Oleksii Kuchaiev @kuchaev Twitter profile

Last Seen Profiles

@northfieldaca

@donmezmustafa

@OfficialBuck103

@hanjsjsjsjj

@l1tttxk1ve

@JIUDAFU

@odachandango

@Domino6544

@Aatos5

@Marinove_1989

@lknurkse10

@cnietor

@Humairah_Ak

@marlongerson

@69asuna18

@AKMURAD1971

@AissaLille

@youngsince77

@mdmehmetduman

@brendantp0

@kalopop

@realdreams_ai

@TowardsGalaxys

@_80085__

@Sebihaarik

@nttuvannalu

@profmelnik

@linsdau_lily

@OTolmachev

@LovelsEV0L

@zyyyski

@Yojanelchakalit

@adnanmumtaz382

@otayfff11

@crisicolmondo

@i2jongwoo

Oleksii Kuchaiev

@kuchaev

2 months

Today we are happy to release best open models for synthetic data generation. 340B parameters, includes base, instruct and reward models. As well as new human preference dataset HelpSteer2. 340B-Reward model is #1 on the Reward Bench leaderboard.

NVIDIA Releases Open Synthetic Data Generation Pipeline for Training Large Language Models

Nemotron-4 340B, a family of models optimized for NVIDIA NeMo and NVIDIA TensorRT-LLM, includes cutting-edge instruct and reward models, and a dataset for generative AI training.

blogs.nvidia.com

18

84

460

Oleksii Kuchaiev

@kuchaev

5 months

In LLM pre training, curating and preparing data is perhaps the most impactful step. NeMo data curator is now open source with lots of features you will need. We used it to curate trillions of tokens for our own models training.

GitHub - NVIDIA/NeMo-Curator: Scalable data pre processing and curation toolkit for LLMs

Scalable data pre processing and curation toolkit for LLMs - NVIDIA/NeMo-Curator

github.com

3

43

233

Oleksii Kuchaiev

@kuchaev

2 months

For those wondering, "june-chatbot" on @lmsysorg is exactly this model posted on @huggingface

nvidia/Nemotron-4-340B-Instruct · Hugging Face

huggingface.co

4

27

142

Oleksii Kuchaiev

@kuchaev

2 years

NeMo speech recognition models published on @huggingface hub are now at the top of HF Speech Bench for all languages where we published the models so far: English, German, French and Chinese. Often beating other models without external LMs and with fewer parameters. #DeepLearning

2

25

127

Oleksii Kuchaiev

@kuchaev

9 months

Our team just released a new dataset for LLM model alignment, called HelpSteer on @huggingface Hub under CC-BY-4 license! This one should be used for reward model training, especially with SteerLM method.

nvidia/HelpSteer · Datasets at Hugging Face

huggingface.co

3

28

96

Oleksii Kuchaiev

@kuchaev

11 months

Our team is happy to share SteerLM, a simpler alternative to RLHF which allows dynamic model controls during inference (humor, verbosity, etc.). To appear in Findings of EMNLP 2023. It is implemented in NeMo (open-source) and example model is on HF

2

21

92

Oleksii Kuchaiev

@kuchaev

4 years

Very happy to share our latest NeMo release. We re-designed NeMo to work with @PyTorchLightnin and @Hydra_Framework projects from @PyTorch ecosystem. Train your own #ASR , #NLP and #TTS models or re-use one of the many pre-trained models we have.

PyTorch

@PyTorch

4 years

NeMo, @NVIDIA ’s open-source toolkit based on #PyTorch , allows you to quickly build, train, and fine-tune conversational AI models. See how speech recognition, natural language processing and speech synthesis can be improved in this tutorial:

0

177

598

1

20

56

Oleksii Kuchaiev

@kuchaev

2 months

You can try Nemotron-4-340B-Instruct model here

NVIDIA NIM | nemotron-4-340b-instruct

Experience the leading models to build enterprise generative AI apps now.

build.nvidia.com

1

9

55

Oleksii Kuchaiev

@kuchaev

2 months

@agihippo I guess you missed the part where we used 1000x less human data (10K vs 10M) for alignment than llama3. this is about synthetic data generation, literally says so in the blogpost. Also we released reward model and training data for it, all under commercial friendly license.

1

52

Oleksii Kuchaiev

@kuchaev

2 months

@razomforukraine @POTUS @WhiteHouse @POTUS and @WhiteHouse you must act on this

0

8

51

Oleksii Kuchaiev

@kuchaev

4 months

@srush_nlp Maybe it is because de-noising objective is "wasting" tokens compared to autoregressive models. E.g. when you mask 15% of tokens, then after 1 epoch you've backpropogated loss from only 15% of your tokens, compared to 100% in next token prediction loss.

