[00:00] (0.88s)
How impressive is Gro 4 for you?
[00:03] (3.12s)
If you look at the AIM benchmark, which
[00:05] (5.60s)
is an advanced math quiz, Grock 4 scored
[00:08] (8.64s)
100% on it. You're literally running out
[00:13] (13.76s)
It's got to be driving uh Google nuts
[00:16] (16.08s)
that Elon got this done in 28 months
[00:18] (18.32s)
from a from a cold start.
[00:20] (20.08s)
When he said he was going to put this
[00:21] (21.36s)
huge cluster together, every AI expert
[00:24] (24.00s)
in the world said, "You cannot get power
[00:26] (26.16s)
laws and coherence at that scale. You
[00:28] (28.08s)
just can't do it." Every AI expert is
[00:30] (30.08s)
like, "Oh, god dang, you did it."
[00:32] (32.00s)
The amount of compute and resources
[00:34] (34.08s)
again are going exponential. Now, it's
[00:36] (36.16s)
the real quality that differentiates the
[00:38] (38.16s)
top models between each other.
[00:39] (39.76s)
My big question is, where do we go from