2 — Cool things this weekby@samin
139 reads

2 — Cool things this week

by Shreya AminOctober 12th, 2018
Read on Terminal Reader
Read this story w/o Javascript
tldt arrow

Too Long; Didn't Read

Week 2: Oct 6, 2018 — Oct 13, 2018
featured image - 2 — Cool things this week
Shreya Amin HackerNoon profile picture

Week 2: Oct 6, 2018 — Oct 13, 2018

See last week’s 1- Cool things this week

Category: Humans n.0 and/or World n.0 (n=1, 2,…)

This category will include breakthroughs that have the potential of enhancing humans or the world we live in.

Life Extension & Human Longevity with Dr. Aubrey de Grey

Aubrey de Grey is a pioneer in the field of “rejuvenation”. According to him, life extension is a side effect of making and keeping people healthier — he claims there is no intrinsic limit to how much of our lives this could extend. His research focuses on regenerative medicine and whether it can prevent the aging process. He is also the Chief Science Officer of the SENS Research Foundation, which is doing pioneering work on extending our lifespans.

In the interview they discuss concepts such as “pro aging trance”, “longevity escape velocity” and “comprehensive damage repair” that can sustain a human body.

(I’m in the process of writing some blogs on this topic. So if you’re interested in learning about it, follow me here).

Category: All about AI (and machines)

I’ll try to keep this category short, even though there are so many interesting things happening.

Optical Illusions Images Dataset (Arxiv link)

Robert Max Williams, Roman V. Yampolskiy

The authors compiled a database of over 6,500 images of optical illusions (collected from two websites and a smaller dataset of 500 hand-picked images) and then trained a neural network to recognize them.

Then they built a generative adversarial network (with no hyperparameter optimization) to create optical illusions for itself, but “nothing of value was created after 7 hours of training on an Nvidia Tesla K80”.

They talk about how naively applying methods from recent work on GANs doesn’t yield the same results. This optical illusion dataset is too small and GANs use large datasets (30,000 high resolution images of face — for example).

Apart from needing models capable of learning from such a small and limited dataset, they say a deeper understanding of human vision is also needed.

Concentric circles

You can download the dataset:

Images are currently hosted on the machine learning cloud platform “Floydhub.”

Could the world’s mightiest computers be too complicated to use? (Link)

Bronson Messer, an astrophysicist at Oak Ridge National Laboratory (ORNL) in Tennessee, is building supernovae simulations for Aurora, a machine that is due to come online in the early 2020s — it could be the first machine capable of one exaflop — a billion billion “floating point operations” per second.

This article talks about some challenges to building such a machine:

  • Power — “If for some magical reason you could actually build an exascale machine from today’s technology, nobody could afford to power it,” says McIntosh-Smith. Such a behemoth would need hundreds of megawatts, he says. As a rule of thumb, “one megawatt’s about a million dollars for a year”.”….“Moving data within a large-scale computer system consumes over an order of magnitude more energy than is needed to compute the data.”
  • Parallelism — building software and algorithms to run across potentially billions of cores is “tricky”. I love this example: “The trouble is an inherent weakness in parallelism, noted in 1967 by computer scientist Gene Amdahl. His eponymous law says that any program’s speed will always be limited by its least parallel part. Gropp compares it to transporting a planeload of passengers from Chicago to New York. Even if you replace the flight with instant teleportation, it will still take each passenger 2 hours to clear security.” Usually to deal with this problem, a system would split problems into the smallest possible sub-problems and then decide how to use the processors. But you still need to worry how the processors communicate with each other, and this is where the data movement problem from the above point comes up again. For example, Messer’s supernova simulations would split the star into separate parts, each simulated separately — but “calculating the impact of gravity at each time” step requires each processor to broadcast the mass of its chunk to all the other processors. Brain simulations also present a similar problem because each neuron connects to many others neurons. This approach, of breaking up the problem and then communicating information between processors, works fine at the petascale — millions of billions of flops — but would “simply devour too much time and memory at the exascale.” (Someone please tell me how to think about this better).

Rise of the machines

Even if we can solve these problems, they ask a deeper question: is the focus on flops the best way to improve performance? Many researchers don’t think this is necessarily the best goal anymore.

“Satoshi Matsuoka, who leads the development of Japan’s “exascale” machine, says the country has deliberately set different goals. The objective is to run applications up to 100 times faster than on their existing K supercomputer — a 10-petaflop machine — by focusing development on data movement rather than raw flops. He says candidly that the post-K computer will not hit 1 exaflop, but he is also confident it will run science applications faster than competing first-generation exascale machines.”

Category: Exploring space

This category will have all things related to space: making sense, travel, colonization, etc.

Chris Hadfield teaches space exploration (Link)

“Explore the unknown: Impossible things happen. At age nine, Chris Hadfield knew he wanted to go to space. He eventually went there three times, becoming a commander of the International Space Station. In his MasterClass, Chris teaches you what it takes to explore space and what the future holds for humans in the final frontier. Learn about the science of space travel, life as an astronaut, and how flying in space will forever change the way you think about living on Earth.”

Must watch trailer!!!

I haven’t taken this course, but I think I will. The trailer so super inspiring — check it out. If you want to take the course together so we can discuss, debate, learn, and motivate each other, then message me.

Upcoming telescopes should be able to detect mountains and other landscapes on exoplanets (Link)

“A total of 3,726 exoplanets have been confirmed in 2,792 systems, with 622 systems having more than one planet (as of Jan. 1st, 2018). And in the coming years, scientists expect that many more discoveries will be possible thanks to the deployment of next-generation missions.”

“The transit method directly measures the sky-projected area of a planet’s silhouette relative to that of a star, under the assumption that the planet is not luminous itself… This fact implies that there is indeed some potential for transits to reveal surface features, since the planet’s silhouette is certainly distorted from a circular profile due to the presence of topography.”

Category: What’s going on around me?

I’m going to try to highlight cool things around me — in NJ/NY area.

How artificial intelligence robots can support NJ’s underwater infrastructure (Link)

“We would like the robot to be able to do infrastructure inspections, ideally to assess the integrity of underwater infrastructure, make sure everything is intact, working properly, that there are no damages or defects. Or potentially from a security standpoint, that there are no anomalies planted on an underwater piece of infrastructure.” — Dr. Brendan Englot, professor of mechanical engineering.

Category: I know, but I can’t help it

This is exactly what the category title says.

I want this!!!

Everyone should want this

Nope, I don’t think so :

That’s it. See you next week.

If you liked what you read, be sure to comment or clap — as a new writer, it means a lot.

I’ll be doing this on a weekly basis, so please follow if you’re interested.