{
  "schema_version": "1.0",
  "topic": "Artificial general intelligence",
  "compiled_date": "2026-10-22",
  "editorial_note": "All entries are paraphrases with source pointers, not verbatim quotations unless explicitly marked. Different definitions, claims, research, policy and measurements should not be conflated with a scored forecast.",
  "entries": [
    {
      "id": "agi-S01",
      "date": "2026-03-17",
      "speaker_or_authors": "Ryan Burnell and Oran Kelly, Google DeepMind",
      "venue": "Google DeepMind research announcement",
      "statement_type": "RESEARCH",
      "headline": "AGI needs a multi-ability test, not one benchmark",
      "summary_paraphrase": "DeepMind proposed a taxonomy of ten cognitive abilities and a human-baseline evaluation protocol for measuring progress toward AGI.",
      "qualification": "Measurement proposal, not an arrival prediction or proof of AGI.",
      "original_source": "https://blog.google/innovation-and-ai/models-and-research/google-deepmind/measuring-agi-cognitive-framework/"
    },
    {
      "id": "agi-S02",
      "date": "2026-02-17",
      "speaker_or_authors": "Gary Marcus, Walter Quattrociocchi and Valerio Capraro",
      "venue": "Gary Marcus essay discussing their Nature response",
      "statement_type": "CRITICAL VIEW",
      "headline": "Benchmarks are not equivalent to general intelligence",
      "summary_paraphrase": "The authors argued that high benchmark performance should not be confused with flexible, real-world general intelligence.",
      "qualification": "Explicit skeptical counterpoint; this is their interpretation, not an independently resolved AGI outcome.",
      "original_source": "https://garymarcus.substack.com/p/rumors-of-agis-arrival-have-been"
    },
    {
      "id": "agi-S03",
      "date": "2025-12-20",
      "speaker_or_authors": "Gary Marcus",
      "venue": "Marcus on AI: 2026 predictions",
      "statement_type": "FORECAST / SKEPTICAL",
      "headline": "No AGI in 2026 or 2027",
      "summary_paraphrase": "Marcus predicted that AGI would not arrive in 2026 or 2027 and argued that current LLM systems remained insufficiently reliable.",
      "qualification": "A dated negative forecast; no numerical probability or shared AGI test specified.",
      "original_source": "https://garymarcus.substack.com/p/six-or-seven-predictions-for-ai-2026"
    },
    {
      "id": "agi-S04",
      "date": "2025-04-09",
      "speaker_or_authors": "Yann LeCun",
      "venue": "AI Inside podcast interview",
      "statement_type": "CRITICAL VIEW",
      "headline": "Human-level AI requires more than language models",
      "summary_paraphrase": "LeCun discussed limitations of current LLMs, the need for world understanding, abstract reasoning and planning, and a longer research path toward human-level AI.",
      "qualification": "Recorded interview and methodological position; not evidence that AGI has arrived.",
      "original_source": "https://aiinside.show/video/yann-lecun-human-intelligence-is-not-general-intelligence-ai-inside-63"
    },
    {
      "id": "agi-S05",
      "date": "2025-04-02",
      "speaker_or_authors": "Anca Dragan, Rohin Shah, Four Flynn and Shane Legg",
      "venue": "Google DeepMind: Taking a responsible path to AGI",
      "statement_type": "OUTLOOK / SAFETY",
      "headline": "AGI could emerge within the coming years",
      "summary_paraphrase": "The authors described human-level capability on most cognitive tasks as potentially arriving in the coming years and called for proactive testing and safety planning.",
      "qualification": "Broad qualitative time horizon, not a date-certain prediction or a numeric probability.",
      "original_source": "https://deepmind.google/blog/taking-a-responsible-path-to-agi/"
    },
    {
      "id": "agi-S06",
      "date": "2025-03-19",
      "speaker_or_authors": "Thomas Kwa, Ben West, Joel Becker and METR colleagues",
      "venue": "METR: Measuring AI Ability to Complete Long Software Tasks",
      "statement_type": "EMPIRICAL MEASURE",
      "headline": "Agent task horizons doubled roughly every seven months",
      "summary_paraphrase": "METR reported an approximately seven-month historical doubling time for the length of tasks generalist agents could complete at a specified reliability level, while noting extrapolation limits.",
      "qualification": "Measures a particular autonomous-task capability, not AGI itself.",
      "original_source": "https://metr.org/blog/2025-03-19-measuring-ai-ability-to-complete-long-tasks/"
    },
    {
      "id": "agi-S07",
      "date": "2025-01-06",
      "speaker_or_authors": "Sam Altman",
      "venue": "Personal essay: Reflections",
      "statement_type": "EXECUTIVE CLAIM",
      "headline": "Altman said OpenAI knew how to build AGI as traditionally understood",
      "summary_paraphrase": "Altman expressed confidence in a route to traditionally understood AGI and expected the first AI agents to begin materially affecting workplaces in 2025.",
      "qualification": "A first-person capability outlook, not a publicly independently validated AGI milestone.",
      "original_source": "https://blog.samaltman.com/reflections"
    },
    {
      "id": "agi-S08",
      "date": "2024-10",
      "speaker_or_authors": "Dario Amodei",
      "venue": "Personal essay: Machines of Loving Grace",
      "statement_type": "SCENARIO",
      "headline": "What a world with powerful AI could look like",
      "summary_paraphrase": "Amodei explored the potential upsides of powerful AI and repeatedly cautioned that predictions about its consequences involve considerable uncertainty.",
      "qualification": "His defined powerful AI is not automatically equivalent to Cortist's proposed AGI threshold.",
      "original_source": "https://darioamodei.com/essay/machines-of-loving-grace"
    },
    {
      "id": "agi-S09",
      "date": "2024-07-21",
      "speaker_or_authors": "Meredith Ringel Morris, Shane Legg and Google DeepMind coauthors",
      "venue": "ICML 2024: Levels of AGI",
      "statement_type": "RESEARCH",
      "headline": "AGI should be measured along breadth and performance",
      "summary_paraphrase": "Researchers proposed levels that separate an AI system's generality from how proficiently it performs, rather than treating AGI as a single simple switch.",
      "qualification": "Academic measurement framework; no claim that a specific model had reached full AGI.",
      "original_source": "https://deepmind.google/research/publications/66938/"
    },
    {
      "id": "agi-S10",
      "date": "2023-02-24",
      "speaker_or_authors": "OpenAI",
      "venue": "OpenAI: Planning for AGI and beyond",
      "statement_type": "POLICY / OUTLOOK",
      "headline": "OpenAI argued AGI needs gradual deployment and independent oversight",
      "summary_paraphrase": "OpenAI described AGI as a possible transformative but risky capability and called for safety evaluations, public scrutiny and governance while acknowledging uncertain timelines.",
      "qualification": "Institutional policy statement, not a probability forecast or achieved milestone.",
      "original_source": "https://openai.com/index/planning-for-agi-and-beyond/"
    }
  ]
}