Sunday, October 25, 2015

Reflections on "don't break the chain"

For about 1 month I tried writing up one idea per day on this blog. This was largely a selfish task, a way to get some ideas down that had been swimming around my head and distracting me. I find that once I do this they stop diverting my attention, and it allows me to start working toward an implementation or something actionable (these are also the main reasons I use TODO lists, Wunderlist being my current favorite).

I was also testing the Seinfeld "don't break the chain" productivity approach.

Some observations

  • The biggest problem with the "uninterrupted chain" approach is that blog writing represents only a small fraction of my overall writing output, and my writing output is itself only a small fraction of my overall work output. I think that writing something everyday is key, but writing up new ideas is probably less useful than working on papers or posts that communicate the technical work I am doing. Perhaps I need a "meta-calendar" that would allow me to string together days that I write in any forum for a particular audience. For example, I should probably keep an academic writing calendar separate from a calendar keeping track of public-facing posts.
  • Writing new ideas in a public forum is tricky when you have to worry about patent issues, which is yet another reason why the patent system is so badly broken.
  • Writing a new idea every day is a good forcing factor, but it might be better to linger on particularly good ideas for a few days. In this spirit I am planning to return to some of my posts and beef them up a bit with mock-ups.
  • It is much more fun to write when you aren't too worried about related work. I think in academia the literature survey can be a terrible, idea-quashing instrument. Sure, somebody probably did something like your thing at PARC in 1992, but it is best to take a few minutes to get your idea down first. Almost certainly yours has some novel components, and it might be more than that. At worst, you've lost just a few minutes of your time, and as time wasters go writing down ideas isn't so bad.

Thursday, October 15, 2015

Asynchronous, multimedia email

This is actually a very old idea, but it is remarkable we don't see it yet. It would be useful to record video snippets via webcam (or screencapture) and send them in a threadable conversation the same way we do email. Such a tool would also be a useful compliment to video conferencing tools (especially when connectivity makes synchronous sharing impossible).

Why isn't this a thing? With modern speech-to-text, the content would be searchable. Bandwidth isn't really an issue. Is it only reception? Reading a screen is easier (and more socially acceptable) than listening to a series of clips (if you don't have headphones). Is that all?

Patent mashups

So many patents are simply combinations of previous ideas, I wonder if you could build a tool that generates new patents, or new ideas at least, by intelligently selecting claims to combine together. Of course, glomming claims together willy-nilly won't provide much value, so the tool would need to learn combinations that are rational.

Tuesday, October 13, 2015

Compounding knowledge?

This fascinating piece on animal intelligence adds to the growing body of work suggesting that human intelligence is different from animals only by a matter of "degree, not kind." If we admit that, though, we must also admit that, at least at this point in history, overall human production is different from animals in kind. One way to reconcile these two conclusions is the theory that human intelligence is different precisely in that it facilitates knowledge compounding over time. Just like a small initial difference in an investment can turn into a large difference after many years of compounding interest, so too can an incremental difference in intelligence eventually lead to a vast difference in achievement (for the moment, I'm setting aside the issue of whether this achievement is good or bad).

This line of reasoning, though, could lead one to something like the Singularity: if human abilities are growing exponentially then eventually they reach a near-infinite level. I think this conclusion is largely a fallacy of metaphor. Closer to the truth is that humans compound knowledge at different rates in different domains. The phones in our pockets run as fast as supercomputers did only decades ago, but we've still not put anyone on Mars (or even returned to the Moon, for that matter). Or to use a more mundane example: besides the ubiquity of mobile devices the interiors of homes are largely unchanged over the last 60 years.

Perhaps it is best to say that human ability is an irregular field whose area is monotonically increasing.

Monday, October 12, 2015

Solid data on firearms

This one is short, sweet, completely obvious, not novel, yet somehow incredibly difficult.

Regardless of what side of the debate you're on, we've got to have more data on gun use in the US.

Sunday, October 11, 2015

Education is reputation

The term disruption is hackneyed in Silicon Valleyyet there are a few massive bureaucracies that do seem dearly in need of it but that are stubbornly resistant to change. The medical-insurance complex is one, but the sheer amount of money involved coupled with massive legal hurdles at least makes its resilience understandable.

Academia less so. While any bureaucracy has significant political power as well as legal and fiscal interests, they would seem less pronounced in academia than in e.g. medicine. I don't quite understand the glue that keeps it together. These experiences of a university math lecturer are a typical refrain of a system creaking under a terrible weight of bureaucracy, with its in-fighting, internal power struggles, and political gambits taking a front seat to intellectualism.

Probably most of the glue is in reputation networks. This implies that new educational ventures should cooperatively leverage established systems.

