{"id":"c2c1cc9dc72f698e6a4906bde8ab79a0e4c9cf8af350ec96099b4da6dfe18536","title":"MIT Technology Review reports OpenAI’s iGPT experiments with image generation","argument":"MIT Technology Review reported on July 16, 2020, that OpenAI was adapting GPT-2 to generate images. The resulting model, iGPT, was trained by turning images into sequences of pixels and asking the model to predict missing parts.\n\nThe article says iGPT could predict the second half of an image from its first half, producing completions that appeared sensible to a human. It describes the work as a new path for unsupervised learning, which trains on unlabeled data and can reduce the labor of labeling examples.\n\nThe article also raises a risk: it says iGPT could generate plausible images without training only on a narrow category such as faces, unlike the GANs it discusses. It notes that training remained computationally expensive, while warning that this barrier might not last. These results concern image completions, not demonstrated real-world reliability or the future video generation discussed in a quoted forecast by OpenAI policy director Jack Clark.","topicIds":["artificial-general-intelligence"],"domain":"other","claimType":"attributed reading","limits":"Text as retrieved on October 28, 2026; originally published July 16, 2020. The page is marked as edited after publication and includes an update; passages may reflect later changes to the page. The article’s prediction about future video generation is not evidence that it occurred. The article describes image completions but does not provide measures of their performance across unfamiliar images or the computation required to generate them.","speaker":{"name":"MIT Technology Review","roleAtSource":"Publisher"},"source":{"title":"OpenAI’s fiction-spewing AI is learning to generate images","publisher":"MIT Technology Review","format":"Article","url":"https://www.technologyreview.com/2020/07/16/1005284/openai-ai-gpt-2-generates-images/","publishedAt":"2020-07-16T13:59:09Z","dateBasis":"Date from the publisher’s own records, through its API."},"evidence":[{"text":"MIT Technology Review reports that OpenAI trained iGPT on images converted into pixel sequences, and that it could predict an image’s second half from its first.","excerpt":"they unfurled the images into a single sequence of pixels. They found that the new model, named iGPT, was still able to grasp the two-dimensional structures of the visual world.","url":"https://www.technologyreview.com/2020/07/16/1005284/openai-ai-gpt-2-generates-images/","relationship":"Source-stated"},{"text":"The article presents unsupervised learning as a way to train on unlabeled data and reduce manual labeling work.","excerpt":"it allows an AI system to learn about the world without a human filter, and significantly reduces the manual labor of labeling data.","url":"https://www.technologyreview.com/2020/07/16/1005284/openai-ai-gpt-2-generates-images/","relationship":"Source-stated"},{"text":"The article says iGPT could make plausible images without being trained only on a narrow category, while noting that training was computationally expensive.","excerpt":"While training the model is still computationally expensive, offering a natural barrier to its access, that may not be the case for long.","url":"https://www.technologyreview.com/2020/07/16/1005284/openai-ai-gpt-2-generates-images/","relationship":"Source-stated"},{"text":"Escape Meridian reads the article’s account as leaving open whether iGPT’s image completions generalize beyond the examples described.","excerpt":"Given the sequence of pixels for the first half of an image, it could predict the second half in ways that a human would deem sensible.","url":"https://www.technologyreview.com/2020/07/16/1005284/openai-ai-gpt-2-generates-images/","relationship":"Escape Meridian interpretation"}],"sourceMedia":{"status":"unchecked","sourceUrl":"https://www.technologyreview.com/2020/07/16/1005284/openai-ai-gpt-2-generates-images/","checkedAt":null,"items":[]},"assumptions":[],"whatWouldChangeView":{"attribution":"Open question","text":"How well would iGPT’s image completions hold up across unfamiliar images, and how much computation would generating them require?"},"editorialComparisons":[],"live":{"title":"MIT Technology Review reports OpenAI’s iGPT experiments with image generation","summary":"MIT Technology Review reported on July 16, 2020, that OpenAI was adapting GPT-2 to generate images. The resulting model, iGPT, was trained by turning images into sequences of pixels and asking the model to predict missing parts.