As AI's novelty fades, apps face high churn. The solution is personalization through memory and continual learning. This is a difficult systems problem because it requires a paradigm shift from today's stateless inference to a stateful model where weights are updated dynamically based on user interaction.

The next major evolution in AI will be models that are personalized for specific users or companies and update their knowledge daily from interactions. This contrasts with current monolithic models like ChatGPT, which are static and must store irrelevant information for every user.

Pre-training on internet text data is hitting a wall. The next major advancements will come from reinforcement learning (RL), where models learn by interacting with simulated environments (like games or fake e-commerce sites). This post-training phase is in its infancy but will soon consume the majority of compute.

Many AI projects fail to reach production because of reliability issues. The vision for continual learning is to deploy agents that are 'good enough,' then use RL to correct behavior based on real-world errors, much like training a human. This solves the final-mile reliability problem and could unlock a vast market.

Adaption.AI is bucking the trend of building larger static models to focus on continual learning. Their core mission is to 'eliminate prompt engineering,' viewing it as a crutch that signifies a model's failure to truly adapt and learn from user interaction in real-time.

Static data scraped from the web is becoming less central to AI training. The new frontier is "dynamic data," where models learn through trial-and-error in synthetic environments (like solving math problems), effectively creating their own training material via reinforcement learning.

The next evolution for AI agents is recursive learning: programming them to run tasks on a schedule to update their own knowledge. For example, an agent could study the latest YouTube thumbnail trends daily to improve its own thumbnail generation skill.

Demis Hassabis argues that current LLMs are limited by their "goldfish brain"—they can't permanently learn from new interactions. He identifies solving this "continual learning" problem, where the model itself evolves over time, as one of the critical innovations needed to move from current systems to true AGI.

A major flaw in current AI is that models are frozen after training and don't learn from new interactions. "Nested Learning," a new technique from Google, offers a path for models to continually update, mimicking a key aspect of human intelligence and overcoming this static limitation.