Andrew Sullivan’s long but essential read on how “technology almost killed me” in NYMag paints a vivid and insistent picture of a world in which there is no relief or let up from “the endless bombardment of news and gossip and images [that] has rendered us manic information addicts“. A world in which many are afflicted with distraction sickness. After reading the article, you can’t help but notice the symptoms next time you go out:
Just look around you — at the people crouched over their phones as they walk the streets, or drive their cars, or walk their dogs, or play with their children. Observe yourself in line for coffee, or in a quick work break, or driving, or even just going to the bathroom. Visit an airport and see the sea of craned necks and dead eyes. We have gone from looking up and around to constantly looking down.
Sullivan outlines his journey to at least understanding the problem if not entirely curing himself of it. The article has the added bonus of a legitimate inclusion of Henry David Thoreau’s classic quote from Walden, a personal favourite:
just as modern street lighting has slowly blotted the stars from the visible skies, so too have cars and planes and factories and flickering digital screens combined to rob us of a silence that was previously regarded as integral to the health of the human imagination. … This changes us. It slowly removes — without our even noticing it — the very spaces where we can gain a footing in our minds and souls that is not captive to constant pressures or desires or duties. And the smartphone has all but banished them. Thoreau issued his jeremiad against those pressures more than a century ago: “I went to the woods because I wished to live deliberately, to front only the essential facts of life, and see if I could not learn what it had to teach, and not, when I came to die, discover that I had not lived. I did not wish to live what was not life, living is so dear.”
Thoreau were he living today would probably eschew the wonders of a modern smartphone with its growing artificial intelligence capabilities. It’s very difficult for this blog given the nature of the topics covered to try and hold two opposing viewpoints in equal consideration. Indeed the erstwhile luddite view seems doomed to be overwhelmed by developments. However, it’s essential that doubts remain in a headlong collective rush towards the future that seems to have become more tangible than ever across every aspect of human endeavour largely powered by the evolving AI revolution. Silent space for contemplation devoid of digital assistance seems a distant memory.
Fortune Magazine primer on Deep Learning offers a good non-specialist starting point for understanding the transformative potential of the technology if you like this sort of thing:
“AI is the new electricity. Just as 100 years ago electricity transformed industry after industry, AI will now do the same.”
One of those quoted by Fortune is DeepMind CEO Demis Hassabis who spoke at the Royal Academy of Engineering last week in the UK simply entitled “Towards General Artificial Intelligence” in which he outlined a variety of ways the company’s Deep Reinforcement Learning technology could be applied in the real world. Also opting for a more generalist platform, noted AI philosopher Nick Bostrom will deliver a keynote at IPExpo in London this week on the prospects for Artificial Intelligence and humanity.
Bostrom has been a leading proponent of the view that we need to create supranational institutions for governing the progress of AI technology for the benefit of all. Google, Amazon and Facebook are among the organisations that have signed up to the Partnership AI .org initiative to attempt to do precisely that.
Deep Learning has had a significant impact on language translation advances in research at Google over the last couple of years culminating in the release of an updated Google Translate which is “nearly as good as a human translator“:
“The new method, called Google Machine Neural Translation [GMNT], cuts down errors by 80% compared to its current algorithm, and is nearly indistinguishable from human translation on standardized tests”
Wired pointed out how faster processing backed by specialised hardware is at the core of the progress:
“much of the speed is driven by Google’s tensor processing units, chips the company specifically built for AI. With TPUs, the same sentence that once took ten seconds to translate via this LSTM model now takes 300 milliseconds.”
Google released a technical paper outlining how GMNT works explaining some of the other software elements responsible for the speed up:
Our model consists of a deep LSTM network with 8 encoder and 8 decoder layers using attention and residual connections. To improve parallelism and therefore decrease training time, our attention mechanism connects the bottom layer of the decoder to the top layer of the encoder. To accelerate the final translation speed, we employ low-precision arithmetic during inference computations. To improve handling of rare words, we divide words into a limited set of common sub-word units (“wordpieces”) for both input and output.