2

4

49

Oleksii Kuchaiev

@kuchaev

9 months

NeMo Aligner, a toolkit for scalable AI model alignment is now Open Source! Currently it support PPO-based RLHF, SteerLM and DPO. #EMNLP2023 #nlproc

GitHub - NVIDIA/NeMo-Aligner: Scalable toolkit for efficient model alignment

Scalable toolkit for efficient model alignment. Contribute to NVIDIA/NeMo-Aligner development by creating an account on GitHub.

github.com

1

6

38

Oleksii Kuchaiev

@kuchaev

2 months

Nemotron-4-340B-*Reward* model is now available via API on :) Give it a try.

0

12

37

Oleksii Kuchaiev

@kuchaev

2 months

Models on @huggingface under Nvidia Open Model License

Nemotron 4 340B - a nvidia Collection

huggingface.co

0

4

36

Oleksii Kuchaiev

@kuchaev

9 months

Check out llama2-70B-SteerLM model which gets 7.54 on MT-bench. This model is NOT using outputs of stronger (ChatGPT) models during alignment which allowed us to keep llama2 license. Try now on NGC Catalog . Also on @huggingface hub

1

5

26

Oleksii Kuchaiev

@kuchaev

2 years

NeMo now has Ukrainian speech recognition model on @huggingface hub. This is a CitriNet model tuned by our intern working from Kyiv As of today, I think, this is the best Ukrainian ASR model freely available.

4

5

25

Oleksii Kuchaiev

@kuchaev

2 years

Oh wow! Look what model is #1 on @huggingface speech bench for English speech recognition

Somshubra Majumdar

@HaseoX94

2 years

The largest NeMo ASR model is finally public on @huggingface ! This is a 600 M params Conformer Transducer X-Large, probably the largest public checkpoint trained with multiple datasets.

2

19

88

0

3

22

Oleksii Kuchaiev

@kuchaev

23 days

@paulg I grew up near Chernobyl and have always belived that its most devastating impact is a subsequent push back against nuclear power. Interesting fact - the Chernobyl station kept working after disaster (1986) until 1999 when it was fully shut down under the Western pressure.

1

2

22

Oleksii Kuchaiev

@kuchaev

5 years

A paper about our latest speech recognition model - QuartzNet has been accepted to #ICASSP 2020. Head over to for implementation and pretrained models. #DeepLearning #asr

0

5

21

Oleksii Kuchaiev

@kuchaev

2 months

@fleetingbits @lmsysorg @NVIDIAAI Another way to put it: we used 1000x less (10K vs 10M) human data for alignment than llama3 by using synthetic data. This release is about synthetic data generation which is why our license explicitly allows it.

0

21

Oleksii Kuchaiev

@kuchaev

2 months

@lmsysorg @NVIDIAAI With a license that permits synthetic data generation and commercial use.

1

20

Oleksii Kuchaiev

@kuchaev

2 months

@fchollet people claiming that some LLM is at "high schooler" or even "kindergartner" level should spend more time around kids. No AI system today is near the level of even 3 year old when it comes to general intelligence. Moreover, the progress towards that level is unclear.

0

19

Oleksii Kuchaiev

@kuchaev

3 years

Just finished listening to "Viral" audiobook by @Ayjchan and @mattwridley . This is an excellent account of the most likely origins of COVID-19 pandemic. A must read for anyone (which really should be everyone) interested in the the origins of #COVID19

0

3

16

Oleksii Kuchaiev

@kuchaev

2 years

Two very real steps anyone in the world can help: 1) Consider donating to humanitarian relief efforts, such as @razomforukraine and there are many others. If your company has a match - please make sure you make use of that. #UkraineRussiaWar

1

7

16

Oleksii Kuchaiev

@kuchaev

1 year

@mattrickard you missed NeMo models (1, 5 and 20B GPTs with commercialy friendly license)

Models - Hugging Face

huggingface.co

1

2

16

Oleksii Kuchaiev

@kuchaev

2 months

@geoff_l generating synthetic data for alignment of smaller models is key use case we have in mind.

1

0

15

Oleksii Kuchaiev

@kuchaev

2 years

#ChatGPT is very impressive! In the dialogue below, it comes up with a suboptimal solution, and argues a little without admitting a mistake. Then takes a hint, admits the mistake, and fixes its solution!

