Atomic Auto-focus - Computerphile

Machine Learning has allowed nano-scientists to Autofocus their equipment for the first time - at an atomic level. Professor Moriarty on the paper he wishes his group had published.

This video was filmed and edited by Sean Riley.

[infomaniac]

2:55 Jam it in and wiggle it around. Sounds very scientific.

[rich]

Hilarious that the bottleneck of atomic microscopes is the banging rocks together part. Full circle.

[AleifrLeifrson]

Excellent application of convolutional neural networks. Sure, it's neat to have a CNN that distinguishes pictures of dogs from cats, but this is a really innovative application that solves a real problem. Very cool.

[rentzepopoulos]

Shame I cannot like more than once... I feel the frustration and relief at the same time of Professor Moriarty. This is how science should be. I am thrilled!!!

[constructivist]

I love how you just let the profs talk, really makes Nottingham look like a place you want to go to uni for anything STEM.

[JustOneAsbesto]

2:13 I never realised Prof Moriarty liked big buts.

[bogdanyer]

this article single handedly dropped percent of phd positions in AFM/STM field by 90%

[trefod]

The wonder of science is that something as cutting edge as this, is readily understandable and perfectly logical without resorting to "trust me on this!" or "have faith!"

[dermathze]

Our school once visited a lab where they created images with *almost* this level of precision, so I kinda saw this live and I still can't believe how precice things like this can get.
Things like these are absolutely amazing.

[nigeldupaigel]

Shout out to the cameraman. Brilliant questioning.

[dipi]

3:06 … and this is exactly why I love the Nottingham videos and Professor Moriarty. Great stuff! Cheers!

[pixlark]

This is a really awesome application of machine learning. Glad to see videos here on computerphile talking about applications in other fields

[zacharyhizon]

Laboratory instrumentation is just amazing. It's the intersection of so many sciences: quantum physics, chemistry, molecular biology, computer science, machine learning, and mathematics topics such as geometry, linear algebra, calculus, and networks. Well tbh it _is_ laboratory instrumentation, and laboratory instruments are used in generally all the sciences. But building the instrument, it's so interdisciplinary and at the same time intricate...

[BigJonYT]

Isn't the last comment a chicken and egg problem? If you have single bonds sticking out of the reference samples to use to image and train the tip, how do you know they are single bonds ? And if you can reliably create single bonds on the reference material can't you use the sane process to create tips !?

[Bring_MeSunshine]

It's an interesting point that the body of ths topic highlights the inter-disciplinary nature of the problem. Nature doesn't draw the lines as we do; the physics, chemistry, maths and engineering are an augmented process. Amazing work and another intersting vid. Thanks, Sean and Phil

[underpowerjet]

Machine Learning could be used to help create complex nano structues in the future. Woah, I never really though of this would be a possibility. Super cool

[makinosfly]

amazing job guys... those videos should be mandatory to any student!

[_jb_]

This is just pure genius. Well done guys

[nonchip]

i like the area at 5:39 marked "tip shaping spot" ... nice way to say "giant crater from ramming a needle over and over into our carefully designed test pattern" :D

[johnhanson]

I'm in High School so most of these concept I don't grasp, but this is so interesting. I wish I understood it better...