Category Archives: Technology

CryptoTulip of the Year Watch: ICO = Initial Crap (sh)Oot ?

In this year’s CryptoTulip competition, Ethereum continues to run well, having still not solved its scaling and governance challenges—compounded by the technological suicide pact built in to the software (the “difficulty bomb”)  and a new “war” on ASICs.

Several new technologies may become dark horse contenders for the CryptoTulip award, including and maybe even FOAM, which just went live.

But the main competition for defending champion Ethereum must surely be ‘the ICO’.  Once enthusiastically vaunted as the coming thing (was it only a year ago?), ICOs have sunk into a disappointing wave of bipolarity.

Some ICOs work well (if not necessarily fairly) and some raise piles of cash.  But analysis shows that many are flat out scams.

And a new report finds that almost half of all ICOs raise no money at all [1].  (What’s worse than a scam?  A scam that doesn’t make money!)

In both reports, some ICOs did indeed raise money.  The Research LeadMeri Paterson report says that 40% of ICOs made $1 million for the investors.  Notably, they found that even the successful ICOs, the gains are really short term: the money comes in the first few weeks.

So, the data indicates that investing in an ICO (if you are able to get in) is nearly a coin toss.  You might win, you might lose. I guess the good news is that you’ll know pretty quickly, because it will be over in weeks.

There seems to be a never ending supply, so you can play this lottery over and over if you want to.

ICOs are a very Nakamotoan “innovation”. They “disrupt” the foundations of raising capital by offering opaque, unregulated, and instant trades—with a huge, huge dose of caveat emptor.

And if there is anything that beats mania for virtual tulip, it’s mania for fake virtual tulips!

Surely ICO technology is a strong candidate for CryptoTulip of the Year in 2018.

  1. William Benattar, Meri Paterson, Eitan Galam, and Emmanuel Alamu,  Charting the growth of cryptocurrencies. GreySpark Partners 2018.
  2. Aditi Hudli (2018) Report: Nearly Half of ICOs Failed to Raise Funds Since Start of 2017. Coindesk,


Cryptocurrency Thursday

A New Catalog of Robots

The subtitle says it all “IEEE Spectrum is building the world’s largest, coolest robotics catalog” [1].

This on-line fan-pedia is a collection of real—not fictional—robots.

“a vast zoo of humanoids, drones, exoskeletons, quadrupeds, and other kinds of automatons”

The Beta is up and running, with a pretty good collection so far.

Robots: Your Guide to the World of Robotics (Beta)


Of course, this immediately made me think of how this dataset might be used for research.  You can’t dangle the “largest robot catalog” in front of me and not have me want to be able to dredge it!

So here are some suggestions.

First, it would be nice to have a open API of some kind, so we can build apps to do our own analyses and presentations of the data.

<<imagine a link here to dynamically retrieved image or record from robot db>>

If I had such an API that I could plug in to other software, I could imagine research projects, especially about robot design.  Machine learning could chew through the images and profiles, queries could extract interesting subsets for analysis, etc.

Some ideas for what might be done, off the top of my head.

One interesting question is how designers create ‘humanoid’ appearance. It would be interesting to look at the range and design dimensions that are used.

For that matter, it might be interesting to try to measure the “humunoid-ness” of robots. I.e., a robot might be indented to be humanoid but be pretty abstract, and it might be intended to do something specific, but also look humanoid.  And so on.  This would involve coming up  with an index of ‘humanoid-ness’.

In general, it would be interesting to cluster robots on a variety of dimensions, visual and on other features.  This might be correlated with other information about the use, success, and attractiveness of the robots.  It might also lead to insights into the design space, perhaps suggesting “unoccupied niches” to explore.

(I wouldn’t be surprised to find a ‘family tree’ of robot designs as well.)

The images in this collection might be really useful as stimuli in psychological studies of how robots are perceived.  There are an increasing number of such studies, and to date they have mostly picked robots they happen to have.  This would make it possible to try to sample the space of robots more systematically.

And finally, following from my recent thinking, I immediately notice just how white these robots seem to be.  But is this impression accurate or just an illusion?  This database could document important, if unintended, design trends.

Anyway, let’s all politely request access to the data, so we can muck about and learn all kinds of stuff.