1

4

14

Oleksii Kuchaiev

@kuchaev

9 months

@natolambert @srush_nlp Yes, PPO is much more expensive then DPO in terms of infra but in all our experiments so far, on the same data (and no online setting) PPO>DPO on MT-bench.

3

0

13

Oleksii Kuchaiev

@kuchaev

2 years

This holiday seasons please consider supporting Ukraine which is being attacked by the terrorist federation. Today 76 rockets were fired at critical civilian infrastructure with typical terrorist intents of generating fear and humanitarian disaster.

3

11

Oleksii Kuchaiev

@kuchaev

3 years

NeMo 1.0.0 is ready! #asr #NLProc #DeepLearning #AI

Accelerating Conversational AI Research with New Cutting-Edge Neural Networks and Features from...

The 1.0 update brings significant architectural, code quality, and documentation improvements as well as a plethora of new state-of-the-art neural networks and pretrained checkpoints in several…

developer.nvidia.com

0

4

11

Oleksii Kuchaiev

@kuchaev

4 years

Thank you, Mozilla Common Voice! (New ASR models in NeMo are coming ;) ) —— 2020 End-of-Year Common Voice Dataset Release - Common Voice - Mozilla Discourse

2020 End-of-Year Common Voice Dataset Release

Happy end of 2020! While it has been a tumultuous year for all, the Common Voice team is excited to announce the end of year data set release! Firstly, we could not have made it through this year...

discourse.mozilla.org

1

2

9

Oleksii Kuchaiev

@kuchaev

1 year

@_willfalcon You missed some of the open-source, commercially friendly (CC-BY-4) models built using lightning :)

Models - Hugging Face

huggingface.co

1

0

8

Oleksii Kuchaiev

@kuchaev

9 months

Our team has studied the tradeoffs between performance and the number of trainable params in LoRA. This work would be especially useful to those building and scaling AI customization services. Great work by @rendu_a and Tugrul Konuk

Tied-Lora: Enhancing parameter efficiency of LoRA with weight tying

We introduce Tied-LoRA, a novel paradigm leveraging weight tying and selective training to enhance the parameter efficiency of Low-rank Adaptation (LoRA). Our exploration encompasses different...

arxiv.org

0

2

8

Oleksii Kuchaiev

@kuchaev

3 years

We have an exciting new job opportunity for NLP researcher on our team. Please check job description and apply here if interested #NLProc #NLP #OpenSource

0

5

8

Oleksii Kuchaiev

@kuchaev

7 months

Our team is hiring Sr. Applied Scientists to work on AI model alignment and customization (text and multimodal). If you have strong track record and experience with: LLMs or RL or multi-modal, please apply. Can be in-person or remote. #NLP #hiring

Work With us and Transform Industries

Learn about our culture and much more. #NVIDIA Careers.

www.nvidia.com

1

4

8

Oleksii Kuchaiev

@kuchaev

4 months

Latest release of NeMo-Aligner adds TRT-LLM integration which speeds up RLHF rollouts up to 7x compared to pure Pytorch implementation

0

7

Oleksii Kuchaiev

@kuchaev

8 months

I was taught that trees aren't supposed to have any cycles!

0

7

Oleksii Kuchaiev

@kuchaev

19 days

@karpathy It seems like the recipe to achieving superhuman performance in domain X is now well known: have perfect reward model in X (e.g. game rules, physics) + Good transformer-based heuristic + graph reasoning (MCTS/A*, etc.). Perfect reward model for LLMs would indeed be a game changer

0

7

Oleksii Kuchaiev

@kuchaev

2 years

DALLE-mini by @borisdayma and team is pretty good!

0

6

Oleksii Kuchaiev

@kuchaev

3 years

@sergeykarayev We do (much more to come in the next months)

0

1

7

Oleksii Kuchaiev

@kuchaev

9 months

Santa Barbara weather never disappoints

0

7

Oleksii Kuchaiev

@kuchaev

9 months

@alexgraveley yes, currently very limited on human preference data. You might want to add new dataset we've published this week which can be used with DPO. Btw, in most of our experiments SFT < DPO < SteerLM <= PPO. So while simple, DPO lags behind PPO and SteerLM.

2

0

7

Oleksii Kuchaiev

@kuchaev

13 days

@natolambert or we can just drop “R” from RLHF for “Learning from Human Feedback”

1

0

6

Oleksii Kuchaiev

@kuchaev

9 months

@srush_nlp @natolambert Some experiments are summarized in Table 3 from this paper (though it isn't the main point of the paper).

