This is Bloomberg Opinion Today, an updated Bayeux Tapestry of Bloomberg Opinion's opinions. Apple Inc. recently added audiobook narration to the growing list of occupations where algorithms are poised to replace humans alongside graphic designers, college essayists and limerick writers. Luckily, the fine art of newslettering remains (ahem) far beyond the capabilities of even the most sophisticated artificial intelligence software. Still, hope is at hand for those not fortunate enough to toil in the newsletter mines but still seeking gainful employment that won't disappear as robots take control.

Craft is a supplier intelligence company helping organizations accelerate data-informed business decisions. Our unique, proprietary data platform tracks thousands of real-time signals across millions of companies globally, delivering best in class monitoring and insight into global supply chains, among other company cohorts. Our clients, including Fortune 100 companies, government and military agencies, SMEs, asset management groups, and others, use our technology for supply chain intelligence, market intelligence and related use cases. Through our modular, secure, customizable portal, our clients can monitor any company they are working with and drive critical actions in real-time. We are a well-funded technology company with leading investors from Silicon Valley and elsewhere, but are not your typical data or SaaS startup.

We propose Continual Repeated Annealed Flow Transport Monte Carlo (CRAFT), a method that combines a sequential Monte Carlo (SMC) sampler (itself a generalization of Annealed Importance Sampling) with variational inference using normalizing flows. The normalizing flows are directly trained to transport between annealing temperatures using a KL divergence for each transition. This optimization objective is itself estimated using the normalizing flow/SMC approximation. We show conceptually and using multiple empirical examples that CRAFT improves on Annealed Flow Transport Monte Carlo (Arbel et al., 2021), on which it builds and also on Markov chain Monte Carlo (MCMC) based Stochastic Normalizing Flows (Wu et al., 2020). By incorporating CRAFT within particle MCMC, we show that such learnt samplers can achieve impressively accurate results on a challenging lattice field theory example.