Some helpful info in this thread where I asked Jon about using Parallax for training larger frontier models. Their focus is to train models that are actually usable, rather than training a high parameter model just for the sake of training it (even if they potentially could already).
See also around 37:50 into this podcast Jon did with @markjeffrey where he mentioned ambitions about creating a frontier "Bittensor model" with Parallax: https://t.co/nZBxtJ9Gre.
Seems promising.
https://t.co/lu0Do44Pwf
8b was the lab, testing long context and the algorithms in general out to ~1t tokens. Next model is 5-10x larger (still setting exact arch from findings on 8b), then we learn grow repeat. Usage is the only metric of success IMO though for new models so there's no point jumping to massive models without proving this small-medium scale first.
@RadoTsc They clarified they won't be participating as a miner for this purpose, so no miner emissions will be directed to this. To my understanding, they believe they can fund this with revenue (which targets to exceed any costs related to it).
@CryptoTrigger1@chutes_ai@lium_io Yes agreed. I do think they need to bring back the buys and burns - seems to be the simplest form to get people's attention these days. Then you educate them on the rest once you got their attention.
Bittensor subnet @chutes_ai has partnered with Tencent Cloud.
Chutes is a launch partner for @TencentHunyuan’s new flagship AI model, Hy4, now available through its API.
Day-one access to leading AI models is business as usual on Bittensor.
@beffjezos https://t.co/fKhXn2FCN7
Check this out from @chutes_ai (building on Bittensor). I bet you'll be hearing/seeing a lot more of Chutes' Parallax in the months/years to come.
October on SN44, in a few lines.
-More public tracking tasks running on the subnet
-3 new public elements out
-A big chunk of example data from the cricket private track is now public
-A lot of quiet work on security and scale
-We can spin up new tasks faster now, with plenty more coming
And following IBC, one of the largest real-time graphics software companies is now testing a Studio endpoint, running on the football model built earlier this year under our previous incentive mechanism.
Meanwhile, we’re collecting feedback from Studio users for what comes next.
From here the pace only picks up, with way more tasks every month and one goal, making computer vision a commodity.
@jon_durbin@cjohan247 That makes sense, thanks for responding. If you could give a ballpark estimate, what would you consider the ETA being for usable small-medium models that can be used in a commercial context (assuming things go as 'expected')?