I think MS are super keen to be leading edge in the AI space and given they have deep pockets, they'll spend their money all over the place because they think the return will be many times any investment they make.

Most of the AI companies are way over valued compared to their actual value generation so MS aren't alone in doing this.

So they won't change direction, they'll just throw more money into the hole.

If KANs turn out to be the next great way to make a model, it will totally be the GIT hub heroes leading the charge. Which is why MS invest in open source too.

META who are all about stealing anything they can are also leading the open source charge.

So KANS will get adopted and the bleeding edge of their development will be new startups who will use different compute to move forward, and lots of home bods with their machines.

Some of those home bods will even use the existing LLMs to help them cut the code to move the whole thing forward.

People wonder why I'm so convinced that we are going to see an even greater rate of change and I mentione Ray Kurzweil's law of accelerating returns.

The idea that every advancement combined with massive communication pipes and information channels these days means the whole ends up greater than the sum.

We'll all have bottled intelligences at our hip in any endeavour from here on, and those BI's will only get smarter and faster.

Lots of interesting stuff to come. And none of the invested compute will be wasted. A slow horse is still a useful horse as long as one gives it tasks that suit slow horses.