Brana Rakic@BranaRakic
AGI will come from agents that learn through experience and share memories, not training.
A new paper from @GoogleDeepMind team outlines moving beyond stuffing human data into LLMs to an era of autonomous Agents learning through real-world experiences.
"A single LLM can perform tasks spanning from writing poetry and solving physics problems to diagnosing medical
issues and summarising legal documents. However, while imitating humans is enough to reproduce many human capabilities to a competent level,
this approach in isolation has not and likely cannot achieve superhuman intelligence across many important
topics and tasks. In key domains such as mathematics, coding, and science, the knowledge extracted from
human data is rapidly approaching a limit.
...
To progress significantly further, a new source of data is required. This data must be generated in a way that
continually improves as the agent becomes stronger; any static procedure for synthetically generating data
will quickly become outstripped. This can be achieved by allowing agents to learn continually from their
own experience, i.e., data that is generated by the agent interacting with its environment.
...
Our contention is that incredible new capabilities will arise once the full potential of experiential learning
is harnessed. This era of experience will likely be characterised by agents and environments that, in addition
to learning from vast quantities of experiential data, will break through the limitations of human-centric AI
systems in several further dimensions:
• Agents will inhabit streams of experience, rather than short snippets of interaction.
• Their actions and observations will be richly grounded in the environment, rather than interacting via
human dialogue alone.
• Their rewards will be grounded in their experience of the environment, rather than coming from human
prejudgement.
• They will plan and/or reason about experience, rather than reasoning solely in human terms
We believe that today’s technology, with appropriately chosen algorithms, already provides a sufficiently
powerful foundation to achieve these breakthroughs"
The key enabler for this shift is neuro-symbolic agentic memory. For agents to truly learn from experience, they need more than temporary context. They require semantically structured, persistent, and verifiable memory to collect and contextualize the richness of experience. That’s what the @origin_trail Decentralized Knowledge Graph (DKG) delivers.
More than a storage layer, the DKG is a symbolic memory infrastructure enabling:
•Structured representation with RDF graphs for logical reasoning and long-horizon retrieval (as some say "retrieval is reasoning")
•Provenance and trust via blockchain, anchoring knowledge in time
•Data ownership and sharing through NFTs and UALs, modular, interoperable, secure
In this architecture, knowledge graphs are not just a backend, rather the perfect symbolic memory companion to LLMs. Where LLMs offer fluent generalization, KGs provide logic, powerful retrieval and rich semantics. Together, they enable agents to store experiences, reason with logic and retrieve with purpose.
This is the foundation required for neuro-symbolic reinforcement learning systems, such as demonstrated by AlphaGo and AlphaProof, the kind that evolve, self-correct, and make explainable decisions in the open world.
Imagine agents that:
•Justify decisions with grounded, time-stamped knowledge states
•Query deeply contextual memory to guide multi-step plans
•Share segments of their memory selectively—with humans or other agents
In the era of experience, memory becomes a first-class citizen.
And @origin_trail DKG is where that memory will thrive.
Paper link: storage.googleapis.com/deepmind-media…
