raulpuri.eth
@TheRealRPuri
Followers
6,829
Following
329
Media
19
Statuses
432
AI things @ OpenAI - Her, GPT4V, GPT4, GPT3.5, Codex | past: NVIDIA - megatron, sentiment neurons | go bears 🐻
Joined March 2014
Don't wanna be here?
Send us removal request.
Explore trending content on Musk Viewer
Mbappe
• 867743 Tweets
Real Madrid
• 719067 Tweets
Fauci
• 702350 Tweets
Claudia Sheinbaum
• 623573 Tweets
AMLO
• 245864 Tweets
Farage
• 204652 Tweets
River
• 137898 Tweets
WNBA
• 133628 Tweets
Venom
• 130664 Tweets
Reform
• 114435 Tweets
Aziz
• 112981 Tweets
Jadue
• 104904 Tweets
مدريد
• 93360 Tweets
Monica
• 76445 Tweets
Marjorie Taylor Greene
• 71605 Tweets
Acun
• 67112 Tweets
sabrina
• 64625 Tweets
Bienvenido
• 63439 Tweets
Florentino
• 61910 Tweets
Djokovic
• 53387 Tweets
Clacton
• 48921 Tweets
Rodrygo
• 42525 Tweets
Botafogo
• 40067 Tweets
Barriga
• 37000 Tweets
Stephen A
• 32271 Tweets
Larry Allen
• 27924 Tweets
#المقاومه_العراقيه_تدك_حيفا
• 26335 Tweets
Talleres
• 23423 Tweets
Cerundolo
• 18370 Tweets
الريال
• 16414 Tweets
Marge
• 15751 Tweets
RESPIRATOR OUT NOW
• 14375 Tweets
Ducati
• 14135 Tweets
Bosnia
• 13644 Tweets
#loveIsland
• 12314 Tweets
#مبابي
• 10151 Tweets
Fun fact: I’m a terrible web dev. I used gpt-4V at some point to write code for some of its own prototype GUIs.
You can give ChatGPT a picture of your team’s whiteboarding session and have it write the code for you.
This is absolutely insane.
677
5K
31K
22
50
632
We fuckin sent it
Say hello to GPT-4o, our new flagship model which can reason across audio, vision, and text in real time:
Text and image input rolling out today in API and ChatGPT with voice and video in the coming weeks.
3K
14K
65K
17
7
411
We Just released a cool
#PyTorch
#NaturalLanguageProcessing
project we've been working on: training an 8.3B GPT2 model with model parallelism. Check it out...
Details:
Training Code:
3
51
149
Every time I open this image I end up having to play where’s Waldo again to find myself
9
35
112
Team is committed to the mission
6
26
107
Something magical is happening in SF this weekend for sure.
The universe is winking.
not gpt-5, not a search engine, but we’ve been hard at work on some new stuff we think people will love! feels like magic to me.
monday 10am PT.
1K
3K
28K
5
6
99
If you wonder what we’ve been doing with vision since march’s gpt-4 announcement: Safety!
We've included a system card focused on the vision capabilities, building on the work from the GPT-4 system card.
Thank you to all our expert testers and red teamers for helping to inform this work!
4
12
110
15
6
91
Also very big shout out to
@DustMason
for helping us get this out to all you users
It’s been really amazing watching
@TheRealRPuri
work so hard across the whole stack to make image inputs a reality.
Congrats to core contributors
@shgusdngogo
, Jamie Kiros,
@longouyang
,
@daniellevy__
,
@choniong
,
@SandhiniAgarwal
, awesome work all.
11
21
221
8
8
76
Prompt engineering is just technical/document writing for robots instead of humans
Is prompt engineering dead?
No, it’s SoTA.
GPT-4 with good prompts (dynamic k-shot + self-generated CoT + choice-shuffled ensembles) beats Med-PaLM 2 on all nine of the MultiMedQA benchmarks it was fine-tuned for, without fine-tuning:
23
160
1K
4
41
43
Shipping multimodal AI is hard. Urban planning is harder
The most implausible prediction from the movie "Her" is not the AI but high-density walkable Los Angeles.
21
78
2K
1
5
41
Gpt-4V for your dating life
✨ My first 100% ChatGPT startup is now monetized:
❤️🔥 - $9.99 to get a full dating profile review for your Tinder, Bumble or Hinge
The site has NO landing page, which is new for me too, I just use the Stripe payment link as the landing page, which after
175
80
2K
2
33
37
Galaxy brain
CMU ml phd really coming in clutch for building gmail snooze button
never thought i'd be using nonhomogenous poisson processes for email scheduling but 2024 here we come
1
1
9
2
33
36
2
1
31
@NandoDF
I remember our rejections after submitting to sysML. One reviewer thought we weren’t a systems paper. Another reviewer pointed out that our architecture/model wasn’t novel. …… wat
0
0
30
Excited to share our
#acl2020
work on Large Scale Multi-Actor Generative Dialog Modeling, done with amazing coauthors Alex Boyd,
@MostofaPatwary
, Mohammad Shoeybi,
@ctnzr
!
