This is a really interesting article on why the current approach to AI will fail and how different approaches are necessary.
Interestingly it states that current models are wrong ~1/4 of the time at best and that the scale up approach to compute driving the AI bubble is doomed to fail.
When something is wrong 1/4 of the time it is literally selling the customer lemons 🍋 by design.
Interestingly it states that current models are wrong ~1/4 of the time at best and that the scale up approach to compute driving the AI bubble is doomed to fail.
When something is wrong 1/4 of the time it is literally selling the customer lemons 🍋 by design.
