Re: "NASA to study "city-killer" asteroid as odds of hitting Earth go up"

In article <vojokq$2jp83$3@dont-email.me>, Titus G  <noone@nowhere.com> wrote:

>
But something so complex as AI with computer vision must take umpteen
billion lines of code. How long would it take for American scientists to
read and understand the Chinese open source?

Probably as long as it took for the Chinese scientists to read and understand
the American open source of the last generation.

But the honest truth is that the ML algorithms are not that difficult... it is
just one big array of weights and training the algorithm sets the weights
up and then you evaluate new input based upon those weights.

What is difficult is the training process... and then once you have trained
it, being able to debug it since you can't really tell what changes are
tracked down to what training input very easily.  But this is also the
difficult part about natural intelligence too.

I can't speak directly about Deepseek (which is both an algorithm and that
weighting set... as well as a service) because I am explicitly forbidden
from looking at it.
--scott
--
"C'est un Nagra. C'est suisse, et tres, tres precis."