AI That Plans in Plain English

Kari Jaaskelainen

12 Nov 2025 — 1 min read

What if AI made plans in plain English before acting? This research shows that giving an agent a language “brainstorm” step can make it smarter and more reliable.

Instead of choosing thousands of tiny moves directly, the system first writes a short plan in natural language, then a second model executes it. The team built a tough real-time strategy game where many units must be coordinated over long periods. From 76,000 human play examples of instructions paired with actions, they trained an “instructor” and an “executor.”

Agents that use language as a hidden plan beat those that just imitate human clicks. The compositional structure of language—breaking big goals into smaller steps—was key to better decisions.

Why it matters: clearer, more controllable decisions; easier transfer to new tasks; and a path to human–AI collaboration. Code, models, and data are available. Read more: http://arxiv.org/abs/1906.00744v5

Paper: http://arxiv.org/abs/1906.00744v5

Register: https://www.AiFeta.com

#AI #NLP #ReinforcementLearning #HierarchicalRL #GameAI #MachineLearning #HumanAIInteraction

Automating GDPR Compliance: A Roadmap for Companies and Law Firms

GDPR compliance is more than checkboxes. A new roadmap from the Privatech project shows how automation and machine learning can help companies and law firms assess—and even generate—privacy compliance. * Shift the focus to data processors’ real workflows: drafting policies, mapping data uses, documenting decisions. * Break compliance into machine-ready

FPGAs for Faster, Leaner Deep Learning: A Review of CNN Accelerators

Deep learning drives image search, robots, and medical scans. Most systems lean on CPUs and GPUs. This review asks: what if we run convolutional neural networks (CNNs) on FPGAs—reconfigurable chips you can tailor to the model? * Why FPGAs: custom dataflows, low latency, and strong energy efficiency—great for cameras,

Dynamic-K: Recommendations That Know When to Stop

Most apps show a fixed number of “top” items—say 10 movies or 20 products—assuming there are always enough good options. But that’s not always true: sometimes there are few relevant items, or some users are extra picky. The result? Filler recommendations. Dynamic-K flips the script. Instead of

Teaching chatbots to stop contradicting themselves (DECODE)

Teaching chatbots to stop contradicting themselves Ever had a bot say one thing, then the opposite a few turns later? This study introduces DECODE—a new task and dataset for spotting contradictions in everyday conversations, drawn from both human-human and human-bot chats. * New data beats existing natural language inference (NLI)