To the author thanks a lot for sharing this and for taking the time for the interview.
On the person interviewed... I must say I was very underwhelmed... It sounds like you're hearing politic speeches "yes we have this, that is great, and you will and to use and it's so much better than the others you can't even compare. It's very difficult to understand but we did it"...
I was surprised to see this out of such a high profile individual.
I would have loved to hear a bit more about the actual technical innovations and practical accomplishments.
I would love to see it running actual code. I kind of understood how they adapt the topology to the instruction and data flows, but I suspect software will be the biggest challenge here. In order to have high utilization, it'll need to fine tune not only the topology on the fly, but also the data flowing into the code.
To the author thanks a lot for sharing this and for taking the time for the interview.
On the person interviewed... I must say I was very underwhelmed... It sounds like you're hearing politic speeches "yes we have this, that is great, and you will and to use and it's so much better than the others you can't even compare. It's very difficult to understand but we did it"...
I was surprised to see this out of such a high profile individual.
I would have loved to hear a bit more about the actual technical innovations and practical accomplishments.
I would love to see it running actual code. I kind of understood how they adapt the topology to the instruction and data flows, but I suspect software will be the biggest challenge here. In order to have high utilization, it'll need to fine tune not only the topology on the fly, but also the data flowing into the code.