0

6

Oleksii Kuchaiev

@kuchaev

7 months

Congrats Som and entire NeMo Speech team! Top 4 models on @huggingface ASR leaderboard are now NeMo models, pushing Whisper-large to #5 .

Somshubra Majumdar

@HaseoX94

7 months

To read up about Token and Duration Transducers, visit the TDT is research from our team that enabled ASR inference at speeds approaching Non Autoregressive inference, but with an Autoregressive model !

0

2

8

1

0

6

Oleksii Kuchaiev

@kuchaev

2 years

@sama detailed paper about it

0

6

Oleksii Kuchaiev

@kuchaev

5 years

On Friday (12/13), I will be presenting NeMo at MLSys Workshop at #NeurIPS2019 . NeMo is a toolkit for building conversational AI (ASR, NLP and TTS) applications. If you are at #NeurIPS and are interested in this work ping me to have a chat. #ai

1

0

6

Oleksii Kuchaiev

@kuchaev

9 months

I'm heading to Singapore for #EMNLP2023 DM if you want to chat about model alignment and customization

0

6

Oleksii Kuchaiev

@kuchaev

4 years

Watch @poonamchitale and @_willfalcon doing NeMo walk-through and discussing our collaboration

Lightning AI ⚡️

@LightningAI

4 years

You can build, train and fine-tune speech recognition, TTS, and NLP SOTA models at scale with just a few lines of code with @nvidia NeMo and Lightning.

0

25

83

0

4

6

Oleksii Kuchaiev

@kuchaev

4 years

@lexfridman @navalny you should moderate an interview with @elonmusk and @navalny instead of Putin. Because Putin will not age well in history books.

0

6

Oleksii Kuchaiev

@kuchaev

2 years

NeMo Megatron GPT models: 1.3B, 5B, and 20B are now available on @huggingface hub! #GTC22 #GPT

Models - Hugging Face

huggingface.co

1

5

6

Oleksii Kuchaiev

@kuchaev

1 month

@bindureddy At least Nemotron-4 gets it

0

1

6

Oleksii Kuchaiev

@kuchaev

4 months

@paulg @VABVOX A quote I've heard (attributed to L. Landau, not sure if he truly said it though), could be roughly translated as: "What makes University great is not quality of it's professors or courses but the topics which students discuss while drinking vodka in their dorms"

0

5

Oleksii Kuchaiev

@kuchaev

3 years

New NeMo update! Adds new models for ASR (CitriNet, Conformer-CTC), NLP (Machine translation models) and TTS (HiFiGan, MelGan, GlowTTS, UniGlow SqueezeWave). 60 new pre-trained models: ASR in En, Zh, Ru, Es, Pl, Ca, It, Fr and De. NMT in 5 language pairs.

Release NVIDIA Neural Modules 1.0.0rc1 · NVIDIA/NeMo

Release 1.0.0rc1 This release contains major new models, features and docs improvements. It is a "candidate" release for 1.0.0. To install from Pip do: apt-get update && apt-get i...

github.com

1

2

5

Oleksii Kuchaiev

@kuchaev

2 years

Always double check #ChatGPT before using it for scheduling your meetings. (correct answer is 8am according to Google)

0

5

Oleksii Kuchaiev

@kuchaev

2 years

Controlled Adaptation of Speech Recognition Models to New Domains | NVIDIA Technical Blog by @HaseoX94

Controlled Adaptation of Speech Recognition Models to New Domains | NVIDIA Technical Blog

Using adapters for parameter-efficient training reduces the effects of catastrophic forgetting of general speech recognition.

developer.nvidia.com

0

2

5

Oleksii Kuchaiev

@kuchaev

1 year

@Yampeleg no, if you increase you batch size by x, you should first try increasing lr by sqrt(x)

0

5

Oleksii Kuchaiev

@kuchaev

2 years

Contact your political representatives asking them to provide more support to Ukraine. Ukraine needs air defense, longe range rockets, fighter jets and tanks. To contact your US Senators check: #UkraineRussiaWar

0

3

5

Oleksii Kuchaiev

@kuchaev

1 month

@ylecun awesome work by Meta team! Can you please also release reward models like we did with Nemotron?

