Week 6

Deep Learning

The impact of Artificial Intelligence and Machine Learning on society are increasingly important preoccupations for this blog.  The palpable sense that these technologies have overhauled the more amorphous ‘Big Data’ in mindshare was underscored by a Medium post this week which compared their trajectory in Google Trends. Over the last year AI, Machine Learning and Deep Learning have all risen sharply while Big Data has fallen away.

Deep Learning certainly feels like it is taking centre stage with each remarkable breakthrough around the world inching us closer to AGI in stochastic gradient descent style.  Witness this week MIT’s Technology Review reported on Iranian research which demonstrated that the idiosyncrasies of human face recognition can be successfully replicated in a 7-layer deep neural network which ‘copies’ structures found through experimental observation of monkeys.  The hidden layers of the net detect and operate at the level of “face patches” to build the overall recognition.  It’s a development that provides implicit support to proponents of intelligence as an ’emergent’ property of brain hardware:

The process behind this work is almost as fascinating as the result. These guys have taken certain structures found in monkey brains, built synthetic system based on the structures and then found that the artificial behaviour matches the biological behaviour.

Meanwhile the Guardian once again effusively lionized Demis Hassabis, the founder of DeepMind and poster boy of Deep Learning at least in the UK.  It’s their third profile of him in 8 months underscoring the media mainstreaming of the technology. The article provides several interesting insights into DeepMind – apparently the company has a 100% retention record and the integration of multidisciplinary skillsets are a key aspect of its work culture:

Vital to the fabric of DeepMind are what he calls his “glue minds”: fellow polymaths who can sufficiently grasp myriad scientific areas to “find the join points and quickly identify where promising interdisciplinary connections might be, in a sort of left-field way.” Applying the right benchmarks, these glue people can then check in on working groups every few weeks and swiftly, flexibly, move around resources and engineers where required.

While Hassabis is clearly and openly driven by a desire and curiosity to combine ANI into something approaching AGI, there are plenty of people who have serious reservations and concerns about the ramifications of that sort of “progress” for humanity including one quoted at length in another Guardian article this week:

Expert Moshe Vardi told the American Association for the Advancement of Science (AAAS): “We are approaching a time when machines will be able to outperform humans at almost any task. … “I believe that society needs to confront this question before it is upon us: if machines are capable of doing almost any work humans can do, what will humans do?”

The tech giants will continue to press on regardless – the stakes are simply too great not to innovate with these technologies.   Even for Microsoft who just launched their Fetch! app to feed us their own take on “human-centric” AI:

Mobile Devices

  • Onto more prosaic tech.  VoLTE is a big deal for network operators and likely to be heavily pushed at MWC next week. Dean Bubley explains why it’s all a mirage because VoLTE is not the future of mobile voice:

VoLTE is the right answer, 7 years too late. It should have been part of LTE at launch on Day 1, designed-in and standardised during Year Minus-3. The industry forgot about it, and has been in panicky catch-up mode ever since, trying to shoehorn immature packet voice onto a complex new air-interface and core network. The focus has been on network QoS, standards and spectrum re-farming, rather than communications purpose, user-experience, analytics, devices or applications.

Any time there’s a way to quick launch the camera on a phone, I immediately learn it and make it a part of my workflow, because it really does make a difference when trying to get that quick picture.

  • The Nitro Duo is a “convergence” product designed to run both Android and Windows 10 in the same case with a 2 in 1 twist:

Nitro Duo is essentially two computers sharing a 5-inch touch display, ports, and peripherals. The first system is a Freescale quad core iMX-6 with 2GB of DDR2 Ram and boots a Droid OS on its own drive. The secondary system is an Intel quad core Atom with 4GB of DDR3 Ram and a 256GB SSD for its OS storage and applications. The Intel system boots a full x86 version of Windows 10.

Wearbles, IoT and Hardware

  • Fitbit have built a unicorn off the back of the “10000 steps” .  This fascinating post explains the surprising backstory behind that number and why it’s actually not really that useful an aid.  You really only need about 5000 steps a day to make a real impact on lowering your resting heart rate.  To achieve further reductions requires a focus on intensity rather than quantity:


  • It’s that time of year when many ready themselves for the annual trip to MWC in Barcelona. I haven’t been for a couple of years now and don’t miss it all that much – the energy and momentum seems to be draining out of mobile. Sounds like Dean Bubley (again) isn’t much of a fan either:

  • The existential question all Smart Ring products eventually face – what are they actually for?

  • The IoT Expo this week at Olympia, London was in many ways much more interesting and insightful than MWC.  The intimate scale and grass-roots enthusiasm of the participants made for a much more enjoyable occasion.  Aldebaran Systems’ ‘humanoid robot’ Pepper made its UK debut and proved a popular draw albeit one that only seemed to show a sad face in mirror mode for me:


  • A particular standout was SamLabs and their eponymous IoT kit, “lego for the Internet generation”.  I ended up buying a set and it was surprisingly easy to set up for a product that involved the unpromising combination of native PC software and a Bluetooth dongle:


Service Design

We demand that new technologies do not only what they promise, but what we imagine. We demand that they behave in a way that we can guess, without ever having used them before. That’s what lies behind the mysterious notion of user friendliness.