Alison Carnwath, Chair of construction giant Land Securities on the mass scale technological unemployment that robotics will unleash upon the 2.3million strong UK construction sector presumably right after it has done with wreaking havoc on driving jobs:
“Five years ago I’d have smiled wryly if somebody had said to me that robots would be able to put up buildings in the City of London – I tell you we’re not that far off, and that has huge implications”
Music is likely to be disrupted too with Quartz previewing the first song ever written by an artificial intelligence. Here it is:
It stands in curious contrast with an attempt to restore the first computer generated music created courtesy of Alan Turing and his team in 1951.
No surprise then that Microsoft’s CEO Satya Nadella sees the company’s future as being all about Artificial Intelligence and is forming a 5000 person AI division to back up his vision.
Deep Learning remains hard to approach for many developers lacking prior academic background in the field. This recent post by Adrian Rosebrock explaining in some detail how to build a simple deep learning model using Python and Keras is therefore very welcome. An interesting startup called Bonsai goes even further and offers the prospect of genuine easy to access deep learning functionality for all. This Steven Levy article outlines their platform proposition which is able to create a deep learning version of the Atari game Breakout in 37 lines of code. According to Bonsai CEO Mark Hammond:
“with tools like Google’s TensorFlow, AI is now in the assembly language era, which makes it easier for the scientists building neural nets, but still limits the field to those who really understand how those nets work. His idea was to provide the equivalent of a compiler, to really open things up. … They have built a system with several components including Brain, a cloud-based system that constructs neural nets; a scripting language called Inkling; and Mastermind, an “integrated development environment” that gives programmers all the tools they need in one place.”
Another contender for democratising machine learning in the enterprise is Splunk. The latest v6.5 update to the popular enterprise data analytics tool includes the introduction of a range of wizards for applying a variety of techniques to your data. It looks like Python scikit-learn under the hood.
Quartz post on ordering food by chatbot highlights a startup called Conversable “which is building chatbot services for Whole Foods, Pizza Hut, and TGI Fridays.“
The Internet of Things
Really interesting consumer research on Amazon Echo contains a wealth of insight into what ordinary users are doing with their units:
Alexa now has 3000 skills with the list growing daily. Most of them are admittedly more gimmick than essential. It is nevertheless probably the most likely product to kickstart the consumer smart home market:
Relatedly, Wired on why your car is watching you and wants to sell you stuff. Underpinning this shift to the software-defined car of the future is a comprehensive API-encapsulated software platform. All the car OEMs seem to trying to build one. This article outlines Ford’s vision of what it means:
“Ford is reinventing itself as a platform company, whether that’s a cloud technology platform (which includes FinTech, autonomy, data management and connectivity as the four main pillars) or a physical vehicle platform on which it can more rapidly innovate new vehicle types and use cases.”
Mobile and Devices
Blackberry will make phone no more after reaching an almost classic rise and fall in fortunes. The company is now right back where they were ten years ago:
Meanwhile Apple continue to make a healthy margin from the iPhone. According to this analysis, the BOM cost of an iPhone7 is one third its retail cost to end consumers. Still, it does have an FPGA inside it presumably for local machine learning applications?
Android Authority on Allo’s most annoying feature namely App Preview Messaging:
While a cool idea on the surface, Allo’s ability to message people even if they don’t have the app (and then prompting them to download Allo) led some users’ friends and family to ask them to stop using it.
AWS have created a new monster P2 EC2 instance type offering up to 16 GPUs targeting heavy machine learning applications in the cloud:
These instances were designed to chew through tough, large-scale machine learning, deep learning, computational fluid dynamics (CFD), seismic analysis, molecular modeling, genomics, and computational finance workloads.
The flaw at the heart of the Gartner “bimodal IT” is that it is not an either-or to trade off responsiveness and reliability. You can have your cake and eat it and the best outfits do just that:
high performers achieve both higher levels of throughput and higher levels of stability than their low performing peers. They achieve this by implementing a high performance, lean culture and adopting continuous delivery practices across all of their products and services.
