3 ms·
Maybe I missed it in the paper but they did post the source code (Github) for their implementation? Is anyone working on creating their own infrastructure based
by FestiveHydra235 4y ago
Maybe I missed it in the paper but they did post the source code (Github) for their implementation? Is anyone working on creating their own infrastructure based on the paper?
- anigbrowl 4y agoUgh, and allow peasants to touch it? Tbh seeing Google research at the top of a paper these days feels like a red flag that I shouldn't get too invested in whatever cool new thing is on show. They're basically commercials for nerds - I still find their output interesting, but it's probably not going to be actionable.
- zoba 4y agoI have been working on this even before I was aware of the paper. Feels a bit weird to have something almost identical released. Stay tuned, I guess. I plan to keep working on my version.
- aymeric 4y agoAre you talking about errand-runner, or something else? Are you planning on open sourcing it?
- zoba 4y agoSomething else. Calling it GPTRPG at the moment. There really isn't much to share right now, other than this basic demo that doesn't even have AI: https://gptrpgagent.web.app/ https://gptrpgagent.web.app/ Yes I plan to make it open source.
- colanderman 4y agoI'll be interested to see your approach. I've been bouncing some ideas in my head but not implemented anything yet (and I might never as I'm ethically conflicted here, as agents gain properties associated with sentience/consciousness). Their approach to memory is interesting. I had been considering a tiered command-based approach -- "short-term memory" being an automatic summary of recent sensory inputs/command outputs; "long-term memory" being a detailed database queryable by the agent.