chatting with OpenAI’s @karinanguyen_ who joined OpenAI earlier this year and within 6 months co-created and shipped Canvas. I really respect teams that can move fast. That OpenAI, even as a large-ish company, can ship at this pace is fantastic!

Tweet Image 1


only OGs will remember the creativity of claude 1.3, the claude instant 1.2 price-performance, the magic of 100k context, edit toggle in claude.ai interface to modify model's outputs for devs, inability to stop claude's streaming bc of silly next.js, long lived…


Karina Nguyen Reposted

Excited to open-source a new hallucinations eval called SimpleQA! For a while it felt like there was no great benchmark for factuality, and so we created an eval that was simple, reliable, and easy-to-use for researchers. Main features of SimpleQA: 1. Very simple setup: there…

Tweet Image 1

New paper! SimpleQA is new factuality benchmark that contains 4,326 short, fact-seeking questions that are challenging for frontier models. Designing good evals is hard. Here we used the following criteria: - High correctness via robust data quality verification / human…

Tweet Image 1
Tweet Image 2

Factuality is one of the biggest open problems in the deployment of artificial intelligence. We are open-sourcing a new benchmark called SimpleQA that measures the factuality of language models. openai.com/index/introduc…



we’re in an intelligence overhang—the problem now is scaling and distribution (eg not all developers have truly embraced in-context learning magic imo). it will take time, but people are underestimating how profound the consequences will be


Show changes feature now in canvas! You can now visually see yours or model's code diffs and edits :)

When using canvas, you can now see what's changed in your writing and code by selecting the "Show changes" icon. Enjoy!

Tweet Image 1


insane 5 months at oai not going to lie, 1 month == 1 yr in these labs


fun to look back at old explorations, some dreams can come true!

Tweet Image 1
Tweet Image 2

I oftentimes use language models to brainstorm and think through ideas during the writing process to get unstuck or find different perspectives. So, I built a writing environment with GPT-3 integrated directly into the editor called Synevyr.



Autocompleting a human’s thought during creative processes like writing or coding is the ultimate personalization — other efforts that people claim as “personalization” is a waste of time


Karina Nguyen Reposted

OpenAI’s canvas seems to really understand the writing process. How is this possible? I recorded myself trying it out at writing for the first time and it’s insane how it figures out really quickly what I wanted to improve in the writing. I’ve been writing for a long time and…


Karina Nguyen Reposted

New is always better


Loading...

Something went wrong.


Something went wrong.