--- schema: 1 type: video title: 'Data Quality Is the Compute Multiplier — Ari Morcos, DatologyAI' description: >- Swap compute for data on the scaling curve and the same money buys a better model, which is why Ari Morcos calls data quality the compute multiplier and the most underinvested part of training. His frame is an oil refinery for data rather than a firehose: clean, curate, create, and compose, with quality classifiers, deduplication, and synthetic generation each earning their place, and the sequencing across stages mattering as much as any single step. The scarce resource now is not tokens but signal per token, and finding data that is optimal for a given target is where the leverage hides. The proof points are concrete. Better curated data lets a small multilingual model beat far larger ones trained on many more tokens, and it buys real inference efficiency because a model reaches the same quality with less. Morcos points to DatologyAI's customer results, from Thomson Reuters gaining on proprietary legal data in mid-training to Arcee's Trinity reaching the open frontier on public data alone. The closing argument is blunt: it is cheaper to manufacture high quality data than to buy more compute, so data curation is quietly shaping the future of model training. Speaker info: - https://x.com/arimorcos - https://www.linkedin.com/in/arimorcos/ - http://www.arimorcos.com/ Timestamps: 0:00 - Data is all we think about 0:52 - Why good data became scarce 2:19 - Swapping compute for data on the curve 3:48 - An oil refinery for data 5:52 - Curation work at DatologyAI 6:54 - Proof: small models beating bigger ones 8:58 - Inference efficiency from better data 9:24 - Multilingual gains 12:16 - Synthetic data done right 14:10 - Thomson Reuters and Arcee results 17:43 - Cheaper than buying compute canonical: >- https://liquidgold.cc/channel/aidotengineer/explore/video/data-quality-is-the-compute-multiplier-ari-morcos-d-478874ee429f markdown: >- https://liquidgold.cc/channel/aidotengineer/video/data-quality-is-the-compute-multiplier-ari-morcos-d-478874ee429f.md llms: null parent: 'https://liquidgold.cc/channel/aidotengineer.md' channel: aidotengineer updated: '2026-10-24T06:29:09.222Z' language: null children: [] videoId: _PdK6x7PQNM durationSeconds: 1145 publishedAt: '2026-07-31T23:00:06.000Z' searchChunks: - ordinal: 0 text: null startSeconds: 24 endSeconds: 58 watch: 'https://liquidgold.cc/watch/_PdK6x7PQNM?t=24' heldReason: rights - ordinal: 1 text: null startSeconds: 58 endSeconds: 70 watch: 'https://liquidgold.cc/watch/_PdK6x7PQNM?t=58' heldReason: rights - ordinal: 2 text: null startSeconds: 70 endSeconds: 143 watch: 'https://liquidgold.cc/watch/_PdK6x7PQNM?t=70' heldReason: rights - ordinal: 3 text: null startSeconds: 143 endSeconds: 153 watch: 'https://liquidgold.cc/watch/_PdK6x7PQNM?t=143' heldReason: rights - ordinal: 4 text: null startSeconds: 153 endSeconds: 230 watch: 'https://liquidgold.cc/watch/_PdK6x7PQNM?t=153' heldReason: rights - ordinal: 5 text: null startSeconds: 153 endSeconds: 230 watch: 'https://liquidgold.cc/watch/_PdK6x7PQNM?t=153' heldReason: rights - ordinal: 6 text: null startSeconds: 153 endSeconds: 230 watch: 'https://liquidgold.cc/watch/_PdK6x7PQNM?t=153' heldReason: rights - ordinal: 7 text: null startSeconds: 230 endSeconds: 241 watch: 'https://liquidgold.cc/watch/_PdK6x7PQNM?t=230' heldReason: rights - ordinal: 8 text: null startSeconds: 241 endSeconds: 358 watch: 'https://liquidgold.cc/watch/_PdK6x7PQNM?t=241' heldReason: rights - ordinal: 9 text: null startSeconds: 241 endSeconds: 358 watch: 'https://liquidgold.cc/watch/_PdK6x7PQNM?t=241' heldReason: rights - ordinal: 10 text: null startSeconds: 241 endSeconds: 358 watch: 'https://liquidgold.cc/watch/_PdK6x7PQNM?t=241' heldReason: rights - ordinal: 11 text: null startSeconds: 241 endSeconds: 358 watch: 'https://liquidgold.cc/watch/_PdK6x7PQNM?t=241' heldReason: rights - ordinal: 12 text: null startSeconds: 369 endSeconds: 424 watch: 'https://liquidgold.cc/watch/_PdK6x7PQNM?t=369' heldReason: