Will Transformers Replace CNNs in Computer Vision? + NVIDIA GTC Giveaway:
https://www.youtube.com/watch?v=QcCJJOLCeJQ
--
Artificial General Intelligence List: AGI
Permalink:
https://agi.topicbox.com/groups/agi/Tefaeb8e790a54cec-Me5095e2a2845fc9c283c0278
Del
Yeah I've been saying for a while that system identification of PDEs is
likely where it's at given the need to hook up with the empirical world's
priors with a formal system that's Turing Complete -- at least
potentially. Indeed the second connectionist summer would likely not have
happened if not
When I said that ANNs used linear approximations you knew what I meant because
'you are in the club.' But a newbie might have been confused and thought
something like, "So that's how Neural Networks work. They use linear
approximations." Seeing this I will try to find better phrases like - they
Transformer Attention does seem to be more than just those two fundamental
points.
I do not want to spend a lot of time working with NNs (other than on my TinyML
projects) but I do want to get a better understanding about how these things
work and then apply some of the ideas to some slightly m
Looks kewl, I wouldn't tend to trust them very far but it sounds like a
great way to obtain priors for more conventional methods to squash out
the last little bits of epsilon u don't want.
Bill Hibbard via AGI wrote:
> Interesting article:
> https://www.quantamagazine.org/new-neural-networks-solve
Interesting article:
https://www.quantamagazine.org/new-neural-networks-solve-hardest-equations-faster-than-ever-20210419/
Points to a couple arxiv papers:
https://arxiv.org/abs/1910.03193
https://arxiv.org/abs/2010.08895
--
Artificial General Intelligence
On Tuesday, April 20, 2021, at 1:59 AM, Ben Goertzel wrote:
> In general though, I think the academic community has not adapted fast
enough to the shift to online publication, which means that cost of
printing on paper is not an issue anymore.
I never thought of that. I suppose it's this way with
