We’re seeing major technical innovations at an increasing rate.
Just a few examples from the last couple weeks:
@QuasarModels just announced a custom attention architecture targeting 5M token context windows.
@IOTA_SN9 developed a technique that compresses data flowing between distributed GPUs by 128x with little to no loss in training quality, increasing viability of training large AI models across internet-connected machines worldwide.
We're seeing the building blocks start to form whereby competitive large generalized models can eventually be built.
In the meantime, we're also witnessing more targeted, niche players start to pull ahead in their respective fields.
During the presentation, I gave the example of @resilabsai achieving 90% accuracy on their home valuation model, making it the most performant open source model and quickly approaching state of the art.
Quite literally as I was explaining this during the talk, @markjeffrey pointed out they had just achieved 98% accuracy.
In the time between when I prepared the presentation and actually presented, they went from best open source to at or near state of the art - only further highlighting the unique value of Bittensor's open, competitive intelligence creation cycle.
Last week at the @YumaGroup Summit I had the opportunity to present on The State of Bittensor.
That presentation is in the thread below. If you choose to read it, I'd ask that you keep the following three things in mind:
1. This is just one guy's view of what was the most relevant for a 25-minute talk; a difficult filter for such a dynamic industry
2. The slides were designed to supplement a talk; I've done my best to replicate what I recall of the talk in the accompanying X posts
3. The topic of the Summit was "The Tipping Point" - a candid assessment of what could lead to Bittensor's breakout success and what evidence we see of that today - which also thematically anchored this presentation
Let's dive in: