What does HackerNews think of Open-Assistant?

LLama hasn't been fine-tuned with RLHF, so it requires additional prompting, check out the open-assistant[0] project for an open-source ChatGPT equivalent (WIP).

[0]: https://github.com/LAION-AI/Open-Assistant

The folks making Open Assistant [1] (opensource ChatGPT clone) gathered enough data to start initial training, so hopefully there will be something to play with soon.

[1] https://github.com/LAION-AI/Open-Assistant

LAION is working on an open chatgpt like model [0] and there is a project discussed here recently that is about running the llms on single gpus [1]

I agree, there is a very serious risk of newer AI advances being closed by default, and I think it's not being taken seriously enough.

Even "open" players are still for-profit, say Huggingface. There really needs to be Linux or FSF like players that are motivated by freedom.

[0] https://github.com/LAION-AI/Open-Assistant [1] https://github.com/FMInference/FlexGen

I'm stupid in this area and have only a very broad understanding of "AI". But isn't this already happening in this project? https://github.com/LAION-AI/Open-Assistant

I think this is a sad way to go. Hardware is super expensive to develop and the key IP here is in the software. There's probably some fancy microphone tech, but ultimately you could (should?) run Mycroft on a Raspberry Pi with an edge accelerator device.

Technically I think we're getting close to a good offline system. Home Assistant solves the practical aspect of controlling systems and triggered routines - it really works well, at least on par with vendor software. Speech recognition models are usable and getting better. The interesting "glue" is parsing the commands into something reasonable and I think open source LLMs are likely to be the answer. See Open Assistant [0] for example. It doesn't solve the knowledge base issue (e.g. people seem to use Alexa either as a very fancy egg timer, or for querying facts), but it would probably allow for very natural interaction with the device and the model could just translate a query into a command that triggers an action.

My impression is that Amazon's approach (I have no knowledge of what goes on inside) has been to build fairly complicated subsystems to handle very specific queries. Each "Skill" has to be separately developed and if you ask something outside a fairly rigid grammar, the device has no idea.

[0] https://github.com/LAION-AI/Open-Assistant

They made T5 and released the weights, while T5 "only" goes up to 11B params, it's an efficient model for a wide variety of tasks and the base for a lot of other models/projects. Of note imagen[1] and open-assistant[2] use T5 and FlanT5.

[1] https://imagen.research.google/

[2] https://github.com/LAION-AI/Open-Assistant

Open-Assistant - https://github.com/LAION-AI/Open-Assistant - is an interesting open source project I found this morning.

As others have pointed out, running a truly large language model like GPT-3 isn't (yet) feasible on your own hardware - you need a LOT of powerful GPUs racked up in order to run inference.

https://github.com/bigscience-workshop/petals is a really interesting project here: it works a bit like bittorrent, allowing you to join a larger network of people who share time on their GPUs, enabling execution of models that can't fit on a single member's hardware.

oh damn https://github.com/LAION-AI/Open-Assistant

cool stuff, thanks

There’s LAION working on open source[1] version of chatGPT

[1] https://github.com/LAION-AI/Open-Assistant

LAION’s Open Assistant trying to address that with gamified crowdsourcing.

https://github.com/LAION-AI/Open-Assistant

There is Open Assistant by LAION (WiP): https://github.com/LAION-AI/Open-Assistant