{
  "id": "352a7e075c3b",
  "url": "https://ainowinstitute.org/publication/policy/compute-and-ai",
  "title": "Computational Power and AI - AI Now Institute",
  "note": "This is a very broad and reasonably deep description of how compute power is used to build and train models, and why this is hard and very expensive.\n\n> Compute costs are predictably large: the final training run of GPT-3 is estimated to have cost somewhere between $500,000 to $4.6 million. Training GPT-4 may have cost in the vicinity of $50 million, but the overall training cost is probably more than $100 million because compute is required for trial and error before the final training run.\n\nThose costs are huge, and it is a ton of power too. It is interesting that the big innovations over the last few years — crypto and AI — are both very high compute demand workloads. Crypto has an answer with proof-of-stake. It isn't clear what will make AI costs go down for model building.",
  "voice": "mixed",
  "saved": "2023-10-06T00:05:40Z",
  "host": "ainowinstitute.org",
  "kind": "page",
  "tags": [],
  "record": "https://links.thingelstad.xyz/links/352a7e075c3b/",
  "markdown": "https://links.thingelstad.xyz/links/352a7e075c3b.md",
  "issues": [
    {
      "number": 264,
      "slug": "264",
      "title": "Weekly Thing 264 / Dazed, Zany, Llama",
      "date": "2023-10-07T12:00:00Z",
      "section": "Notable",
      "position": 1,
      "url": "https://weekly.thingelstad.com/archive/264/",
      "page": "https://links.thingelstad.xyz/issues/264/"
    }
  ],
  "connections": {
    "same_host": {
      "count": 0,
      "page": "https://links.thingelstad.xyz/hosts/ainowinstitute.org/",
      "nearest": []
    },
    "same_week": {
      "count": 22,
      "nearest": [
        {
          "id": "e13217efcdc8",
          "title": "We were never supposed to see our own faces this much | Dazed",
          "url": "https://www.dazeddigital.com/beauty/article/60860/1/we-were-never-supposed-to-see-our-faces-this-much-social-media-zoom",
          "record": "https://links.thingelstad.xyz/links/e13217efcdc8/"
        },
        {
          "id": "baefe77cf2ae",
          "title": "bruno",
          "url": "https://www.usebruno.com/",
          "record": "https://links.thingelstad.xyz/links/baefe77cf2ae/"
        },
        {
          "id": "28dc0eff33bc",
          "title": "Amazon launches its Bedrock generative AI service in general availability | TechCrunch",
          "url": "https://techcrunch.com/2023/09/28/amazon-launches-its-bedrock-generative-ai-service-in-general-availability/",
          "record": "https://links.thingelstad.xyz/links/28dc0eff33bc/"
        },
        {
          "id": "67193714aa0a",
          "title": "How To Turn Off Google’s “Privacy Sandbox” Ad Tracking—and Why You Should | Electronic Frontier Foundation",
          "url": "https://www.eff.org/deeplinks/2023/09/how-turn-googles-privacy-sandbox-ad-tracking-and-why-you-should",
          "record": "https://links.thingelstad.xyz/links/67193714aa0a/"
        },
        {
          "id": "c529f926e647",
          "title": "Introducing: Raspberry Pi 5! - Raspberry Pi",
          "url": "https://www.raspberrypi.com/news/introducing-raspberry-pi-5/",
          "record": "https://links.thingelstad.xyz/links/c529f926e647/"
        }
      ]
    },
    "mentions": [],
    "mentioned_by": [],
    "saved_twice": []
  },
  "elsewhere": {
    "wayback": "https://web.archive.org/web/*/https://ainowinstitute.org/publication/policy/compute-and-ai"
  }
}