  1. Erico Guizzo and Randi Klett, Calling All Robots: IEEE Spectrum is building the world’s largest, coolest robotics catalog, in IEEE Spectrum – Robotics. 2018.


Robot Wednesday CryptoTulip Candidate

This fall we see a new candidate for CryptoTulip of the Year:  EOS.IO from the annoyingly named

This is an interesting system, though rather hard to evaluate.  Sounds Tulip-y to me.

It’s big idea is to do dapps better than Ethereum.  That means higher performance and lower cost.  We’ll see.

The key “innovation” is “delegated” consensus that makes the ‘decentralized’ system is much more efficient by centralizing the consensus step [1].

blocks are produced in rounds of 126 (6 blocks each, times 21 producers). At the start of each round 21 unique block producers are chosen by preference of votes cast by token holders. The selected producers are scheduled in an order agreed upon by 15 or more producers.

This should be faster than Ethereum, assuming that it actually works as intended. I’m not sure how secure and fair this system is (voting is scarcely guaranteed to be either).

Another innovation is a charging scheme for the virtual machine that runs the dapps.  The dapp has to buy three resources, storage/bandwidth, cpu, and ram.  This mechanism manages the use of the computational resources of the nodes of the network, and maybe incentivizes participants to run node.  I think.  The whitepaper describes this as a “sender pays” model, contrasting to Ethereum which the user (receiver?) pays.

EOS features an explicit “constitution” that is supposed to apply to all nodes.  This is a human readable document that ‘splains the intent of the code and also “obligations among the users which cannot be entirely enforced by code”.  In an interesting bit of techno-theater, “Every transaction broadcast on the network must incorporate the hash of the constitution as part of the signature and thereby explicitly binds the signer to the contract.”  Right.

“All users are required to indicate acceptance of the new constitution as a condition of future transactions being processed.”

This is all pretty creative, and has attracted interest from developers who are dissatisfied with the performance of Ethereum and the cost of “gas”.  However, David Floyd  reports that all is not perfect in EOS land [2].

“That’s because, whereas ethereum dapps can be costly for the ones using them, EOS dapps can be costly for the teams deploying them.”

Worse, these resources are susceptible to price fluctuations and, it seems, price manipulation.  It may cost a lot to deploy a dapp, compared to Ethereum which accrues costs when run.  So the supposed cost savings are not automatic or simple.

So lets reckon the overall Tulipiness of this

The performance hit of “naïve Nakamotoan” consensus—security by massive redundancy—is replaced with a rotating sample of 21 of the nodes.  This may boil down to security by reputation, with the biggest players having the most influence on the sampling.  Or you might call it “continuously changing centralization”.  The single point of failure changes unpredictably with each decision round.

Execution of the dapps is “pay as you go”, but pay in advance.  Where Ethereum has a single resource (“gas”), EOS has three.   These differences certainly move around choke points on the execution of dapps, though the long term merits of the approach aren’t apparent.  You could also say that Ethereum’s single pain point (gas) is replaced with three different potential pain points.

EOS tackles some of the governance issues that have plagued Ethereum and others with an explicit “constitution”, plus rules that attempt to enforce it.  This would seem to be a “centralized” rule book with subjective interpretations of the code, and therefore a potential single point of failure.  (I.e., if the constitution is suborned, the whole system is compromised.)

I’ll note that none of this has been published in any kind of peer review study that I know of.  In particular, I see no evidence that the protocol has been analyzed by independent parties.  I’ll also note that the resource management scheme does not seem to have been simulated or studied.  (As a veteran of many resource control concepts inside operating systems, I assure you that intuition is not a good guide to how well they will actually work.)

For that matter, the boasts of performance and cost are noticeably undocumented.  How hard would it be to publish even some rough benchmarks of, say, throughput and latency?  If you are supposed to be better than Ethereum, shouldn’t there be at least case studies to prove it?

So, what we have here is a really complicated Tulip, which sets much stock in being “better” than Ethereum. is certainly different than Ethereum, but who knows if it is “better”, and in what cases?

Certainly, should be in the runnig for CryptoTulip of the Year for 2018.


  1., EOS.IO Technical White Paper v2. 2018.
  2. David Floyd (2018) RAM It All: Rising Costs Are Turning EOS Into a Crypto Coder’s Nightmare. Coindesk,


Cryptocurrency Thursday

Bat Inspired Sonar Navigation

The coolest thing about bats is their sonar navigation.

The recent crop of bat inspired robots has mimicked their amazing body form and flight.

A new study form Tel Aviv University reports efforts to mimic bat sonar [1].  In fact, the bat has a ‘mouth’ and two ‘ears’, and autonomously maps its environment. Cool!

“a fully autonomous bat-like terrestrial robot that relies on echolocation to move through a novel environment while mapping it solely based on sound. “ (p. 1)

OK, this specific study is not a bat shaped robot, nor does it even fly.  And the initial version does 2D mapping, not 3D.

More significantly, the research did not process the information the same way as bats do.  (I.e., the bio inspiration was only partial.)  Interestingly, their simple methods using significant computation roughly match the natural bat nervous system, though different methods.

“Altogether, we show how a rather simple signal processing approach allows to autonomously move and map a new environment based on acoustic information. “ (p. 7)

Obviously, it would be cool to port it to a real, flying batbot.  This will require considerable miniaturization and weight reduction of all the components.  The algorithms need to be extended to the 3D case, not to mention to run fast enough for real flight.

  1. Itamar Eliakim, Zahi Cohen, Gabor Kosa, and Yossi Yovel, A fully autonomous terrestrial bat-like acoustic robot. PLOS Computational Biology, 14 (9):e1006406, 2018.


Robot Wednesday

UAV Defensive Swarm

Speaking of swarms of robots….

With the proliferation of small UAVs, warding off intrusive drones is becoming an urgent need.  , a variety of interception methods are being developed, with varying levels of lethality and practicality.  From nets through ray guns, everything is on the table.  And, of course, avian defense forces are the coolest option.

This fall researchers at the University of Luxembourg report an approach that deploys a swarm of UAVs to “autonomously and collaboratively act as a defense swarm to deal with intruder” [1].

This is actually a fairly natural development, considering the long study of swarming algorithms dating back some thirty years [2].

The particular idea is based on the assumption that the intruder is programmed to avoid collisions, which is a common feature of UAVs (though scarcely guaranteed in a weapon system).

The essence of the idea is to surround and herd the intruder to a designated location, controlling it’s motion by placing defenders so as to ‘push’ the intruder to avoid collisions.  (Sort of like bumper cars.) The development includes algorithms to muster the defenders and dynamically coordinate toward this goal.  In principle, it could deal with multiple intruders or a swarm of intruders, as well.

(It should be noted that this study focuses on algorithm development in simulations.  The concepts are not field tested in actual flight.)

This idea has some good points.  It is non-destructive and safer for bystanders than some defenses (such as blazing away with a shotgun, Texas style).  The defenders are not expended, and don’t require expendable ammunition or nets.  And the defense is very flexible, capable of being recalled or redirected in case of other developments, unlike projectile or directed energy weapons.

This concept has distinct limits.  It assumes that the intruder is roughly equivalent is size and no faster than the defenders. As noted, the entire concept assumes that the intruder has a ‘normal’ collision avoidance behavior.  I don’t think this approach is intended to deal with an armed intruder, which might just attack the defenders.

And, of course, similar algorithms could be deployed by the intruders, to push aside the defenders and open paths to the target area.  In fact, there clearly will be algorithmic arms races, pitting suites of algorithmic moves and counter moves against each other.  This will be kind of a cool game, and it’s going to rapidly exceed the well explored territory of swarm behavior (and collision avoidance behavior).

Swarm versus swarm!

In the end, it is likely that clever algorithms will be defeated by speed, size, and numbers of intruders.  Basically, any given defense can be overcome by larger, faster, nastier intruders, or by large number of intruders, even relatively dumb intruders.  (And the best defense will still be to control territory to prevent the launch of attacking drones.)

  1. Matthias R. Brust, Grégoire Danoy, Pascal Bouvry, Dren Gashi, Himadri Pathak, and Mike P. Gonçalves, Defending against Intrusion of Malicious UAVs with Networked UAV Defense Swarms. arXiv, 2018.
  2. Craig W. Reynolds, Flocks, Herds, and Schools: A Distributed Behavioral Model. Computer Graphics, 21 (4):25-34, 1987.


Robot Wednesday Tuesday

Mobile 3D Printing Robots

Hey let’s combine two cool technologies to be even cooler together!

