Alexandr Wang
Chief AI Officer at Meta; founder of Scale AI
Founded Scale AI at 19 to supply labeled training data to AI labs and governments. In 2025 Meta invested in Scale and he joined to lead Meta Superintelligence Labs.
Key arguments & positions
- Argues data quality and evaluation, not model architecture, are the binding constraints on AI progress.
- Frames AI leadership as a national-security matter and supports US government adoption of frontier systems.
- Has publicly argued that current benchmarks overstate model ability and that better measurement is the field's weak point.
Accomplishments
- Founded Scale AI in 2016 at 19 after dropping out of MIT; became the youngest self-made billionaire on record.
- Built Scale into the main data-labeling and evaluation supplier to major AI labs and to the US Department of Defense.
- Joined Meta in 2025 as Chief AI Officer leading Meta Superintelligence Labs, following Meta's investment in Scale.
Papers & key writings
Not a research author. His public arguments appear in congressional testimony, Scale AI reports and conference talks.
Links
In the library
Nothing of theirs is filed in the hub yet. Browse the full library.
Recommended next
Hand-picked from the hub based on what Alexandr Wang covers.
Companies Are Selling Employees' Slack Messages and Emails to Train AI (IBTimes UK)
An IBTimes UK article on companies selling workplace emails, Slack messages, and documents to AI developers, and the questions this raises about employee privacy, consent, and who profits.
Why this: Also about training data
CausalMix: Data Mixture as Causal Inference for Language Model Training
arXiv preprint (July 2026) proposing a causal-inference framing for choosing data mixtures in language-model training.
Why this: Also about training data
OpenDataLab Arena
Open evaluation arena from OpenDataLab where AI models are compared on shared tasks.
Why this: Also about model evaluation
OpenWALDO
Self-described open-source AI project and community that develops training data, tools, models, governance and provenance records in public.
Why this: Also about training data