Saturday, October 10, 2015

Better tools for everyday capture and sharing

With the near ubiquity of GoPro cameras, this feels like a solved problem. But I don't think we are quite there yet. I think that what people most want is a way to record and share their own experiences without having to think about technology at all. This would require not only capture devices and software capable of taking high quality shots, but also automated filtering and rendering algorithms (either pre- or post-capture). The tool should also focus on photos, quick videos, and animated PNGs.

20th century critiques of photography often centered on its unnaturalness. As Sontag and others wrote, it has a way of freezing something in time so that the photographer can gain control of the moment. There may be some truth in this, but we now know that these type of critiques are largely beside the point, which is that the network is at least as important, if not more so, than the photograph itself. This is because the value of a photograph is more about providing others with some context about your situation than it is documenting moments for one's own sake. It is for this reason that longer video is not as useful: it is harder for others to consume (and, again, harder on the network). This also implies that volume can quickly make photographs useless because, again, a recipient can't be expected to sift through a massive archive.

If the main focus is on sharing, then this also implies that quality is somewhat less important. At least, it implies that the bar might be low enough for automated techniques to have a chance.

One key problem is the capture device itself. Drones are problematic for many reasons, not the last of which is that they are likely to be illegal in many outdoor (and indoor) settings. GoPros are close, but are probably too cumbersome for most activities. We need something like a panoramic camera that can be attached flexibly to equipment or clothing that will take photos at an extremely high resolution, and coupled with an automated editing process that will create a personalized and relatively terse archive of photos and animations selected from portions of the panorama at key moments.

Finally, many automated techniques focus on creating a narrative of an event, I doubt this is necessary. I think people just want to feel like they've captured and communicated the gist of an event (in a way that makes them look good, if possible ;)

Friday, October 9, 2015

Smartwatch app for meteor logging

Marking down the path and intensity of meteors during showers can provide useful data for astronomers. It can also simply be a nice way of logging what can be a spectacular event. But I always found it difficult to mark down a meteor's path on a paper sheet. It would be much easier to directly retrace the path through the sky with a gesture. Intensity would be slightly harder to log, but it is usually done with reference to another star or planet of similar magnitude. Perhaps the app could have four stages: 1) gesture path; 2) feedback that path has been accepted with option to manually correct it on an associated phone; 3) point to a reference star; 4) feedback that magnitude has been set.

A similar approach might be useful for other logging tasks as well, for example wilderness tracking, perhaps with speech-to-text to tag the species, etc.

Thursday, October 8, 2015

Everyday microloans with microinterest

ING Direct, now Capital One 360, has a nice feature where if your checking account goes below zero you get a loan for the negative amount from the bank at a competitive rate. In most cases this happens only rarely and only for small sums, so that the actual amount that customers pay for the loan is on the order of pocket change. Still, this benefits the bank since they profit from the scale of the loans.

Perhaps it would be useful for people to do the same with each other. Of course you'll loan a friend money with no strings attached, but attaching a small amount of interest might make everyday microloans more likely among people who don't necessarily have strong social ties. I am not sure about to mitigate the risk of defaults, though. Perhaps if the transactions were publicly inspectible (similar to Bitcoin)?

Wednesday, October 7, 2015

PR-to-scientific sentiment ratio for products

In high school economics we learn that 20th century economic theory was more-or-less based on the assumption of rational actors: people and corporations allocate assets such that overall utility is maximized. A more recent vein of research, behavioral economics, makes an argument that feels self-evident in hindsight: people don't make rational decisions at all times, but are in fact often motivated by completely irrational internal or external factors.

In their recent book (reviewed here), Akerlof and Shiller go further to suggest that people are not only making sometimes irrational decisions but that they are beguiled into thinking they need things they don't. More than that, they argue that the very nature of capitalism guarantees this kind of behavior from companies.
Phishermen know how to give rise to temptations, thus generating novel “needs.”
This has to be a disturbing thought to anyone in the technology industry. Are we fulfilling actual needs people have, or monkey-on-the-shoulder needs invented out of whole cloth?

I think we can safely say that we do need more products that help consumers make better decisions about purchases. One somewhat snarky app I'd like to have would scan a product barcode and do a quick calculation of a lobbyist/PR-to-scientific research sentiment ratio. The higher the ratio the more suspicious the product: a company that spends tons on PR and lobbyists but which scientific research has found to be problematic is probably pushing something that is not in the public's best interest.

Tuesday, October 6, 2015

Measuring the impact of sports on mental state

Dios Mio. Stephortless.

My favorite thing about this is how ecstatic everyone is.

Sports are often poo-pooed in academia, but they are a big part of most people's lives. Perhaps it would help their legitimacy if we had better data on the extent to which they improve people's moods. This is difficult to get a handle on and might require mining social media posts, etc, to get a rough sense of how being a fan changes people's outlook, emotional state, and overall health.

One thing I suspect is that the negative effects of alcohol might well erase most of the benefits of positive affect. Still, surely enjoying Curry or The Brow over 1 or 2 drinks with friends makes ones life better in a measurable way.

