2 ms·

We don't have the same definition of "without too much trouble".

• Vecr 2y ago
  It doesn't rely too much on further AI progress. Sure, you'd need a really big training cluster, and an inference cluster, and a supercomputer to put the results together and shuffle the large transformer model I/O, and a good team, and a huge budget...