The Post-Transformer Era: A Research Dinner with Pathway
Transformers have become the dominant architecture in AI, but fundamental questions remain around memory, continual adaptation, reasoning, efficiency, and learning from experience. At the same time, new architectures must work within the realities of large-scale training, inference, data, and compute infrastructure.
This research dinner is intended exclusively for registered attendees of ML in PL Conference 2026. It will bring together experienced researchers and technical practitioners to discuss where current approaches fall short, which alternatives appear most promising, and what evidence and systems would be needed to evaluate them.
Joining the conversation
Jan Chorowski (LinkedIn | Google Scholar) is the Chief Technology Officer and co-founder of Pathway, where he oversees the engineering direction of the company’s AI architecture. A pioneer in neural network models, Chorowski was the first person to apply attention in speech recognition. His research — including collaborations with Geoffrey Hinton and Yoshua Bengio — has been cited more than 14,000 times and helped shape the field of sequence-to-sequence learning. Before co-founding Pathway, he held research and engineering roles at Google Brain and MILA. At Pathway, Chorowski ensures that the company’s research breakthroughs are translated into scalable systems capable of reasoning, adapting, and learning continuously in real-world environments. Co-author of the BDH architecture.
This will be a relaxed discussion around the table. The evening is intended as an exchange among peers working on advanced AI research and systems.
Who should apply
The dinner is intended primarily for:
Senior and principal research scientists.
Research and technical leads.
ML infrastructure and AI platform leaders.
Research engineers and distributed-systems specialists.
Technical practitioners working on large-scale model training, inference, data, or compute infrastructure.
Postdoctoral researchers and exceptional PhD candidates with directly relevant work.
Relevant areas include foundation models, sequence modeling, memory and continual learning, reasoning, recurrent and state-space architectures, distributed training, inference infrastructure, optimization, high-performance computing, interpretability, and efficient ML systems.
Practical details
📆 Friday, 9 October 2026
⏰ 18:00–20:00 CEST
📍 Powiśle, Warsaw
Dinner and drinks are hosted by Pathway. There are 22 guest places.
Registration is free but subject to approval. Applications will be reviewed based on research or technical relevance, substantive contributions, potential contribution to the discussion and the overall composition of the group.
The exact venue and arrival instructions will be shared with confirmed guests.