TellWorldAI World Archive
Ask中文
News
Models
Organisations
People
Papers & concepts
Products & agents
Chips & robots
History
Topics
Ask中文

Home›People

Person

Ilya Sutskever

Israeli-Canadian computer scientist born in 1986 who co-built AlexNet, co-founded OpenAI and served as its chief scientist, and later founded Safe Superintelligence Inc.

Born 1986

AffiliationsOpenAIGoogle BrainUniversity of TorontoStanford UniversityAll 5

2024 · 2 20122025
Events per year
11Events5Organisations15Works

Overview

Ilya Sutskever is an Israeli-Canadian computer scientist who works on neural networks and deep learning. He was born in 1986 in Nizhny Novgorod, then called Gorky, in the Soviet Union. His family moved to Israel when he was 5, and he grew up in Jerusalem; at 16 he moved to Canada.[1]

Education and doctoral work

He studied at the University of Toronto, where he received a BSc in mathematics in 2005, an MSc in computer science in 2007 and a PhD in computer science in 2013. His doctoral thesis was titled "Training Recurrent Neural Networks", and his doctoral advisor was Geoffrey Hinton.[1] A recurrent neural network processes a sequence one step at a time and carries a state forward from step to step, which makes it suited to ordered data; the thesis dealt with how such networks are trained.

AlexNet

In 2012, together with Alex Krizhevsky and Geoffrey Hinton, he built AlexNet, a neural network for recognising objects in images. ACM's Turing Award announcement states that this work nearly halved the error rate for object recognition on ImageNet, a large collection of labelled images used to measure progress in image recognition.[1][6] After a brief postdoc with Andrew Ng at Stanford in 2012, he joined Hinton's company DNNResearch; Google acquired that company in 2013, and he became a research scientist at Google Brain.[1][6]

At Google

At Google he was first author of "Sequence to Sequence Learning with Neural Networks" (2014), written with Oriol Vinyals and Quoc V. Le. The approach in that paper uses neural networks to map one sequence onto another, rather than assigning a single label to a single input. He also worked on TensorFlow, Google's software library for building and training neural networks, and he co-authored the 2013 word2vec paper on representations of words and phrases as vectors, as well as the 2016 AlphaGo paper.[1][7]

OpenAI

At the end of 2015 he left Google to co-found OpenAI, where he served as chief scientist.[1] There he co-authored the papers on GPT, GPT-2, GPT-3, CLIP and DALL-E; the GPT-3 paper describes him as "an early advocate for scaling large generative likelihood models".[1][5] These papers cover language models trained to generate text, the CLIP model that connects images with text descriptions, and the DALL-E image generation model.

The board vote of November 2023

In November 2023, while serving as a member of OpenAI's board, he voted to remove Sam Altman as CEO. Altman was reinstated about a week later, Sutskever left the board, and he later said that he regretted his role in the events.[1]

Departure and Safe Superintelligence

His departure from OpenAI was announced on 14 May 2024, after nearly ten years at the company; Jakub Pachocki succeeded him as chief scientist.[2] On 19 June 2024 he launched Safe Superintelligence Inc. (SSI) with Daniel Gross and Daniel Levy. The company has offices in Palo Alto and Tel Aviv and describes its work as having "one goal and one product: a safe superintelligence".[3][4] He became SSI's CEO in July 2025, after Gross left for Meta.[3][4]

Recognition

He was elected a Fellow of the Royal Society in 2022. He received NeurIPS Test of Time awards in 2022, in 2023 for the word2vec phrase paper, and in 2024 for the sequence-to-sequence paper.[1][8][9]

Place in the field

Sutskever's public career runs through several of the developments that made deep learning a working technology: image recognition with AlexNet in 2012, sequence-to-sequence learning and word vectors at Google, and the scaling of large generative models at OpenAI.[1][5][7] The AlexNet result is the one singled out in the Turing Award citation for nearly halving ImageNet error, and the sequence-to-sequence and word2vec papers were later recognised with NeurIPS Test of Time awards.[6][8][9] Since 2024 his stated focus has shifted from building larger general models to the single goal named in the title of his new company, a safe superintelligence.[3][4]

CapabilitiesLarge language models

TechniquesDeep learning

Governance & standardsSafety & governance

HistoryResearch history

Career timeline

20
  1. 1827
    • University of Toronto Worked at

      A public research university in Toronto where Geoffrey Hinton has been affiliated since 1987 and where AlexNet, the 2012 deep convolutional network, was developed.

  2. 2012
    • AlexNet Contributed to

      AlexNet is the eight-layer convolutional network that won the ILSVRC2012 image classification task by a wide margin, showing that a large network trained on a large labelled dataset with GPUs could beat hand-designed features.

