Army of Bots
[avatar user=”malm” size=”small” align=”left” link=”file” /]
This post from Benedict Evans positions conversational UI in general as the next big service interface paradigm:
It’s certainly a hot topic which has already been highlighted in this blog in the context of both chatbots and Personal Digital Assistants. This week more evidence of the trend came with the Economist suggesting they represented the ‘next frontier’ in mobile services pegging its date of birth of the idea to the launch of Telegram less than a year ago in June 2015:
a new software ecosystem has started to emerge. Text-based services have been around since the dawn of internet time, but the birth of the bot economy can be dated to June last year, when Telegram, a messaging app with Russian origins and more than 100m users, launched a bot platform and a “bot store”. It now counts thousands of bots
It also came in the form of further revelations by TechCrunch of the nature of the chatbot support that Facebook are intending to introduce within Messenger and how it will allow brands to engage with users:
TechCrunch has reviewed a presentation sent by Facebook to some Messenger chatbot developers. It details how beyond just text chatbots will be able to respond with what it calls “Structured Messages.” These include a title, image, a description, a URL and calls to action such as visiting a website, viewing an e-commerce order or making a restaurant reservation.
Microsoft’s “big bet on bots” is also highlighted by The Verge notwithstanding the somewhat embarrassing recent debut performance of their Tay chatbot. Nadella is on record with a vision where everyone has their “own personal army of bots“.
Bloomberg suggested that rather than representing utopia or dystopia, Microsoft’s approach was in danger of becoming a Third Way of meh, a Clippy 2.0 for those that remember the infamous Paper Clip:
After spending some time with Tay, Microsoft’s new chatbot software, it was easy to see a third possibility: The AI future may simply be incredibly annoying.
One thing is clear. There is a massive software technology chasm between narrow closed domain systems built on short conversations using a retrieval-based model at one end of the spectrum and open domain systems built around unstructured arbitrarily long conversations using a generative model. The former is where many of the currently hyped chatbot propositions (assist.ai for instance) really lie. The latter is more the territory of Her (the film) and where the likes of Facebook and Microsoft are focussing their deep learning efforts. This excellent blog post outlines the technical landscape here with RNNs using LSTM (long short term memory) cells offering the most promise right now. However, the current state of the art results leave a lot to be desired judging by the results highlighted below in red which bring to mind John Searle’s famous Chinese Room thought experiment:
Artificial Intelligence and ML
- This is a great introduction to Scikit-Learn (sklearn) from the recent PyData conference:
- Sklearn is the go-to workhorse for many applied machine learning problems and helps explain the rising significance of PyData. It’s under the hood of Splunk’s machine learning app for instance:
- This insightful Artificial Intelligence market overview from the VC perspective highlights the sheer breadth of effort going on involving hundreds of companies operating within a whole host of different niches. AI is an foundational enabler technology not a separate industry vertical.
- Salesforce presumably hope to profit from their acquisition of AI startup MetaMind by eventually offering the holy grail of a value-add “executive AI” data mining SaaS proposition built upon their vast multi-tenant data collection. In theory at least, if they can build a compelling enough proposition, key business-oriented customers ought to be able to leverage it without needing any especial data analysis expertise. The description of the problem they’re attacking is familiar enough:
The problem for many companies right now, Block told Fortune, is that they’ve been collecting huge amounts of data, but they’ve failed to do much with it. He cited a recent meeting with the CEO of a large oil and gas company. “They have upstream and downstream devices collecting data but they’re not taking any action on it. They’re looking for ways to interpret that data and get to the next step,” he noted.
- The “father of the Singularity” Ray Kurzweil at Marvin Minsky’s remembrance service. Minsky has been put into cold storage and according to Wikipedia:
“Kurzweil believes that Minsky was cryonically preserved by Alcor and will be revived by 2045”
- It’s not clear if such an event would be pre or post Singularity. This NYT post baldly asserts the Singularity won’t happen in our lifetimes and that “the field of artificial intelligence has a long history of over-promising and under-delivering“.
Apps and Services
- McKinsey article on what it takes to really listen to what your customers are saying highlights the critical importance of having an open/transparent quantitative framework on which to rapidly assess the success/failure of app feature changes:
“In our experience, three core elements are critical for transforming a middling approach to customer-experience measurement into a value-creating system
1. Journey-centric feedback. The heart of journey-centric measurement is the organizing principle of measuring customer experience at the journey level, as opposed to looking only at transactional touchpoints or overall satisfaction
2. Hardwired backbone. investment in technology is necessary to support a superior customer-experience-measurement system. Numerous vendors provide a variety of functionalities and price points for these platforms to integrate the customer-experience-measurement strategy into day-to-day work routines.
3. Continuous-improvement mind-set. In our experience, there are two areas where establishing a continuous-improvement regimen matters most in achieving a superior customer-centric mind-set. The first is at the front line, with employees closing the loop with customers on direct feedback, then using that insight to change the way the process is designed or executed. … The second important area of impact is making feedback part of an approach to continuous improvement in service design. Product companies long have known that rapid iteration using customer insights is the way to get to a winning product, but service companies have rarely invested the same way in service design.”