  • Why Twitter is ‘fundamentally broken’ because the more successful and therefore prominent you are on it, the more frequently you will be targeted for abuse:

  • How to Snapchat like a teen champion if you must, though it sounds like a lifestyle choice that is inherently unsustainable for anyone with any other interests or responsibilities.  Among other mind blowers from the 13-year old sister the insight is this one – it’s worth reflecting on what’s going on here..:

If you want to take a screenshot without your friend knowing, turn on airplane mode, take the screenshot, log out of the app immediately, turn off airplane mode, and then load the app back up.

Software as a Service

With the SharePoint 2016 release fast approaching, it’s becoming evident that the rise of the API and SaaS model of getting technology to do stuff you want will be far quicker and cheaper (and without the use of server rooms).

Security and Privacy

Canonical is keen to support the development of blockchain applications and their expansion from business cloud solutions to mobile consumer applications. Working with ConsenSys and BlockApps on building a platform for the consumption and creation of blockchain-based mobile apps is the realisation of the promise of Ubuntu as the reinvention of personal mobile computing

Python and Coding

9781409591511-lift-the-flap-coding  9781409599357-coding-scratch-new-2

good mocking requires a different mindset than good development. Development is about making things, while mocking is about faking things.

# line 1
license = 'https://opensource.org/licenses/MIT'
chosen_word = ''.join(filter(str.isalpha, __import__('random').choice(open('/usr/share/dict/words').readlines()).upper()))
guesses = set()
scaffold = '|======\n| |\n| {3} {0} {5}\n| {2}{1}{4}\n| {6} {7}\n| {8} {9}\n|'
man = list('OT-\\-//\\||')
guesses_left = 10

# line 2
while not all(letter in guesses for letter in chosen_word) and guesses_left: 
 _ = map(guesses.add, filter(str.isalpha, raw_input('%s(%s guesses left)\n%s\n%s:' % 
 scaffold.format(*(man[:10-guesses_left] + [' '] * guesses_left)),
 ' '.join(letter if letter in guesses else '_' for letter in chosen_word))).upper()))
 guesses_left = max((10 - len(guesses - set(chosen_word))), 0)

# line 3
print 'You', ['lose!\n' + scaffold.format(*man), 'win!'][bool(guesses_left)], '\nWord was', chosen_word

Gravitational Waves

  • The announcement of the discovery of gravitational waves by scientists working on the LIGO detector in the US this week was universally heralded as a major milestone in scientific understanding and the final piece of the jigsaw of General Relativity.  The New Yorker published a good deep dive on the long and winding road that led from dusty technical committee discussions going back to the 70’s through an obstacle course of funding meetings to the present day.  Several blogs suggested this video was the best explanation of the science behind gravitational waves and the LIGO detector:

  • The Independent meanwhile outlined why you should care about their discovery even if you can’t see any obvious practical benefit today.  These sort of breakthroughs tend to take years to filter through to universal application.

One of the primary data analysis “pipelines” which listens out for gravitational waves runs on Python. It looks like you noticed we used Matplotlib in the paper too! I am a big fan of Python and Matplotlib and I try to use it wherever possible. Traditionally in my part of the field we have used tools like Matlab for plotting, but gradually we’re moving over to open source alternatives as they support more and more of the features we’re used to.

Some thoughts on Leadership

for your most critical projects, the 10 to 30% of projects that absolutely change the game for you and separate your company from competitors, you have to break the process completely.  … If you follow process on those, it will take far too long, and you’ll lose your advantage

  • LinkedIn opinion piece on the five biggest trends for Millennials in 2016 namely:
    • Breaking news
    • Virtual Reality
    • Video Games
    • Streaming Services
    • Fintech
  • Streaming breaking news service delivered via a Google Cardboard-based virtual Monopoly game anyone?


  • But when it comes to criticising ideas, you want to give it five minutes and try to avoid shooting from the hip.  Ideas at first are very fragile and need intensive support and encouragement to get off the ground.


I’ve been clean a year and a half now, and I’m doing fine. I get plenty of work, I don’t miss invitations, and I’m no longer scared of my own thoughts.  …  I got a landline and I got more sleep. I look people in the eye. I eat food instead of photographing it and am not driving half a ton of metal into oncoming traffic while looking down at a tiny screen.


If man really sat back and thought about his impending termination, and his terrifying insignificance and aloneness in the cosmos, he would surely go mad, or succumb to a numbing sense of futility. Why, he might ask himself, should he bother to write a great symphony, or strive to make a living, or even to love another, when he is no more than a momentary microbe on a dust mote whirling through the unimaginable immensity of space? … This shattering recognition of our mortality is at the root of far more mental illness than I suspect even psychiatrists are aware.

Future Shock


  • Donald won’t go away and to the growing consternation of many liberal Americans, he may yet win the ultimate prize so it’s worth asking, what would a Trump presidency look like?   In short, expensive chaos would ensue if he actually tried to enact the more controversial measures he’s built his platform on (ie. the Mexican wall and repatriation, moving manufacturing jobs back to the US, barring Muslims from entering the country).  The only one that has even the slightest merit relates to manufacturing but even that would take years to realise and would constitute a Pyrrhic victory at best.