Hacker News new | ask | show | jobs
by jfarlow 2076 days ago
One issue is that the scales don't well-match. And there are actually many time-scales beneath us, not just a few (and that biology/life is working at them all - including those on the order of millions of years.

An event like a "protein folding" can take milliseconds, in a tube [1]. While atomic/biophysical/biochemical simulations have time-steps of femto-seconds.

[1] https://www.youtube.com/watch?v=gFcp2Xpd29I

From https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3890418/

"Traditional MD [Molecular Dynamics (protein folding)] simulations are limited in length by timestep limits. Studies by our group and others have shown that traditional MD is limited to timesteps of about 2 fs due to high-frequency resonance frequencies.1–3 Many biologically relevant motions occur on the microsecond to millisecond range, which is 9 to 12 orders of magnitude greater than the timesteps possible with traditional MD. Further, each step requires a costly force calculation (O(N) to O(N2)). As such, simulating medium-size proteins often requires months of computer time on a large distributed cluster such as Folding@home4,5 to simulate milliseconds of dynamics, while simulating a large protein (e.g. the β-2 Adrenergic Receptor) on biologically-relevant time scales (milliseconds through hours) using a standard desktop computer would take years. Thus, it is not feasible to simulate timescales of biological interest without substantial advances in MD methods."

12 orders of magnitude of difference in time is akin to the difference between causal events happening once per millisecond and those happening once per century. How do you show a movie capturing the nuance of someone's blink reflex, along with their birth life and death...

https://www.youtube.com/watch?v=uHeTQLNFTgU

https://www.youtube.com/watch?v=VdmbpAo9JR4

1 comments

> One issue is that the scales don't well-match. [...] many time-scales

Indeed, teaching of scale pervasively fails. Physical, temporal, and other. But it's also taught very very badly. Thus it seems an open question how well it might be taught, and to whom.

Might we teach it better? This[1] (mine) illustrates one speculative approach to teaching size down to atoms, for a young and outreach audience. And here's[2] an old attempt at helping develop a feel for torque, down to picoNewton-nanometers. Many years back I started on a temporal zoomer - don't know if it would have worked, but I've not seen anything similar since. I recall some work on teaching deep time in intro geology as being rather nice.

Interactives with physically realistic molecular motion are vanishingly rare, and I've never seen a one with temporal zoom. So I suggest we're not even trying yet to explore whether we can teach this well.

Those last two "Inner Life ..." videos are regrettably even more misleading than the original. Motion in the original was aphysical, but at least it was simple. Those two retain that aphysicality, and add aphysical jiggle, further obscuring how badly you're being misled.

Creating educational content to support an excellent understanding of science is ghastly hard. So much so, that I suggest we're not even really trying yet. Which would mean how hard it will be to teach, once such content exists, is necessarily an open question. I wish I knew of folks exploring it.

[1] first section of http://www.clarifyscience.info/part/Atoms (page loads slowly - was meeting prep, not intended to be public) [2] http://www.clarifyscience.info/part/ZoomB?v=A&p=CK6Ji&m=torq...

Very cool! I've actually built similar models for myself to build that intuition. They're so helpful and generally hard to find. I agree that scale is generally poorly taught.

I personally think having even a modest intuition for how to map physical knowledge to its appropriate 'scale' (time/space, from plank to universe) is one of the most straightforward ways to be "smart". And a great way to get to know the limits of our knowledge.

I still haven't seen a video that matches my intuition. Which is unfortunate. I want to build a version of your scaler there, but for VR that you can slide up and down along at least time/scale axes - and maybe additional ones too.

:) It does seem physical scale can serve as skeleton and scaffold to support learning and integrating knowledge of the physical world. I don't know of any attempts to really push on that.

> for VR

An unfinished lockdown project was a RL version of that hands-apart-to-zoom cartoon, with an atomic-bonding interactive, with realistic electron density, as content. Educational XR might eventually be so much fun.

> maybe additional ones too

Maybe Ashby diagrams in XR?! Someday.

I'd love an "measure explorer" - a densely fleshed out space of order-of-magnitude overviews. Length; velocity; accel; jerk; length/dollar; length/dollar^2; length/mass; etc, etc.

Years ago I prototyped an interactive where oom diagrams were tied with other content. So as you slide mass, you get different animals, and as metabolic rate scales with animal mass, you get heart beat. Life expectancy, etc. Heart to Hz, keyboard, and sound ripples spreading in space. Or explore equations: select ideal hand crank for torque, with a meter arm, a bacteria for mass, and a second later, hey, relativistic E. coli. :)

I've found the bottleneck on creating such to be finding usable media. Hopefully google will eventually do semantic scene indexing of youtube videos, and of images. If anti-trust doesn't break them. And sci-hub still exists. Finding "a bit of video with an X doing Y" is still prohibitively hard. And licensed for reuse? Perhaps structure the effort to survive being sued for aggressive use of Fair Use, or some country's similar exception, because copyright is a nightmare.