    • ImageNet Classification with Deep Convolutional Neural Networks Contributed to

      The 2012 NeurIPS paper that introduced AlexNet, a large convolutional network trained on GPUs. It won the ILSVRC-2012 image classification competition by a wide margin and is widely seen as a turning point for deep learning in computer vision.

    • Stanford University Worked at 2012–2012

      Stanford University is a private research university in California, home to the Stanford AI Lab (SAIL) and the Institute for Human-Centered AI, which publishes the AI Index.

  3. 2013
    • Google Brain Worked at 2013–2015

      Google Brain was a deep-learning research team started inside Google's X in 2011. Its cat-recognition experiment, TensorFlow, sequence-to-sequence models and the Transformer made it central to modern AI; it merged into Google DeepMind in 2023.

  4. 2014
    • Dropout: A Simple Way to Prevent Neural Networks from Overfitting Contributed to

      A 2014 paper by Srivastava, Hinton, Krizhevsky, Sutskever and Salakhutdinov introducing dropout, a training trick that randomly removes units so large neural networks overfit less.

    • Sequence to Sequence Learning with Neural Networks Contributed to

      Ilya Sutskever, Oriol Vinyals and Quoc Le's 2014 paper showed that two LSTMs, one reading the input and one writing the output, could translate sentences and outperform a phrase-based statistical baseline. It set the encoder–decoder pattern behind later translation models.

  5. 2015
    • OpenAI Founded、Worked at 2015–2024

      OpenAI is a San Francisco AI research and deployment company founded in 2015 as a nonprofit. It makes the GPT model family and ChatGPT and filed confidentially for an IPO in June 2026.

  6. 2016
    • Mastering the game of Go with deep neural networks and tree search Contributed to

      The 2016 Nature paper describing AlphaGo, which combined deep neural networks with tree search. In October 2015 its distributed version beat the European Go champion Fan Hui 5-0, the first time a program defeated a professional player in full-sized Go without handicap.

  7. 2018
    • Improving Language Understanding by Generative Pre-Training (GPT) Contributed to

      OpenAI's 2018 paper introducing GPT, a Transformer decoder pre-trained to predict the next token and then fine-tuned for specific tasks. It improved the state of the art on 9 of 12 tasks studied and set the recipe the GPT series followed.

    • GPT-1 Contributed to

      GPT-1 was OpenAI's first generative pre-trained transformer, published in June 2018. It showed that one language model, pre-trained on unlabelled text and then fine-tuned, could improve results on many language tasks.

  8. 2019
    • Language Models are Unsupervised Multitask Learners (GPT-2) Contributed to

      OpenAI's 2019 paper on GPT-2, a 1.5-billion-parameter Transformer trained on web text. It showed that a model trained only to predict the next word could perform several tasks when prompted, and its staged release sparked public debate.

    • GPT-2 Contributed to

      GPT-2 is a large language model that OpenAI announced in February 2019. Its largest version has 1.5 billion parameters. OpenAI first withheld the full model, then released it in stages through November 2019.

  9. 2020
    • GPT-3 Contributed to

      GPT-3 is an autoregressive language model with 175 billion parameters, described in a May 2020 paper. It could perform new tasks from a few examples in the prompt, without any gradient updates, and it was first offered through OpenAI's API.

    • Language Models are Few-Shot Learners (GPT-3) Contributed to

      The GPT-3 paper showed that a language model with 175 billion parameters could perform many tasks from a few examples in its prompt, with no fine-tuning, and that this ability grew with model size.

  10. 2021
    • DALL·E Contributed to

      DALL·E is OpenAI's text-to-image model unveiled on 5 January 2021: a 12-billion-parameter transformer trained on 250 million text-image pairs that generates a picture token by token from a caption.

    • Learning Transferable Visual Models From Natural Language Supervision (CLIP) Contributed to

      CLIP is an OpenAI model trained on 400 million image-text pairs to map pictures and sentences into one shared space, so it can classify images zero-shot from a written description.

    • OpenAI Codex (2021 model) Contributed to

      OpenAI Codex is a 2021 GPT language model fine-tuned on public code from GitHub. It translates natural language into code, and was introduced alongside the HumanEval benchmark for testing code correctness.

  11. 2022
    • Whisper Contributed to

      Whisper is OpenAI's open-source speech recognition system, released in September 2022 and trained on 680,000 hours of audio, which transcribes speech in many languages and translates it into English.

  12. 2024
    • Safe Superintelligence Inc. (SSI) Founded

      Safe Superintelligence Inc. (SSI) is an American AI company founded in June 2024 by Ilya Sutskever, Daniel Gross and Daniel Levy to build a safe superintelligence as its only product, without releasing anything in between.

Worked at4

OpenAI2015–2024Google Brain2013–2015University of Toronto–2013Stanford University2012–2012

Founded2

OpenAI2015Safe Superintelligence Inc. (SSI)2024

20251

  1. FundingSafe Superintelligence is valued at $30 billion

    In March 2025 a Greenoaks-led round valued Safe Superintelligence at $30 billion, six times its September 2024 valuation, for a company with no revenue and about 20 employees.

20242

  1. MilestoneIlya Sutskever launches Safe Superintelligence Inc.

    On 19 June 2024 Ilya Sutskever announced Safe Superintelligence Inc. on X, a company with one stated goal: building safe superintelligence. It raised $1 billion in September 2024.

  2. DepartureOpenAI dissolves the Superalignment team after Sutskever and Leike exit

    In May 2024 OpenAI's chief scientist Ilya Sutskever left the company, followed days later by Superalignment co-leader Jan Leike; OpenAI then disbanded the team, and Leike publicly said safety had "taken a backseat to shiny products".

20232

  1. DepartureOpenAI's board fires Sam Altman, who returns days later

    In November 2023 OpenAI's board removed CEO Sam Altman, citing a lack of candour; after an employee revolt and a Microsoft offer, he was reinstated within days under a new board.

  2. MilestoneOpenAI announces its Superalignment team

    In July 2023 OpenAI created a team co-led by Ilya Sutskever and Jan Leike to solve superintelligence alignment within four years, committing 20% of its compute and proposing AI systems to help supervise AI.

20211

  1. ReleaseOpenAI introduces DALL·E

    On 5 January 2021 OpenAI announced DALL·E, a neural network that creates images from text captions, and CLIP, which learns visual concepts from natural-language supervision.

20201

  1. PaperOpenAI publishes the GPT-3 paper

    In May 2020 OpenAI published the GPT-3 paper, describing a 175-billion-parameter language model that performs new tasks from a few examples given in its text input, without additional training.

20191

  1. ReleaseOpenAI announces GPT-2 and holds back the full model

    On 14 February 2019 OpenAI announced GPT-2, a 1.5-billion-parameter Transformer that performed several language tasks without task-specific training. It did not release the full model at first, instead releasing it in stages.

20181

  1. PaperOpenAI publishes the first GPT paper on generative pre-training

    On 11 June 2018 OpenAI published a paper on generative pre-training: a language model first learns from unlabeled text and is then adapted to specific tasks. The approach improved the state of the art on 9 of the 12 tasks studied.

20151

  1. MilestoneOpenAI is founded as a nonprofit AI research lab

    In December 2015 a group of entrepreneurs and AI researchers founded OpenAI in San Francisco as a non-profit research company whose stated aim was to advance digital intelligence for humanity's benefit rather than financial return.

20121

  1. BenchmarkAlexNet wins the ImageNet ILSVRC 2012 challenge

    In October 2012 AlexNet, a deep neural network built by a University of Toronto team, won the ImageNet image-classification challenge by a wide margin. Its result showed that large neural networks trained on GPUs could outperform hand-designed vision methods.

Contributed to15

Notes and sources

  1. Ilya Sutskeverhttps://en.wikipedia.org/wiki/Ilya_Sutskever
  2. Ilya Sutskever, OpenAI co-founder and longtime chief scientist, departshttps://techcrunch.com/2024/05/14/ilya-sutskever-openai-co-founder-and-longtime-chief-scientist-departs/
  3. Ilya Sutskever, OpenAI's former chief scientist, launches new AI companyhttps://techcrunch.com/2024/06/19/ilya-sutskever-openais-former-chief-scientist-launches-new-ai-company/
  4. Safe Superintelligence Inc.https://ssi.inc/
  5. Language Models are Few-Shot Learnershttps://arxiv.org/abs/2005.14165
  6. Fathers of the Deep Learning Revolution Receive ACM A.M. Turing Awardhttps://www.eurekalert.org/news-releases/508417
  7. Sequence to Sequence Learning with Neural Networkshttps://arxiv.org/abs/1409.3215
  8. Announcing the NeurIPS 2024 Test of Time Paper Awardshttps://blog.neurips.cc/2024/11/27/announcing-the-neurips-2024-test-of-time-paper-awards/
  9. NeurIPS 2023 Test of Time Award talk: Distributed Representations of Words and Phrases and their Compositionalityhttps://neurips.cc/virtual/2023/test-of-time/83333

Cite this page: Ilya Sutskever. TellWorld, https://www.tellworld.net/people/ilya-sutskever. Updated 2026-11-08