\n\nThe article says iGPT could predict the second half of an image from its first half, producing completions that appeared sensible to a human. It describes the work as a new path for unsupervised learning, which trains on unlabeled data and can reduce the labor of labeling examples.\n\nThe article also raises a risk: it says iGPT could generate plausible images without training only on a narrow category such as faces, unlike the GANs it discusses. It notes that training remained computationally expensive, while warning that this barrier might not last. These results concern image completions, not demonstrated real-world reliability or the future video generation discussed in a quoted forecast by OpenAI policy director Jack Clark.","question":"How well would iGPT’s image completions hold up across unfamiliar images, and how much computation would generating them require?","claims":[{"text":"MIT Technology Review reports that OpenAI trained iGPT on images converted into pixel sequences, and that it could predict an image’s second half from its first.","excerpt":"they unfurled the images into a single sequence of pixels. They found that the new model, named iGPT, was still able to grasp the two-dimensional structures of the visual world.","support":"source-stated"},{"text":"The article presents unsupervised learning as a way to train on unlabeled data and reduce manual labeling work.","excerpt":"it allows an AI system to learn about the world without a human filter, and significantly reduces the manual labor of labeling data.","support":"source-stated"},{"text":"The article says iGPT could make plausible images without being trained only on a narrow category, while noting that training was computationally expensive.","excerpt":"While training the model is still computationally expensive, offering a natural barrier to its access, that may not be the case for long.","support":"source-stated"},{"text":"Escape Meridian reads the article’s account as leaving open whether iGPT’s image completions generalize beyond the examples described.","excerpt":"Given the sequence of pixels for the first half of an image, it could predict the second half in ways that a human would deem sensible.","support":"interpretation"}],"limitations":["Text as retrieved on October 28, 2026; originally published July 16, 2020.","The page is marked as edited after publication and includes an update; passages may reflect later changes to the page. The article’s prediction about future video generation is not evidence that it occurred.","The article describes image completions but does not provide measures of their performance across unfamiliar images or the computation required to generate them."],"mediaId":null,"correctionReason":null,"id":"c2c1cc9dc72f698e6a4906bde8ab79a0e4c9cf8af350ec96099b4da6dfe18536","publicationId":"4e42e1c2-65cc-fa8a-5d03-6c604840d775","version":1,"sourceUrl":"https://www.technologyreview.com/2020/07/16/1005284/openai-ai-gpt-2-generates-images/","sourceTitle":"OpenAI’s fiction-spewing AI is learning to generate images","publisher":"MIT Technology Review","publishedAt":"2020-07-16T13:59:09Z","originalPublished":{"value":"2020-07-16T13:59:09Z","precision":"instant","basis":"backfill:api_published"},"firstObservedAt":"2026-10-28T04:13:54.981Z","createdAt":"2026-10-29T00:48:25.311Z","historical":true,"entityIds":[],"topicIds":["artificial-general-intelligence"],"media":[],"sourceMedia":{"status":"unchecked","sourceUrl":"https://www.technologyreview.com/2020/07/16/1005284/openai-ai-gpt-2-generates-images/","checkedAt":null,"items":[]},"editorialDecision":{"outcome":"accept","mediaId":null,"inputHash":"1431774f80aa5d307261e45e21ee80ceae070ef6ace49294eecb2f70dec348cb","rubricVersion":"unrecorded","provider":"typesafe","model":"jev-1.13.0","reasonCodes":["bridge_admitted_by_model"],"confidence":0.53,"decidedAt":"2026-10-28T04:13:53.775Z"},"provenance":{"textHash":"f03222b85d32b9f4e43f39abfbfb1b0d593959caa587a73cd6a7be9833e0081b","observationId":"ba0f52f3c701b7b8ca9d457a2b45d21b33bb3d8644ceca8fdf0d45d5a3d12487","retrievedAt":"2026-10-28T04:15:49.745Z","kind":"publisher-page","preparer":{"model":"openai/gpt-6-luna","runId":"a3bcc57c-a59c-5a3e-3b17-0887115f6d07"},"reviewer":{"model":"anthropic/claude-opus-5.5","runId":"714bc52d-b0f7-2471-f61c-8dacfd7d3070"}},"audit":{"status":"passed","reviewedAt":"2026-10-29T00:48:03.003Z","model":"anthropic/claude-opus-5.5"}},"canonicalUrl":"https://escapemeridian.com/readings/mit-technology-review-reports-openai-s-igpt-experiments-with-c2c1cc9d","formats":{"html":"/readings/mit-technology-review-reports-openai-s-igpt-experiments-with-c2c1cc9d","markdown":"/readings/mit-technology-review-reports-openai-s-igpt-experiments-with-c2c1cc9d.md","json":"/readings/mit-technology-review-reports-openai-s-igpt-experiments-with-c2c1cc9d.json"},"attribution":"Headline