1

0

3

Oleksii Kuchaiev

@kuchaev

2 years

@mrm8488 some suggestion on 20B to start with :)

nvidia/nemo-megatron-gpt-20B · Hugging Face

huggingface.co

0

4

Oleksii Kuchaiev

@kuchaev

4 years

0

4

Oleksii Kuchaiev

@kuchaev

6 years

0

1

4

Oleksii Kuchaiev

@kuchaev

21 days

@natolambert @Gradio "Too many post-training researchers don't talk to their models" - what do you think we do all day? 🤣

1

0

4

Oleksii Kuchaiev

@kuchaev

7 months

@natolambert Reward hacking is an issue, and this is assuming your reward model actually learned what you hope it did...

0

3

Oleksii Kuchaiev

@kuchaev

6 years

I’ll be presenting OpenSeq2Seq during #NLPOSS workshop at #ACL2018 . It is a toolkit for distributed and mixed precision training of seq2seq models for machine translation, speech recognition and speech synthesis. Check it out here: #nlproc #DemocratizeNLP

0

4

Oleksii Kuchaiev

@kuchaev

4 years

A blogpost by me on ASR transfer learning from one language to another #DeepLearning #asr #GTC2020 #NVIDIA “

Jump-start Training for Speech Recognition Models in Different Languages with NVIDIA NeMo | NVIDIA...

Transfer learning is an important machine learning technique that uses a model’s knowledge of one task to make it perform better on another. Fine-tuning is one of the techniques to perform transfer…

developer.nvidia.com

0

4

Oleksii Kuchaiev

@kuchaev

2 months

@Scott_Wiener SB 1047 is a bad piece of legislation because it regulates model development instead of their deployments in applications.

1

0

4

Oleksii Kuchaiev

@kuchaev

2 years

#StableDiffusion is a huge hit among kids :)

0

4

Oleksii Kuchaiev

@kuchaev

5 years

@ID_AA_Carmack If you are putting together some ASR pipeline - checkout QuartzNet it is a simple lightweight convolutional model. It does use mel spectrogram but you can switch to mfcc and tweak this

0

4

Oleksii Kuchaiev

@kuchaev

4 years

For the first time since 2011 a crewed launch from USA. And the first crewed launch for @SpaceX !!! #LaunchAmerica

0

3

Oleksii Kuchaiev

@kuchaev

2 years

A fantastic opportunity to work in one of the top speech processing groups!

Mirco Ravanelli

@mirco_ravanelli

2 years

I would like to remark that applications from #Ukrainian #students are particularly welcome!

24

2

20

0

3

Oleksii Kuchaiev

@kuchaev

5 years

Thank you! We love this dataset and already pre-trained and released some models with it!

Josh Meyer

@_josh_meyer_

5 years

It's finally out! (submitted to LREC) Common Voice: A Massively-Multilingual Speech Corpus

11

135

467

1

3

Oleksii Kuchaiev

@kuchaev

2 years

@karpathy shouldn’t there be 3) scalability gap too?

0

3

Oleksii Kuchaiev

@kuchaev

3 years

Are you excited about #OpenSource #DeepLearning tools for Conversational #AI ? We are #Hiring engineers to work on projects like NVIDIA NeMo. If this sounds exciting to you, apply here: #NLP #NLProc

0

2

3

Oleksii Kuchaiev

@kuchaev

2 months

@SelfishPriority helping other models become better is the goal here

0

3

Oleksii Kuchaiev

@kuchaev

6 years

Our #ICLR2018 Poster and Paper on mixed precision training with Volta GPUs

0

3

Oleksii Kuchaiev

@kuchaev

2 years

@danielgross Chinchilla paper showed that more compute during training is still better, just changes what kind of tradeoffs one would make between model size and training data size.

0

3

Oleksii Kuchaiev

@kuchaev

5 years

GTC is always fun #GTC19

0

3

Oleksii Kuchaiev

@kuchaev

2 months

@alexgraveley Nemotron-4 reward model is available both as checkooint and API

1

0

3

Oleksii Kuchaiev

@kuchaev

3 years

Very excited to do more with Common Voice! ASR models trained with it already start to appear in NeMo (Fr, It, Ru, De, Es, Pl, Ca) #asr #GTC21

Jane Scowcroft

@janescowcroft

3 years

I'm so pleased to share that Common Voice is heading into it's next stage of growth as high-impact partners, like NVIDIA, recognize the value of large scale open voice datasets. It also marks a career transition for me, as I join t…

0

5

18

0

1

3

Oleksii Kuchaiev

@kuchaev

8 months

@natolambert Problem is that releasing test sets turns them into dev sets rather fast ...

