Ought (@oughtinc) header
Ought (@oughtinc) profile picture

Ought

@oughtinc

Automate & scale open-ended reasoning Building @elicitorg, an AI research assistant Demos https://nitter.gq/t.co/8UGlbJ1eVE Jobs & product roadmap https://nitter.gq/t.co/oDFh35ACEG

 San Francisco, CA
ought.org
Joined April 2016
  • Tweets266
  • Following18
  • Followers2.3K
14Media
Photo by Ought (@oughtinc)Photo by Ought (@oughtinc)Photo by Ought (@oughtinc)Photo by Ought (@oughtinc)Photo by Ought (@oughtinc)Photo by Ought (@oughtinc)Photo by Ought (@oughtinc)Photo by Ought (@oughtinc)Photo by Ought (@oughtinc)Photo by Ought (@oughtinc)
ought.org/updates/2023-09-25…
Ought has spun off Elicit | Ought

Ought has spun off Elicit | Ought

Elicit is now an independent public benefit corporation and has raised $9 million in venture funding

ought.org
5
25
2,565
ought.org has a new look and feel now thanks to @Mappletons!
Image from Ought (@oughtinc)
4
1
29
7,465
We think that building products and testing out ideas is one of the best ways to have an impact on AI safety. This is what we're doing at Ought! Take a look at our recent post to learn more: forum.effectivealtruism.org/…
AI Safety Needs Great Product Builders

AI Safety Needs Great Product Builders

In his AI Safety Needs Great Engineers post, Andy Jones explains how software engineers can reduce the risks of unfriendly artificial intelligence. E…

forum.effectivealtruism.org
2
8
Welcome to the team Maggie! We’re thrilled to start building with you.
Maggie Appleton (@Mappletons) profile pictureMaggie Appleton@Mappletons15 Jul 2022
Today was my last day @hashai. I still support the team and everything they're working on ❤️ But I'm thrilled to be joining @oughtinc I wrote a bit about the move and Ought here: maggieappleton.com/joining-o… TLDR: Designing AI tools to help researchers with open-ended reasoning
Joining Ought

Joining Ought

A new role at an AI research lab working on tools for open-ended reasoning

maggieappleton.com
5
Lead devops at Ought and build infrastructure for composing language model tasks at scale -
Image from Ought (@oughtinc)
Andreas Stuhlmüller (@stuhlmueller) profile pictureAndreas Stuhlmüller
@stuhlmueller22 Jun 2022
Letter to our future devops lead docs.google.com/document/d/1…
Letter to our future devops lead

Letter to our future devops lead

I’m writing on behalf of Ought, a machine learning lab. We’re building the AI research assistant Elicit and need you to lead devops at Ought. Here’s what I’m hoping we can do together: On your first...

docs.google.com
5
We're creating a dataset of alignment research tasks (ART 🎨) to guide AI research towards helping with alignment. Each contribution is eligible for a $200-$2000 prize!
1
5
2
25
One thing we realized at the Tahoe retreat is that our team cooks & cleans really well together. We seamlessly danced around and with each other while wielding knives, stoking fires, and overcoming fire alarms. youtu.be/nIzLbT6Yz5k
1
3
Excited to welcome James Brady as our new Head of Engineering! jmsbrdy.com/
Image from Ought (@oughtinc)
1
2
7
Thank you SFF & Jaan Tallinn for your support. survivalandflourishing.fund/…
Image from Ought (@oughtinc)
1
17
Kudos @timo_schick!
Andreas Stuhlmüller (@stuhlmueller) profile pictureAndreas Stuhlmüller
@stuhlmueller8 Nov 2021
New leader on raft.elicit.org benchmark: @timo_schick with PET, beating GPT-3. I'm skeptical of much few-shot work (easy to overfit), but RAFT makes it hard to cheat + has real world tasks so seems PET is actually best for few-shot classification github.com/timoschick/pet
RAFT | Real-world Annotated Few-shot Tasks

RAFT | Real-world Annotated Few-shot Tasks

RAFT is a few-shot classification benchmark that tests language models across multiple domains on economically valuable classification tasks with evaluation that mirrors deployment.

raft.elicit.org
Thank you @William_Kiely!
Will Kiely (@William_Kiely) profile pictureWill Kiely@William_Kiely2 Nov 2021
Please join me in supporting @oughtinc. Double your gift with the #FallGivingChallenge and get matched up to $100 by donating today! every.org/ought?donationId=4… via @everydotorg
Ought Inc | Every.org

Ought Inc | Every.org

every.org
1
We now accept donations in cryptocurrency! Help us apply advanced ML systems and language models to complex, open-ended reasoning tasks: ought.org/donate Currently building @elicitorg
1
11
This week's Lunch & Learn: @stuhlmueller and @EthanJPerez answer crowd sourced questions about the basics of language models and the most interesting developments they expect to see in the next 5-10 years. Video: youtube.com/watch?v=v35FrIhG… Notes: docs.google.com/document/d/1…
Lunch & Learn | Intro to NLP and language models with Ethan Perez

Lunch & Learn | Intro to NLP and language models with Ethan Perez

Andreas Stuhlmueller (Ought Cofounder) and Ethan Perez (NLP PhD stu...

youtube.com
4
2
12
We're beta testing research assistant features in @elicitorg using GPT-3. You can now: 1. Find relevant think tank publications
1
4
4
30
2. Search forecasting questions across @CSETforetell, @metaculus, @predictit
1
1
8
Owain Evans (@OwainEvans_UK) profile pictureOwain Evans
@OwainEvans_UK20 Nov 2020
Embed forecasts in LessWrong posts: from @oughtinc and @wrong_less. lesswrong.com/posts/JLrnbThM…
Embedded Interactive Predictions on LessWrong — LessWrong

Embedded Interactive Predictions on LessWrong — LessWrong

Ought and LessWrong are excited to launch an embedded interactive prediction feature. You can now embed binary questions into LessWrong posts and com…

lesswrong.com
1
4
Our vision for automating open-ended reasoning about the future in our latest blog post: ought.org/updates/2020-10-09… cc @elicitorg
1
4
2
14
AGI seems to be the dominant consideration in a lot of LessWrong x-risk predictions. Is this actually accurate, or is it a community bias? You can make your own prediction here: tinyurl.com/x-risk-predictio…
1
We’re excited to announce $3.8 million in new donations from partners including @open_phil, @paulfchristiano, Jaan Tallinn, Survival and Flourishing Fund and many other amazing donors. ought.org/updates/2020-02-14…
Ought Raises $3.8 Million | Ought

Ought Raises $3.8 Million | Ought

We’re excited to share that we've raised $3.8 million in donations

ought.org
1
10
Updates on factored evaluation experiments discerning the validity of arguments about Roger Ebert movie reviews ought.org/updates/2020-01-11…
Evaluating Arguments One Step at a Time | Ought

Evaluating Arguments One Step at a Time | Ought

A technical report on our experiments testing factored evaluation of structured arguments.

ought.org
2
1
3
Over the past 10 months, we identified a new setup for question decomposition experiments, tested it with 60+ contractors working 150+ hours per week, and laid the groundwork for an app that supports many different kinds of question-answer experiments. ought.org/updates/2019-10-28…
1
1
Load more