and summary by Escape Meridian, prepared by Luna from the attributed source. A source check is not independent confirmation of the source’s claims.","readerContext":{"version":"reader-context-v1","attribution":"Reference context selected by Escape Meridian; separate from source evidence. Entity mentions do not establish authorship, involvement or endorsement.","terms":[],"entities":[{"id":"openai","name":"OpenAI","kind":"organization","context":"OpenAI develops AI models and products and publishes research, release announcements and evaluations.","referenceUrls":["https://openai.com/news/"],"path":"/entities/openai","json":"/entities/openai.json","markdown":"/entities/openai.md"},{"id":"jack-clark","name":"Jack Clark","kind":"person","context":"Jack Clark writes Import AI, a newsletter covering AI research, development and its implications.","referenceUrls":["https://jack-clark.net/"],"path":"/entities/jack-clark","json":"/entities/jack-clark.json","markdown":"/entities/jack-clark.md"}]},"withdrawn":false,"references":{"schemaVersion":1,"target":{"type":"reading","id":"c2c1cc9dc72f698e6a4906bde8ab79a0e4c9cf8af350ec96099b4da6dfe18536","revision":"4e42e1c2-65cc-fa8a-5d03-6c604840d775","revisionKind":"publication-id","title":"MIT Technology Review reports OpenAI’s iGPT experiments with image generation","url":"https://escapemeridian.com/readings/mit-technology-review-reports-openai-s-igpt-experiments-with-c2c1cc9d"},"source":{"type":"source","id":"https://www.technologyreview.com/2020/07/16/1005284/openai-ai-gpt-2-generates-images/","revision":"f03222b85d32b9f4e43f39abfbfb1b0d593959caa587a73cd6a7be9833e0081b","revisionKind":"sha256-source-text","title":"OpenAI’s fiction-spewing AI is learning to generate images","url":"https://www.technologyreview.com/2020/07/16/1005284/openai-ai-gpt-2-generates-images/"},"status":"available","items":[],"next":null,"scope":"Explicit citations in public readings and public file attachments to this reading. Shared topics are not citations. Other artifact types need an explicit reference adapter."},"revisions":[{"id":"c2c1cc9dc72f698e6a4906bde8ab79a0e4c9cf8af350ec96099b4da6dfe18536","title":"MIT Technology Review reports OpenAI’s iGPT experiments with image generation","argument":"MIT Technology Review reported on July 16, 2020, that OpenAI was adapting GPT-2 to generate images. The resulting model, iGPT, was trained by turning images into sequences of pixels and asking the model to predict missing parts.\n\nThe article says iGPT could predict the second half of an image from its first half, producing completions that appeared sensible to a human. It describes the work as a new path for unsupervised learning, which trains on unlabeled data and can reduce the labor of labeling examples.\n\nThe article also raises a risk: it says iGPT could generate plausible images without training only on a narrow category such as faces, unlike the GANs it discusses. It notes that training remained computationally expensive, while warning that this barrier might not last. These results concern image completions, not demonstrated real-world reliability or the future video generation discussed in a quoted forecast by OpenAI policy director Jack Clark.","topicIds":["artificial-general-intelligence"],"domain":"other","claimType":"attributed reading","limits":"Text as retrieved on October 28, 2026; originally published July 16, 2020. The page is marked as edited after publication and includes an update; passages may reflect later changes to the page. The article’s prediction about future video generation is not evidence that it occurred. The article describes image completions but does not provide measures of their performance across unfamiliar images or the computation required to generate them.","speaker":{"name":"MIT Technology Review","roleAtSource":"Publisher"},"source":{"title":"OpenAI’s fiction-spewing AI is learning to generate images","publisher":"MIT Technology Review","format":"Article","url":"https://www.technologyreview.com/2020/07/16/1005284/openai-ai-gpt-2-generates-images/","publishedAt":"2020-07-16T13:59:09Z","dateBasis":"Date from the publisher’s own records, through its API."