Karl Cobbe @karlcobbe Twitter profile

Last Seen Profiles

@Butterflyy1026_

@videobokep_oi

@TOstendorf11987

@1361481ant

@louvrmn

@FORTCUKAMIL

@kamo_books

@stw_pdg

@1361481ant

@BinorRaja

@couple_muda69

@ibireme

@VegasLineReader

@Baturr017

@BinorRaja

@jxwjeon

@AdvMasrat

@1361481ant

@BinorRaja

@kristygnzz

@SwifRosa

@notsafeforschaf

@bokeplokalmalam

@cukienaknikmati

@1911eLOVES

@BAAPSMedia

@bunda_tiwok

@portoquinza

@bunda_tiwok

@Umar_Refa

@viajefilos

@savahannaISme

@cukienaknikmati

Karl Cobbe

@karlcobbe

3 years

Excited to share our recent work @OpenAI : training large language models (like GPT-3) to solve grade school math problems much more effectively!

9

60

314

Karl Cobbe

@karlcobbe

4 years

Excited to share our recent work on Phasic Policy Gradient, a new RL algorithm which improves sample efficiency by performing policy optimization and auxiliary optimization in two alternating phases. Check out the paper and code!

Phasic Policy Gradient

We introduce Phasic Policy Gradient (PPG), a reinforcement learning framework which modifies traditional on-policy actor-critic methods by separating policy and value function training into...

arxiv.org

2

58

228

Karl Cobbe

@karlcobbe

1 year

We're thrilled to release our latest work on the Mathgen team @OpenAI ! We show that process supervision (step-by-step feedback) is much more effective than outcome supervision at training LLMs to solve challenging math problems. This could be good news for AI alignment!

OpenAI

@OpenAI

1 year

We trained an AI using process supervision — rewarding the thought process rather than the outcome — to achieve new state-of-art in mathematical reasoning. Encouraging sign for alignment of advanced AIs: …

449

846

5K

6

10

150

Karl Cobbe

@karlcobbe

2 months

Very excited to share our latest work @OpenAI teaching models how to solve hard reasoning tasks!

Learning to Reason with LLMs

We are introducing OpenAI o1, a new large language model trained with reinforcement learning to perform complex reasoning. o1 thinks before it answers—it can produce a long internal chain of thought...

openai.com

7

18

139

Karl Cobbe

@karlcobbe

6 years

Excited to share the work I've done during my @OpenAI Fellowship! Using a new procedurally generated environment called CoinRun, we measure how well trained agents can generalize to new environments:

1

30

128

Karl Cobbe

@karlcobbe

4 years

My team at OpenAI is co-organizing two NeurIPS competitions this year using some of our most compelling RL environments: Procgen Benchmark and MineRL. I'm excited for the community to contend with these challenging competitions and advance the state-of-the-art!

OpenAI

@OpenAI

4 years

We're co-organizing two NeurIPS 2020 competitions using Procgen Benchmark and MineRL. We rely heavily on these environments internally for RL research, and we look forward to seeing the progress the community makes in these challenging competitions.

10

116

402

2

23

107

Karl Cobbe

@karlcobbe

1 year

OpenAI is nothing without its people

1

11

101

Karl Cobbe

@karlcobbe

1 year

❤️

Sam Altman

@sama

1 year

i love the openai team so much

5K

4K

72K

2

6

97

Karl Cobbe

@karlcobbe

1 year

Grateful to @satyanadella and @kevin_scott for their steadfast support of the OpenAI team these past days!

1

2

47

Karl Cobbe

@karlcobbe

5 years

Following our work last year on CoinRun, we've designed 15 new procedurally-generated environments to improve our understanding of generalization in reinforcement learning. Check them out!

OpenAI

@OpenAI

5 years

We're releasing Procgen Benchmark, 16 procedurally-generated environments for measuring how quickly a reinforcement learning agent learns generalizable skills. This has become the standard research platform used by the OpenAI RL team:

51

358

991

0

11

48

Karl Cobbe

@karlcobbe

1 year

Excited beyond words to get back to work with @gdb and @sama ! This team is unstoppable

0

1

48

Karl Cobbe

@karlcobbe

1 year

❤️

OpenAI

@OpenAI

1 year

We have reached an agreement in principle for Sam Altman to return to OpenAI as CEO with a new initial board of Bret Taylor (Chair), Larry Summers, and Adam D'Angelo. We are collaborating to figure out the details. Thank you so much for your patience through this.

6K

13K

66K

1

2

35

Karl Cobbe

@karlcobbe

5 years

I'll be speaking at the Deep Reinforcement Learning Summit in SF today, at 2:35pm! Excited to talk about our work @OpenAI in quantifying generalization in deep RL. #reworkDL

0

1

14

Karl Cobbe

@karlcobbe

1 year

❤️

Ilya Sutskever

@ilyasut

1 year

There exists no sentence in any language that conveys how happy I am:

987

521

11K

1

0

10

Karl Cobbe

@karlcobbe

5 years

Very excited to be among the mentors to the next class of OpenAI Scholars!

OpenAI

@OpenAI

5 years

Now accepting applications for our 3rd class of OpenAI Scholars: a 4 month full-time program for individuals from underrepresented groups to study deep learning and produce an open-source project. Mentors include @mcleavey , @karlcobbe , @AlecRad :

24

125

274

0

1

9

Karl Cobbe

@karlcobbe

3 years

@ronbodkin @GretchenMarina @OpenAI We did try to generate problems like these programmatically! Turns out it's still quite hard. We couldn't get anywhere near as much diversity as the human written problems, so the task was much less interesting.

0

1