Yes, fair point. I was trying to use the same comparison we are currently having between closed weights and open weights and their time gap. If there might be a similar time gap to what is possible with ordinary equipment.
Yes, a small open model that can run on today's hardware and that compared to a historic SOTA closed model with all in. What time difference do we think?
Thanks, yes, I meant even ordinary retail PCs, not specialized GPUs. At some point in time in history, SOTA closed models were at a level that compares to todays open models that can run on ordinary hardware.
Yes, I meant ordinary hardware which you find at home, like a current MacBook Air or equivalent Windows desktop. There must be a time frame when early SOTA LLMs were at a level that compares to open models that can run on ordinary hardware. But it's more like years rather than months. My rough guess would be 2-3 years. Which still would be amazing if we could get OPUS 4.5 quality within 2-3 years on an ordinary computer.
I think there are also implications for governments and states. Imagine when all this is done without human involvement, you want to have your legislation machine and AI ready to remove all the friction. In the future, states who are getting that will have a big advantage. Today, everything is designed to interact with humans, and even governments need to rethink that.
I found the creative idea interesting - like the humanoid robots who do not have front or back - it is always both. So they do not need to turn around.
The trap in my view is that they think they "do something AI" (the AI Adoption) and by thus completely miss the point that in 2-3 years "AI Native" might be a thing.
I see that in companies or at conferences, they are proud of what they have achieved and how good it works. Which makes them kind of blind on what is really coming, the tsunami that's just building up, requiring a complete re-think of what an organization looks like.