Ilya Sutskever

Research

Ilya Sutskever co-authored the 2012 AlexNet paper with Alex Krizhevsky and Geoffrey Hinton, which trained a deep convolutional network on ImageNet using GPUs. In 2014 Sutskever, Oriol Vinyals, and Quoc V. Le published Sequence to Sequence Learning with Neural Networks, which the NeurIPS 2024 award announcement describes as the work that set the encoder-decoder architecture.⁹¹⁰¹¹

Sutskever also co-authored the papers introducing GPT-3, the CLIP image-text models, Codex with the HumanEval benchmark, and the Whisper speech recognition models. NVIDIA's 2026 release also credits Sutskever with contributions to AlphaGo and the GPT models and with leading the research behind reasoning models such as OpenAI o1, and the o1 System Card lists Sutskever among its foundational contributors.⁵⁶⁷⁸¹⁴

Safe Superintelligence Inc.

Sutskever co-founded Safe Superintelligence Inc. in 2024 as a lab with one goal and one product, a safe superintelligence, approaching safety and capabilities together as technical problems. In July 2025 SSI announced that Sutskever was formally CEO, with Daniel Levy as President and the technical team reporting to Sutskever.³²¹

In July 2026, SSI and NVIDIA announced a long-term strategic partnership that SSI says will let it scale its compute by ten times.²¹

Work and contributions

• Whisper Open weights

  Last-listed author of "Robust Speech Recognition via Large-Scale Weak Supervision" (2022), which trains speech recognition models on 680,000 hours of multilingual audio and releases the models and inference code; the repository is under the MIT license.⁸¹²

• CLIP Open weights

  Last-listed author of "Learning Transferable Visual Models From Natural Language Supervision" (2021), which trains image representations by predicting which caption goes with which image and releases code and pre-trained weights; the repository is under the MIT license.⁶¹³

• HumanEval Open data

  Co-author of "Evaluating Large Language Models Trained on Code" (2021), the Codex paper that introduced and released the HumanEval evaluation set; the repository's citation lists the full author list.⁷¹⁴

• OpenAI o1 reasoning research Proprietary product

  Listed among the foundational contributors to reasoning research in the authorship section of the OpenAI o1 System Card (2024).⁴

• Sequence to Sequence Learning with Neural Networks Published research

  First author, with Oriol Vinyals and Quoc V. Le, of the 2014 paper on mapping one sequence to another with neural networks; it received the NeurIPS 2024 Test of Time Paper Award.¹⁰¹¹

• AlexNet Published research

  Co-author, with Alex Krizhevsky and Geoffrey Hinton, of "ImageNet Classification with Deep Convolutional Neural Networks" (NIPS 2012), which received the NeurIPS 2022 Test of Time Award.⁹¹⁵

Timeline

1. 2012Co-authored the AlexNet paper on ImageNet classification, presented at NIPS 2012.⁹
2. Sep 10, 2014Released Sequence to Sequence Learning with Neural Networks with Oriol Vinyals and Quoc V. Le.¹⁰
3. May 28, 2020Co-authored the GPT-3 paper, Language Models are Few-Shot Learners.⁵
4. Feb 26, 2021Co-authored the CLIP paper on learning visual models from natural language supervision.⁶
5. Jul 7, 2021Co-authored the Codex paper that introduced the HumanEval benchmark.⁷
6. Dec 6, 2022Co-authored the Whisper paper on large-scale weakly supervised speech recognition.⁸
7. 2024Co-founded Safe Superintelligence Inc.¹
8. Dec 21, 2024Listed among the foundational contributors in the OpenAI o1 System Card.⁴
9. Jul 3, 2025Became formally CEO of Safe Superintelligence Inc., with Daniel Levy as President.²
10. Jul 27, 2026Announced a long-term strategic partnership between SSI and NVIDIA.¹

Other documented roles

Roles the sources document that are not confirmed as current.

• Author affiliated with OpenAI on the GPT-3 (2020), CLIP (2021), Codex (2021) and Whisper (2022) papers, OpenAI⁵⁶⁷⁸

Honors and recognition

• NeurIPS Test of Time Award (for the AlexNet paper) · Conference on Neural Information Processing Systems (NeurIPS), 2022¹⁵
• NeurIPS Test of Time Paper Award (for Sequence to Sequence Learning) · Conference on Neural Information Processing Systems (NeurIPS), 2024¹¹

Where they work

• Safe Superintelligence Inc. 2 people in the wiki

Related people

• Daniel GrossAlso at Safe Superintelligence Inc.
• Sathwik Tejaswi MadhusudhanLanguage models · Multimodal models
• Evan HubingerSafety and alignment · Language models
• Vaishaal ShankarMultimodal models · Language models
• Nanyun (Violet) PengLanguage models · Multimodal models
• Mira MuratiLanguage models · Multimodal models
• Amanda AskellSafety and alignment · Language models
• Yulia TsvetkovLanguage models · Safety and alignment
• Jared KaplanLanguage models · Safety and alignment
• Zvi MowshowitzLanguage models · Safety and alignment