Excellent and thought-provoking article published on FirstRound outlining “the three infrastructure mistakes your company must not make“. The author, Avi Freeman, spent 10 years at Akamai and has seen over 100 startups scale their infrastructure so he is well worth listening to. The three mistakes being:
- Landing oneself in Cloud Jail.
- Getting sucked in by “hipster tools.”
- Failing to design for monitorability.
Cue a description of how to build a microservice using a hipsterish combination of nodejs and AWS Lambda.
ClassCentral published this excellent guide on the best programming and data science MOOCs available at present to newbies. Top of the pile is The University of Toronto’s “Learn To Program” course available on Coursera.
The Information ask what went wrong in India’s tech sector and provide two main reasons namely “hiring too much staff and offering deep discounts in the push for growth over quality“.
Echoes of Cormac McCarthy’s haunting Blood Meridian in this sobering Guardian Science article which outlines the evidence that lethal violence has long been part of our evolutionary history and that human beings are “natural born killers predisposed to murder”.
Tim Urban of WaitButWhy.com in another epic post explains how you need to understand the ultimate purpose and mission of Space X in order to understand why they are persisting in building the biggest rockets they can:
SpaceX is trying to make human life multi-planetary by building a self-sustaining, one-million-person civilization on Mars.
Astonishing data point on collapsing price of solar power which has the potential to enable this technology to reach an inflexion point by as soon as 2020 provided storage costs can be likewise disrupted. It makes one wonder why governments like the UK insist on making long-term energy supply deals:
These price improvements are not coming primarily from the price of panels dropping. They’re coming from reductions in the total cost to deploy solar, increases in solar capacity factor, ever-lower operating costs, and fierce competition to win bids in the solar industry. … The solar industry is learning faster than expected.
The Atlantic on the curious rise in importance of leisure activities like gaming as a life satisfaction marker and a somewhat inevitable bastion for humans facing technological unemployment:
The increasingly automatic nature of many jobs, coupled with the shortening work week, seem to be creating parallel tensions, which lead an increasing number of workers to look not to work but to leisure for satisfaction, meaning, expression … Today’s leisure occupations are no longer regarded merely as time fillers; the must, in the opinion of both social worker and psychiatrist, also perform to some extent as emotional buffers.
And is it any surprise when one considers the dysfunctional and pointless nature of all too many jobs where you don’t have to be stupid to work here but it helps. Aeon review a book on organisational culture called “The Stupidity Paradox” which looks like a ready made classic. It outlines the fundamental conundrum at the heart of the modern corporate environment:
Organisations hire smart people, but then positively encourage them not to use their intelligence. Asking difficult questions or thinking in greater depth is seen as a dangerous waste. Talented employees quickly learn to use their significant intellectual gifts only in the most narrow and myopic ways.
How to run a terrible meeting. The tl;dr seems to be to invite everyone, have no agenda and don’t agree on an outcome:
There is established and convincing literature that shows decision-making effectiveness sharply declines when the number of attendees grows into double-digit territory.
The Verge publish an interesting clarification on former editor-in-chief Chris Ziegler who appeared to have been moonlighting for Apple towards the end of his tenure at the tech news site.
Thankfully, one way or another, there’s not much longer to go in the US election race thank goodness. And a contrarian analyst who has correctly predicted 30 years of presidential outcomes correctly suggests that Donald Trump is heading for a win. Admittedly that was before the spectacle of the first televised head to head debate which was widely called in favour of Clinton. Still, it’s enough to make you pay attention to the morbid fascination of the New Yorker’s review of President Trump’s first term which many will hope remains firmly in the realm of counter-factual history. Especially this:
According to Bruce G. Blair, a research scholar at the Program on Science and Global Security, at Princeton, Trump encountered a U.S. nuclear-arms negotiator at a reception in 1990 and offered advice on how to cut a “terrific” deal with a Soviet counterpart. Trump told him to arrive late, stand over the Soviet negotiator, stick his finger in his chest, and say, “Fuck you!”