rights - ordinal: 13 text: null startSeconds: 434 endSeconds: 546 watch: 'https://liquidgold.cc/watch/_PdK6x7PQNM?t=434' heldReason: rights - ordinal: 14 text: null startSeconds: 434 endSeconds: 546 watch: 'https://liquidgold.cc/watch/_PdK6x7PQNM?t=434' heldReason: rights - ordinal: 15 text: null startSeconds: 546 endSeconds: 558 watch: 'https://liquidgold.cc/watch/_PdK6x7PQNM?t=546' heldReason: rights - ordinal: 16 text: null startSeconds: 558 endSeconds: 569 watch: 'https://liquidgold.cc/watch/_PdK6x7PQNM?t=558' heldReason: rights - ordinal: 17 text: null startSeconds: 580 endSeconds: 745 watch: 'https://liquidgold.cc/watch/_PdK6x7PQNM?t=580' heldReason: rights - ordinal: 18 text: null startSeconds: 580 endSeconds: 745 watch: 'https://liquidgold.cc/watch/_PdK6x7PQNM?t=580' heldReason: rights - ordinal: 19 text: null startSeconds: 580 endSeconds: 745 watch: 'https://liquidgold.cc/watch/_PdK6x7PQNM?t=580' heldReason: rights - ordinal: 20 text: null startSeconds: 745 endSeconds: 857 watch: 'https://liquidgold.cc/watch/_PdK6x7PQNM?t=745' heldReason: rights - ordinal: 21 text: null startSeconds: 745 endSeconds: 857 watch: 'https://liquidgold.cc/watch/_PdK6x7PQNM?t=745' heldReason: rights - ordinal: 22 text: null startSeconds: 868 endSeconds: 1063 watch: 'https://liquidgold.cc/watch/_PdK6x7PQNM?t=868' heldReason: rights - ordinal: 23 text: null startSeconds: 868 endSeconds: 1063 watch: 'https://liquidgold.cc/watch/_PdK6x7PQNM?t=868' heldReason: rights - ordinal: 24 text: null startSeconds: 868 endSeconds: 1063 watch: 'https://liquidgold.cc/watch/_PdK6x7PQNM?t=868' heldReason: rights - ordinal: 25 text: null startSeconds: 1073 endSeconds: 1145 watch: 'https://liquidgold.cc/watch/_PdK6x7PQNM?t=1073' heldReason: rights - ordinal: 26 text: null startSeconds: 1073 endSeconds: 1145 watch: 'https://liquidgold.cc/watch/_PdK6x7PQNM?t=1073' heldReason: rights --- # Data Quality Is the Compute Multiplier — Ari Morcos, DatologyAI Swap compute for data on the scaling curve and the same money buys a better model, which is why Ari Morcos calls data quality the compute multiplier and the most underinvested part of training. His frame is an oil refinery for data rather than a firehose: clean, curate, create, and compose, with quality classifiers, deduplication, and synthetic generation each earning their place, and the sequencing across stages mattering as much as any single step. The scarce resource now is not tokens but signal per token, and finding data that is optimal for a given target is where the leverage hides. The proof points are concrete. Better curated data lets a small multilingual model beat far larger ones trained on many more tokens, and it buys real inference efficiency because a model reaches the same quality with less. Morcos points to DatologyAI's customer results, from Thomson Reuters gaining on proprietary legal data in mid-training to Arcee's Trinity reaching the open frontier on public data alone. The closing argument is blunt: it is cheaper to manufacture high quality data than to buy more compute, so data curation is quietly shaping the future of model training. Speaker info: - https://x.com/arimorcos - https://www.linkedin.com/in/arimorcos/ - http://www.arimorcos.com/ Timestamps: 0:00 - Data is all we think about 0:52 - Why good data became scarce 2:19 - Swapping compute for data on the curve 3:48 - An oil refinery for data 5:52 - Curation work at DatologyAI 6:54 - Proof: small models beating bigger ones 8:58 - Inference efficiency from better data 9:24 - Multilingual gains 12:16 - Synthetic data done right 14:10 - Thomson Reuters and Arcee results 17:43 - Cheaper than buying compute Canonical: https://liquidgold.cc/channel/aidotengineer/explore/video/data-quality-is-the-compute-multiplier-ari-morcos-d-478874ee429f · Channel: [AI Engineer](https://liquidgold.cc/channel/aidotengineer.md) > Search chunk text is held until public-source provenance is verified. ## Document chunks > Missing: no stored document chunks are available.