Industrial AI Podcast AI in the Forest 2: From Annotation to Automation
6 snips
Oct 22, 2025 Jakub Tomczak, a senior research scientist at the Chan Zuckerberg Initiative and former professor, dives into the fascinating world of generative AI. He discusses how time series analysis represents the next frontier for AI, emphasizing its importance in understanding sequences and making predictions. Jakub challenges traditional machine learning paradigms and advocates for continuous learning systems that can adapt dynamically. Also explored are the implications of LLMs as tool-call interfaces and the critical need for high-quality data in AI applications.
AI Snips
Chapters
Transcript
Episode notes
Autistic Talent Powers Annotation
- Andreas from Responsible Annotation built a company hiring autistic people to do high-quality data annotation in Europe.
- He turned an NGO idea into a business that enables meaningful paid work and European grant eligibility.
Time Series Foundation Models Generalize
- Foundation models for time series can generalize across machines and processes, enabling one-shot or zero-shot transfer.
- This generalization makes industrial AI suddenly more accessible for practitioners without months of bespoke modeling.
Sequence Models Slash Compute Needs
- XLSTM/Tyrex-like architectures can cut CPU usage by ~5x compared to some transformer approaches.
- Using more efficient sequence models could drastically reduce infrastructure needs for large-scale deployments.