// Summary
Hanlin discusses the evolution of Massive Language Fashions and the significance of environment friendly scaling and deployment. He emphasizes the advantages of a decentralized strategy of many small specialised fashions over one big AGI mannequin managed by just a few firms. Hanlin explains the benefits of firms coaching their very own customized fashions, corresponding to knowledge privateness considerations, and offers insights into when it’s applicable to construct your personal fashions and the accessible tooling for coaching and deployment.
// Bio
Hanlin is the CTO & Co-founder of MosaicML, an ML infrastructure startup that permits enterprises to simply prepare large-scale AI fashions of their safe environments. Hanlin was beforehand the Director of the Intel AI Lab, liable for the analysis and deployment of deep studying fashions. He joined Intel from its acquisition of Nervana Methods. Hanlin has a Ph.D. from Harvard College and has printed in main journals and conferences corresponding to NeurIPS, ICLR, ICML, Neuron, and PNAS.