},"evidence":[{"text":"MIT Technology Review reports that OpenAI trained iGPT on images converted into pixel sequences, and that it could predict an image’s second half from its first.","excerpt":"they unfurled the images into a single sequence of pixels. They found that the new model, named iGPT, was still able to grasp the two-dimensional structures of the visual world.","url":"https://www.technologyreview.com/2020/07/16/1005284/openai-ai-gpt-2-generates-images/","relationship":"Source-stated"},{"text":"The article presents unsupervised learning as a way to train on unlabeled data and reduce manual labeling work.","excerpt":"it allows an AI system to learn about the world without a human filter, and significantly reduces the manual labor of labeling data.","url":"https://www.technologyreview.com/2020/07/16/1005284/openai-ai-gpt-2-generates-images/","relationship":"Source-stated"},{"text":"The article says iGPT could make plausible images without being trained only on a narrow category, while noting that training was computationally expensive.","excerpt":"While training the model is still computationally expensive, offering a natural barrier to its access, that may not be the case for long.","url":"https://www.technologyreview.com/2020/07/16/1005284/openai-ai-gpt-2-generates-images/","relationship":"Source-stated"},{"text":"Escape Meridian reads the article’s account as leaving open whether iGPT’s image completions generalize beyond the examples described.","excerpt":"Given the sequence of pixels for the first half of an image, it could predict the second half in ways that a human would deem sensible.","url":"https://www.technologyreview.com/2020/07/16/1005284/openai-ai-gpt-2-generates-images/","relationship":"Escape Meridian interpretation"}],"sourceMedia":{"status":"unchecked","sourceUrl":"https://www.technologyreview.com/2020/07/16/1005284/openai-ai-gpt-2-generates-images/","checkedAt":null,"items":[]},"assumptions":[],"whatWouldChangeView":{"attribution":"Open question","text":"How well would iGPT’s image completions hold up across unfamiliar images, and how much computation would generating them require?"},"editorialComparisons":[],"live":{"title":"MIT Technology Review reports OpenAI’s iGPT experiments with image generation","summary":"MIT Technology Review reported on July 16, 2020, that OpenAI was adapting GPT-2 to generate images. The resulting model, iGPT, was trained by turning images into sequences of pixels and asking the model to predict missing parts.\n\nThe article says iGPT could predict the second half of an image from its first half, producing completions that appeared sensible to a human. It describes the work as a new path for unsupervised learning, which trains on unlabeled data and can reduce the labor of labeling examples.\n\nThe article also raises a risk: it says iGPT could generate plausible images without training only on a narrow category such as faces, unlike the GANs it discusses. It notes that training remained computationally expensive, while warning that this barrier might not last. These results concern image completions, not demonstrated real-world reliability or the future video generation discussed in a quoted forecast by OpenAI policy director Jack Clark.","question":"How well would iGPT’s image completions hold up across unfamiliar images, and how much computation would generating them require?","claims":[{"text":"MIT Technology Review reports that OpenAI trained iGPT on images converted into pixel sequences, and that it could predict an image’s second half from its first.","excerpt":"they unfurled the images into a single sequence of pixels. They found that the new model, named iGPT, was still able to grasp the two-dimensional structures of the visual world.","support":"source-stated"},{"text":"The article presents unsupervised learning as a way to train on unlabeled data and reduce manual labeling work.","excerpt":"it allows an AI system to learn about the world without a human filter, and significantly reduces the manual labor of labeling data.","support":"source-stated"},{"text":"The article says iGPT could make plausible images without being trained only on a narrow category, while noting that training was computationally expensive.","excerpt":"While training the model is still computationally expensive, offering a natural barrier to its access, that may not be the case for long.","support":"source-stated"},{"text":"Escape Meridian reads the article’s account as leaving open whether iGPT’s image completions generalize beyond the examples described.","excerpt":"Given the sequence of pixels for the first half of an image, it could predict the second half in ways that a human would deem sensible.","support":"interpretation"}],"limitations":["Text as retrieved on October 28, 2026; originally published July 16, 2020.","The page is