Join our QA session 5-6 UTC tonight and 21-22pm UTC tomorrow.
1
6
24
This was so fun to work on. So proud of the whole team 🎉🥳
0
1
23
Congrats yall !
ChatQA can outperform GPT-4 on a wide range of conversational QA tasks:
- ChatQA and GPT-4 take the same top-5 chunks from our best retriever, when long documents are involved.
- ChatQA performs very well on tabular data, arithmetic calculation, and “unanswerable” cases!
1
21
87
3
9
19
@ilyasut
Apologizing doesn’t mean one is right or wrong, it does mean however that one values their relationships more than their ego.
Feeling valued. ❤️
OpenAi is nothing without its people.
1
0
20
Def need to do a longer list here, but also wanted to shout out
@MadelaineBoyd
for bringing y’all a safe n trustworthy user experience !
1
0
16
It’s amazing to see how much the
#NaturalLanguageProcessing
community has grown in the past few years and the problems we’re tacklin
GPT-2 6-month follow-up: we're releasing the 774M parameter model, an open-source legal doc organizations can use to form model-sharing partnerships, and a technical report about our experience coordinating to form new publication norms:
31
343
945
0
3
17
Stop by our poster "Zero-shot Text Classification With Generative Language Models" today at the
#neurips19
Meta-Learning workshop.
Learn about the intersection of Meta Learning/NLP and some of the challenges going forward.
paper:
1
7
16
@mckaywrigley
@mckaywrigley
hi author here 👋. Can you purposefully try messing up and see if it can correct code based on a screenshot of the messed up deploy. Really curious to see examples of iterative/interactive usage
1
0
14
🥲
ChatGPT is clearly the child of immigrants because they keep forcing it to be a doctor or lawyer when all it wants to do is creative writing and hallucinating
24
134
1K
0
0
13
@soumithchintala
What qualifies as a step function though? Like bigger model/capability numbers go up?
Personally I feel like the GPT-4V api is a step function and will change what devs build, we just promoted content related to changing how devs build instead.
play with it, lmk...
2
1
10
Y’all 👀
I’ll be giving a behind the scenes look at how we scaled ChatGPT at 10:30. Most of this we’ve never talked publicly about before and I’m super excited to share it here at
@TheLeadDev
West Coast.
5
7
113
0
2
11
If any
#NLProc
researcher is thinking of doing this, please don’t be an idiot and read this insightful thread first.
I've seen several different
#NLProc
folks suggesting today that it would fun/interesting/worthwhile to use BERT or GPT-2 to fill in the redacted bits of the Mueller report. A short thread on why this is a terrible idea /1
9
119
346
1
5
10
Perks of working at OAI: cool paper clips
0
0
8
Correct half the time is when something becomes more useful than unuseful. 📈✈️
Everyone's at the starting line of something new and wonderful. ✨💫
🔥 Exciting Update! We've manually evaluated
#GPT4V
using the playground chatbot on
#MathVista
, our newest benchmark for visual mathematical reasoning.
🚀
#GPT4V
soared with a 15.1%⬆️ improvement over
#Bard
, setting a new record at 49.9%! 🎉
🌐
Yet,
3
28
136
0
3
8
@kevin_zakka
Another take - Maybe a healthy way to make an impact on robotics isn’t to work directly on robotics
0
0
7
MEGATRON Update:
4x more data leads to new WikiText and LAMBADA sota.
Check out our updated blog post and
@WeCNLP
poster (
#21
).
#NaturalLanguageProcessing
1
6
7
We need an osmosis jones reboot for COVID.
Thank you for coming to my TED talk.
0
3
7
Academic graduations, other ceremonies, etc. are beautiful. You feel so connected to the fabric of humanity. Wish I could bottle the feeling and share it with others.
0
0
6
@mezaoptimizer
A story from creating gpt-4V:
2 years ago while moving apartments I almost threw away my old college notes. Ppl didn’t understand why I wanted to hold on to them. Back then I knew where we were headed. I knew I’d eventually be able to plug them into an AI capable of reading them.
1
0
6
Aight yall. The team sent it on this one.
Sleepy time now. Will see yall on the other side.
1
0
6
@eukaryote314
~800M gpt2. this is largely due to the memory required to House parameters + optimizer states. If one uses a smaller optimizer than Adam training something larger should be possible. Make sure to turn on activation checkpointing with —checkpoint-activations
0
0
6
@kimmonismus
“All that remains is to wait for the roll-out not to take another 6 months.”