This summer researchers at Nanyang Technological University in Singapore demonstrated 3D printing with mobile robots [2].

3D printing is cool, but conventional printers are actually pretty simple beasts.  The program is a bunch of moves in the X, Y, and Z planes.  The printer itself is the geometric frame of reference and, barring malfunction, it’s all really straightforward geometrically.

This mobile system puts the printer on wheels, specifically on a mobile robot, so the printer can move to the place to print, and then print.  This makes it possible to build things that are much larger –larger even than the printer itself–and also to print in situ—handy for constructing a building. And, as Evan Ackerman notes “once you’ve decided to go that route, there’s no reason not to use multiple robots to speed things along.” [1]

This is really cool!

One of the tricky things, of course, is that the geometry is w-a-y harder, because you need to know where the robot is, at least relative to where you want the printed object to be.  With sub-millimeter tolerances!

The demonstration accomplishes this by a extension of the slicing algorithm to include planning for where one or more printers are for the printing.  This requires (multi) robot motion planning, potentially including nasty details about the topography of the site and topology of the printed output.  I.e., the site might have obstacles, and the structure itself becomes an obstacle as it is built.  And remember that the robot wants to get back home after printing, so there is now a potential to ‘print yourself into a corner’.

(It also occurs to me that wind and vibrations and birds perching and so on might perturb the robot during the print.  This is going to take some serious error detection and correction of the ‘current position of the printer’.)

As the researchers say, this approach is very flexible, and really opens up what might be done, especially at the scale of a building.  And a coordinated swarm of robots should be very efficient.

I’ll note that the swarm might ultimately be augmented by resupply robots to shuttle materials to the printers, track preparation robots to smooth and grade paths for the printers to travel and sit on, and maybe even helpers that act as support scaffolding during construction. On the last idea, imagine a robot that holds up a form to support printing of an arched doorway, and then extracts the support when the printing is self supporting.


I can envision this approach as a way to rapidly throw up a temporary structure, perhaps a fireproof windbreak or hearth.  (I’m imagining it extruding adobe-like material.)

Or, how about applying this to cake decoration – at scale!  For your next big party, how would you like a gazebo made of extruded sugar!  Edible architecture!

I also imagine a performance art work, in which a swarm of robots silently build a cage around a sleeping or immobile subject.   Or a swarm that builds a bower of love around a couple….

(And, of course, it won’t be long before these robots are hacked, to take them over for sabotage or just vandalism.  Construction robots hacked to write out messages in giant letters across a main highway….)

  1. Evan Ackerman, Mobile Robots Cooperate to 3D Print Large Structures, in IEEE Spectrum – Robotics. 2918.
  2. Xu Zhang, Mingyang Li, Jian Hui Lim, Yiwei Weng, Yi Wei Daniel Tay, Hung Pham, and Quang-Cuong Pham, Large-scale 3D printing by a team of mobile robots. Automation in Construction, 95:98-106, 2018/11/01/ 2018.


Robot Wednesday Monday

China leads the way in solar farms

The costs of solar and wind power is now lower than oil and coal. There will be a rapid replacement of obsolete carbon belching power plants with these new technologies.  This is already happening everywhere, but nowhere more spectacularly than in China.

Chris Baraniuk reports for the BBC that there are approximately 130 gigawatts (theoretical) of PV installed in China [1]. Development of these solar farms have driven up to now by a push-pull of government policies (i.e., subsidies for both production and consumption of PV panels).

The farms are located in remote areas.  There is abundant sunlight and space, but transmission costs and concomitant losses are high. The utilization factor is low (“staggeringly low”, says Baraniuk [1]) .

Subsidies for huge projects are drying up now, so attention may turn to roof tops and other “closer” generation.  But the pump is primed, and prices continue to plummet, so the subsidies may not be relevant soon.

This Chinese boom is also splashing around the world, inspiring and selling similar installations in other places, such as India and Canada.

And there is more than a little ‘Belt & Road’ swagger here.  I mean, stamping your brand on the Earth, visible from space.  Now that’s what I call a logo!

Image from google Maps. Datong Panda Power Plant solar farm in China—with attitude!


There are drawbacks to such large, remote installations, but that is a topic for future posts.

  1. Chris Baraniuk, How China’s giant solar farms are transforming world energy, in BBC News – Future. 2018.