Monday, October 5, 2015

Design pattern recommender

Good artists copy; great artists steal. - P. Picasso (sort of)
Designers, artists, and engineers for that matter are always looking for ideas to integrate into or inspire their own work. This is one reason why Tumblr and Pinterest sites are popular, especially among visual designers: they make it easy for people to mine past work for new ideas.

This research, which uses deep learning to suggest fonts that match those captured in images, can be similarly useful. The difference is that in this case designers can use the real world to not only inspire them but also put them well on their way to utilizing found fonts in their own work.

What other types of recommender systems could be based on image or video recordings of physical objects? Could a photo link you to an architectural design pattern? A chair style? An approach to color use? Layout organization? Dance style?

Sunday, October 4, 2015

Clearing up some confusion: Lone Tree Prairie

I am taking a break from ideas to write-up a mystery that I've cleared up for myself.

The Forest of Nisene Marks is a remarkable area, especially for old growth redwoods. Most folks are familiar with the Advocate Tree near the front of the park, but that barely scratches the surface. At any rate, for a long while I've been a little confused about Lone Tree Prairie. Now after revisiting the site at Aptos Creek Fire Road I see that my issue stems from the fact that the Amended General Plan for The Forest of Nisene Marks State Park is riddled with inaccuracies. Chief among them is confusing Lone Tree Prarie with Sheep Camp Meadow.

It is a reasonable mistake to make. There is no prairie at Lone Tree Prairie, only an old growth redwood (which the plan calls the "Hollow Tree" but which everyone else calls Lone Tree). The two combinations of maps shows that this is the case. First, from the USGS:

And now Google Maps:

From these two maps you can see that there is no prairie at Lone Tree Prairie, but there is one further up the trail. This is Sheep Camp Meadow. For verification we can reference Jeff Thomson's guide to the park:

The Plan makes this mistake everywhere, even labeling grassland where there is none:

Again, this mistake surely comes from confusing Lone Tree Prarie and Sheep Camp Meadow.

I think the fact that the Plan is so error prone clears up another point of confusion for me. It notes that there is a virgin redwood 60 feet to the east of the "Hollow Tree". I've explored the are thoroughly: there isn't. However, there is pretty clearly one to the west. A copy editing mistake, I'm now sure.


It looks like OpenStreetMap also gets the location wrong. Huh.

Saturday, October 3, 2015

Lists and implicit prioritization

  • One problem
  • with lists is that
  • something has to be first
  • and something has to be 
  • last.
This fundamental property of lists is useful for priorization, but problematic when all items should be weighted more-or-less the same or when items can't be organized ordinally. For me, this problem has come up most often when listing authors on academic papers. I've co-written several papers on which each author contributed equally, but nonetheless someone had to be listed first (more than once we resorted to coin flips to figure out author order).

For academic papers there might be a somewhat straightforward solution. Authors usually list their affiliation and contact information just below their name. You could use this same space to list their role with respect to the paper. This could be an indicator of contribution level (primary, secondary, etc.) or explicitly refer to roles (intern, mentor, editor, coordinator, implementor, designer, evaluator, etc.). Obviously you'd want to allow authors to add multiple tags, duplicate tags, or ignore them altogether.

I'm not sure this generalizes to all lists, but it's a start.

Friday, October 2, 2015

Neighborhood digital library

It would be great to have something like this, but for digital documents. The Plasma Poster is one approach, but I think the reason this works is because there's not much of value here (and no electrical components to maintain).

But how can we share digital documents at a completely analog kiosk? We need a way of pressing a link to a document (e.g., a QR-code) into paper. One approach might be a light activated stamp ink. That way you could display a QR code on your phone screen, make an impression in the stamp, then stamp a piece of paper that you could leave at the library. Another low-fi approach (from the kiosk's perspective anyway) would be deformable phone screens. The screen could then deform to match a QR code. With a piece of paper placed over the screen and a pencil you could then easily create a rubbing of the code.

Thursday, October 1, 2015

Better MOOCs

I think MOOCs are the electric cars of academia: poor performance out-of-the-gate but ultimately better in almost every way. Particularly promising are MOOCs focused on tailored, individualized education recreating "one-on-one tutoring" that students get in office hours. This is an approach that academics should be excited about because adaptive instruction will surely drive up the demand for lecturers. Cost savings to students will come from eliminating the brick-and-morter rubbish (stadiums, amenities, and the admins needed to run them).

I recently witnessed a purely digital course that had some flavor of this new model. Students were engaging with each other, course notes and questions, and the professor more than I've seen in most physical classrooms. One of the great benefits of a pure digital approach is that the professor can easily rearrange everything: a lecture goes to a breakout to a group discussion with a button click. Oh and? There’s no sitting at the back of the room playing Minecraft, etc. (Of course, the quality of the instructor still matters immensely, and in that regard the students of this class were very lucky.)