Bruh, like can I get a vacation…
0
0
5
@alexeev_eu
@Miles_Brundage
We’re working on user friendliness before getting to that point. As you can imagine massive models like this aren’t readily accessible to everyone.
1
0
5
Big alpha
BREAKING:
OpenAI and Burger King to merge.
They have tapped the CEO of Mattress Firm to run the combined company.
The former CEO of Mattress Firm (4th CEO of OpenAI in 72 hours) was fired after 7 minutes for Slacking the entire company (the remaining 9 employees) a gif of
131
170
4K
3
0
5
1
0
5
Awesome audio. Awesome authors. All around awesome. 🎶🎵🎶🎵
We just released the paper and code for Mellotron: a multispeaker voice synthesis model that can make a voice emote and sing without emotive or singing training data.
3
119
391
0
1
5
We pretrain large generative language models on multi-actor reddit conversations. Furthermore, by conditioning on users' past reddit conversations we use in-sequence meta learning to control and personalize the dialogue model's response.
1
2
5
Ken griffin buying the constitution is a depressingly poetic reflection of the times: capitalist overlord buys out the us constitution. Orgs like
@ConstitutionDAO
are more necessary than ever. … wagmi
0
0
4
@soumithchintala
I think ppl also underestimate how much harder it is to go from 80->90% on a task than 70->80% on a task. That asymptote at 100% is nasty business. Reaching the asymptote is also necessary for prod reliability. Add the difficulty of generality compounding on top of that as well.
1
1
4
I feel so exposed rn
0
0
4
1
0
4
0
0
4
@RandolphCarterZ
@tszzl
Having been part of both parties’ work… nah. Researchers believe in 99.9% of the words they write. Researchers care deeply about what we work on. We don’t publish something we’re unhappy with. The remaining .1% is disagreement with coauthors, who we ultimately believe in…
0
0
4
@saranormous
Over the last few years we got computers that could use fuzzy logic with discreet inputs. Now we get fuzzy logic and fuzzy inputs. Never before in history have we had a computer that could see
1
0
4
0
0
4
By pretraining the model with a N-way multiple choice title prediction class we can achieve up to a 45% absolute accuracy boost in common downstream text classification tasks.
1
1
3
1
0
3
@agazdecki
@saastr
totally not a founder, just an AI researcher, but would love to join. Would love to see some change in the funding/acquisition landscape for more capex-intense science/eng ventures
0
0
3
@eukaryote314
I mean they’re not mutually exclusive. If you make a model 10x larger while making it 10x more efficient that’s a win in both axis. The point is that now these models can be trained
3
0
3
@SmokeAwayyy
@ChatGPTapp
“- Add a counter or symbol as you approach a text or image usage cap.
- Add a timer showing time until caps are lifted.”
Lmao, ChatGPT becomes clash of clans
0
0
3
@gdb
Looking at the full log stream let’s us leverage the in context learner inside of us all 🙃
0
0
2
So cool to finally see this go live. Can't wait to see what we'll do next.
We’ve developed two neural networks which have learned by associating text and images. CLIP maps images into categories described in text, and DALL-E creates new images, like this, from text.
A step toward systems with deeper understanding of the world.
100
951
3K
0
0
2
@sama
Re: think very long term and execute relentlessly in the short term…… a pragmatic version of this from Jensen Huang is ~”think long term, but make sure you don’t die in the short term”
0
0
3
@royschwartz02
@nlpnoah
Congrats on the really cool paper. Did you guys explore connections to tensor-product transformers ? An input dependent manipulation of attention heads is used in both. This would be a really interesting ablation to inform future architecture designs.
1
0
3
@ariskonstant
IKEA instruction writers are prompt engineers and I’m an overparameterized furniture assembling LM🙃
0
0
3
Sharing some recent work with awesome collaborators :)
Happy to announce that my internship work
@NVIDIAAI
has resulted in this publication. Big thanks to
@wpingnet
@TheRealRPuri
@MohammadShoeybi
@MostofaPatwary
@ctnzr
for being awesome mentors 😁
#NLProc
1
0
7
0
0
2
Y’all if you’re into crypto and digital rights please sign this petition 🙏🏻🙏🏻
0
2
2
0
0
3
Come find out about our plans to incorporate text-to-text generative LMs with fewshot learning and other NLP tasks.
0
1
3
0
0
3
Great mental model for processing any piece of twitter hype. ❤️
It's out! The first
@pagestlabs
issue is on how to think about the buzz in breakthrough technologies like GPT-3 while living in the midst of it. Thanks everyone who subscribed early. Hope you like reading long posts 😅🖖
4
74
234
0
0
3
Big one
Arithmetic circuits were once the craft of human experts, and are now designed by AI in NVIDIA GPUs. H100 chips have nearly 13,000 AI designed circuits! How is this possible? Blog + a thread 🧵👇
25
415
2K
0
0
3
0
0
3
0
0
3