1

0

2

Oleksii Kuchaiev

@kuchaev

5 years

My hometown - even with the picture of a house I grew up in: Inside Slavutych, the city created by the Chernobyl explosion via @CNNTravel

0

3

Oleksii Kuchaiev

@kuchaev

2 years

NeMo Megatron enables training of gigantic language models (<= 1T params) with all types of parallelisms (data, tensor, pipeline, sequence), tuning them for your usecases and deploying them #gpt #T5 #NLP

NVIDIA AI Platform Delivers Big Gains for Large Language Models | NVIDIA Technical Blog

NVIDIA AI platform makes LLMs accessible. Announcing new parallelism techniques and a hyperparameter tool to speed-up training by 30% on any number of GPUs.

developer.nvidia.com

0

3

2

Oleksii Kuchaiev

@kuchaev

2 years

With NeMo Megatron you can train 100B+ LLMs and then “tune” them further for your needs #NLProc #gpt3

Adapting P-Tuning to Solve Non-English Downstream Tasks | NVIDIA Technical Blog

The scarcity of non-English labeled datasets poses serious challenges in applying large NLP models to specific tasks. Learn how we use p-tuning to solve them.

developer.nvidia.com

0

1

2

Oleksii Kuchaiev

@kuchaev

6 months

@natolambert This is awesome! But please, please, don't just say "Mistral-7b" or "Llama-2-7/13/70" without corresponding "-instruct/chat" suffix. So many people are confused about base vs aligned models. All models on the board are aligned.

1

0

2

Oleksii Kuchaiev

@kuchaev

5 years

Hmmm ...

0

2

Oleksii Kuchaiev

@kuchaev

5 years

Yes! Code, docs and pre-trained checkpoint

Josh Meyer

@_josh_meyer_

5 years

@kuchaev Are the models open:)?

0

1

2

Oleksii Kuchaiev

@kuchaev

3 months

@alexandr_wang @scale_AI have you seen CA's SB 1047 though ...

0

2

Oleksii Kuchaiev

@kuchaev

4 months

@srush_nlp This is not from paper (at least I'm not aware of any published study). But just from loss definitions, if you mask x% of tokens in de-noising loss that means you'll backpropogate loss signal from x% of your tokens after 1 epoch, whereas it is 100% in case of autoregressive loss.

3

0

2

Oleksii Kuchaiev

@kuchaev

4 years

NeMo is being developed in the open on GitHub . Try it out and give us your feedback!

1

2

Oleksii Kuchaiev

@kuchaev

9 months

With this data and SteerLM method we can align Llama2-70B and get 7.54 on MT-Bench. Paper with details and ablations studies:

HelpSteer: Multi-attribute Helpfulness Dataset for SteerLM

Existing open-source helpfulness preference datasets do not specify what makes some responses more helpful and others less so. Models trained on these datasets can incidentally learn to model...

arxiv.org

0

1

2

Oleksii Kuchaiev

@kuchaev

4 years

NVIDIA wins all 8 benchmarks for publicly available systems. And all 8 on per-chip basis.

0

2

Oleksii Kuchaiev

@kuchaev

1 month

@ylecun Thank you! this is a huge contribution towards positive AI progress

0

2

Oleksii Kuchaiev

@kuchaev

2 months

@rpanda89 yes, you can use everything now in NeMo Aligner. Methodology is published in paper. We are working on making everything easier to use.

0

2

Oleksii Kuchaiev

@kuchaev

1 year

This is an excellent list worth looking at

George Mack

@george__mack

1 year

What is ignored or neglected by the media -- but will be studied by historians? Here's the full list of 25 examples:

1K

22K

105K

0

1

Oleksii Kuchaiev

@kuchaev

3 years

NVIDIA Academic Hardware Grant Program

0

2

Oleksii Kuchaiev

@kuchaev

9 months

@RepAnnaEshoo Thank you for all your work!

0

Oleksii Kuchaiev

@kuchaev

3 years

This is so horrible and sad.

Allen J. Schaben

@alschaben

3 years

Breaking: First view of major oil Spill off of Huntington Beach and Newport Beach. Booms deployed. Beaches closed @Pacific_Airshow cancelled #oilspill #orangecounty @latimes @latimesphotos

576

3K

4K

0

2

Oleksii Kuchaiev

@kuchaev

2 months

@thegartsy SB 1047 helps OpenAI and hurts open source, not the other way around. The bill is making a fundamental mistake by regulating tech development instead of its application. It is not fair to expect randomly sampled Californians to understand the distinction.

1

0

2

Oleksii Kuchaiev

@kuchaev

4 years

#LaunchAmerica !!! @NASA @SpaceX

0

2