Extending Neural Networks to higher dimensions (article)

[scottdave]

TL;DR Summary

I came across this interesting article: An Idea from Physics helps AI see in Higher Dimensions. The idea is something called gauge equivariance.

Neural networks have come a long way since I first took a course, 25 years ago. Now that I'm in the Online Masters in Analytics program at Georgia Tech, I see these topics come up often. I found the following article an interesting read.

https://www.quantamagazine.org/an-idea-from-physics-helps-ai-see-in-higher-dimensions-20200109/

 

[jedishrfu]

There’s been a real explosion in this area of comp sci with the advent of big data and greater computing power. It’s hard to see how it will all fit together in the end. Basically a collision of big data, comprehensive surveillance, privacy, quantum computing into the beginnings of AGI Ie artificial general intelligence using predictive analytics.

one comment I’ve heard from AI researchers is that the worst solution to any problem is machine learning. And yet that is what is happening right now. I imagine the next steps will be to use the nn to discover a pattern, convert it to a concrete algorithm for legal purposes ie non-discrimination issues or quantum computer algorithm for speed. Already NNs are being used to train NNs by adjusting their hyper parameters, so what will come next.

 

[anorlunda]

AGI Ie artificial general intelligence using predictive analytics

Well, that expands my vocabulary. Did you mean, "AGI i.e. artificial general intelligence using predictive analytics"?

 

[jedishrfu]

Yes, somehow AGI will need to anticipate situations. We do that now from our experience and what we’ve learned. The methods of predictive analytics can provide that anticipation feature.