marked as edited after publication and includes an update; passages may reflect later changes to the page. The article’s prediction about future video generation is not evidence that it occurred.","The article describes image completions but does not provide measures of their performance across unfamiliar images or the computation required to generate them."],"mediaId":null,"correctionReason":null,"id":"c2c1cc9dc72f698e6a4906bde8ab79a0e4c9cf8af350ec96099b4da6dfe18536","publicationId":"4e42e1c2-65cc-fa8a-5d03-6c604840d775","version":1,"sourceUrl":"https://www.technologyreview.com/2020/07/16/1005284/openai-ai-gpt-2-generates-images/","sourceTitle":"OpenAI’s fiction-spewing AI is learning to generate images","publisher":"MIT Technology Review","publishedAt":"2020-07-16T13:59:09Z","originalPublished":{"value":"2020-07-16T13:59:09Z","precision":"instant","basis":"backfill:api_published"},"firstObservedAt":"2026-10-28T04:13:54.981Z","createdAt":"2026-10-29T00:48:25.311Z","historical":true,"entityIds":[],"topicIds":["artificial-general-intelligence"],"media":[],"sourceMedia":{"status":"unchecked","sourceUrl":"https://www.technologyreview.com/2020/07/16/1005284/openai-ai-gpt-2-generates-images/","checkedAt":null,"items":[]},"editorialDecision":{"outcome":"accept","mediaId":null,"inputHash":"1431774f80aa5d307261e45e21ee80ceae070ef6ace49294eecb2f70dec348cb","rubricVersion":"unrecorded","provider":"typesafe","model":"jev-1.13.0","reasonCodes":["bridge_admitted_by_model"],"confidence":0.53,"decidedAt":"2026-10-28T04:13:53.775Z"},"provenance":{"textHash":"f03222b85d32b9f4e43f39abfbfb1b0d593959caa587a73cd6a7be9833e0081b","observationId":"ba0f52f3c701b7b8ca9d457a2b45d21b33bb3d8644ceca8fdf0d45d5a3d12487","retrievedAt":"2026-10-28T04:15:49.745Z","kind":"publisher-page","preparer":{"model":"openai/gpt-6-luna","runId":"a3bcc57c-a59c-5a3e-3b17-0887115f6d07"},"reviewer":{"model":"anthropic/claude-opus-5.5","runId":"714bc52d-b0f7-2471-f61c-8dacfd7d3070"}},"audit":{"status":"passed","reviewedAt":"2026-10-29T00:48:03.003Z","model":"anthropic/claude-opus-5.5"}}}],"audits":[{"publicationId":"4e42e1c2-65cc-fa8a-5d03-6c604840d775","verdict":"passed","findings":[{"dimension":"attribution","status":"pass","explanation":"The reading attributes the report to MIT Technology Review and the research to OpenAI. It labels Jack Clark's video-generation remark as a quoted forecast, which matches the article.","excerpt":"So researchers at OpenAI decided to swap the words for pixels and train the same algorithm on images in ImageNet"},{"dimension":"summary","status":"pass","explanation":"The reading clearly states the actor, the development (GPT-2's algorithm applied to images as pixel sequences) and the key term unsupervised learning in plain language. GANs are mentioned without explanation, but they are peripheral and the contrast still reads clearly. The question is a substantive open issue rather than a leading claim.","excerpt":"Given the sequence of pixels for the first half of an image, it could predict the second half in ways that a human would deem sensible."},{"dimension":"date","status":"pass","explanation":"The stated date, July 16, 2020, matches the UTC calendar day of originalPublished. The page was edited after publication (dateModified 2020-12-10); the update note says only a politician's name was removed. The reading's claims rest on core article passages that are plainly original. Clark's remark is presented as a forecast, not an outcome. The later sidebar stories are not used. Evidence kind is publisher-page, so no archive disclosure is needed.","excerpt":"Update: This article has been updated to remove the name of the politician in the hypothetical scenario described at the end."},{"dimension":"support","status":"pass","explanation":"Each claim is grounded in the article: pixel-sequence training and half-image completion, unsupervised learning reducing labeling work, and broad training versus curated GAN data along with the computational-cost barrier. The interpretive claim is labeled and conditional. The page edit does not affect these core passages.","excerpt":"iGPT, by contrast, simply learns enough of the structure of the visual world across millions and billions of examples to spit out images that could feasibly exist within it."},{"dimension":"media","status":"pass","explanation":"No media was selected and no verified media candidates were supplied, so nothing needs to be checked for identity or relevance.","excerpt":null}],"createdAt":"2026-10-29T00:48:03.003Z","model":"anthropic/claude-opus-5.5"}],"correctionUrl":"https://escapemeridian.com/corrections/"}