The Internet of Things
- The Register on why Nest’s sudden shutdown of its Revolv home automation hub ought to serve as a wakeup call for the consumer IoT industry references a paper written by Nominet which calls for an open DNS for IoT devices because:
When a device is hard-coded to a specific service it is beholden to that service being available. If the company either goes bust or decides to stop running the service, then the product is rendered useless and the customer is left with a very expensive brick. Incidents like this can only make consumers wary and less inclined to buy IoT products in the future.
- The Verge suggests the Revolv shuttering is linked to ongoing problems at Nest. Either way, it is another classic example of the digital platform risk you incur when you choose proprietary technology and it requires us to question more diligently, who provides tech support for the IoT?
- In similar vein to the comments above on the burgeoning AI market, it doesn’t make sense to talk of a singular IOT market. This excellent explanatory post explains why and outlines on what is happening in IoT in 2016. The sheer density of the diagram illustrates the point of the post:
Robots, Drones and Cars
- The disturbing and ethically fraught story of the robot designer and his Scarlett Johansson-lookalike robot says a lot about mainstream male objectification of women and presages more hyperreality nightmares ahead as the technology advances to allow ever closer simulacra:
As AI advances and robotic technology grows cheaper and easier to create at home, other women may soon know what it feels like to have a stranger own and control a version of them.
- On a technical note, nearly 3/4 of the robot’s body was created using 3D-printing:
“A 3D-printed skeleton lies beneath Mark 1’s silicone skin, wrapping its mechanical and electronic parts. About 70 percent of its body was created using 3D printing technology.”
- It’s no coincidence that Johansson was chosen given her starring role in Hollywood dystopian AI classic Her. The film seems to have achieved a form of zeitgeist that every decade or so a handful of films do. It seems destined to become a Bladerunner for the Teenies.
“Robots make it easier to sexually objectify women. Bots modeled after beautiful women like Scarlett Johansson make it easier for men to avoid dealing with rejection.
- London Tesla showroom gives pride of place to a lovely red Tesla Model S keeping the place warm before the arrival of the Model 3:
- If you can’t be persuaded to join the pre-order crowd, maybe this Lego version will suffice:
- Lego aside, Tesla are no toy company for the automotive industry. This Quartz post suggests that we may be witnessing the beginning of the same sort of epochal transformation that afflicted Nokia at the hands of Apple driven by desire for a disruptive new product. In a curious iPhone-like echo, here’s the Tesla pre-order queue in Pasadena.
Quite aside from the political outfall, the trail of IT security lapses that led to the “largest corporate data heist in history” at Panamanian law firm Mossack Fonseca (MF) will reverberate around boardrooms everywhere. If it wasn’t already sufficiently obvious, it will surely be clear now that executive teams cannot regard IT security as a secondary concern in running their business or they mind find themselves without a business. WordFence suggested in an initial post that the entry point for the MF hack may have been an out of date WordPress plugin called Revolution Slider. Once in, the attackers could pull email server credentials from the main wp_config.php file courtesy of one of two other plugins (WP SMTP and ALO EasyMail) which store these credentials in the WordPress database in plain text. That would explain the theft of emails and database content. The documents were likely picked up by exploiting a known collection of vulnerabilities in the old version of Drupal MF were using to power their client portal. Many of their clients must be wishing they’d stuck to old-fashioned paper-based legals.
- In an important development, WordPress have enabled encryption for all WordPress-hosted sites courtesy of LetsEncrypt support. In a post last year, I documented the process and results of shifting this blog over to HTTPS. It is now supported by default but there are still a backlog of broken images to deal with before fully switching off HTTP support.
- This Truecaller exploit links IMEI with account identity information. The mobile service, which has 100 million users, was updated promptly:
Cheetah Mobile researchers told The Hacker News that they were able to retrieve personal data belonged to other users with the help of exploit code just by interacting with Truecaller’s servers.
“Overall, devices running the most recent versions of iOS and Android are comparably protected against offline attacks, when configured correctly by both the phone manufacturer and the end user. Older versions may be more vulnerable; one system could be cracked in less than 10 seconds. Additionally, configuration and software flaws by phone manufacturers may also compromise security of both Android and iOS devices.”
- The Vernee Apollo is an upcoming 128Gb storage, 6Gb RAM, 21 megapixel smartphone running Android on the decacore MediaTek Helio x20 processor. It’s as powerful as an entry-level laptop and will allegedly sell for less than $200:
- The Helio SoC built around the ARM Cortex A72 that is powering the Vernee Apollo is driving a wider revolution in the sub $200 Android device landscape in China that will make it even harder for premium brands to compete.
- The top jobs in 2025 and skills you need to get them. Technology, caregiving, social intelligence, lifelong learning and business acumen are all cited. Not explicitly coding ability though.
- In fact, learning to code won’t save your job according to this surprisingly political post in FastCompany that suggests that coders will not escape the technological unemployment they are inadvertently bringing to other jobs:
“Most of the technologies we’re currently developing replace or obsolesce far more employment opportunities than they create. Those that don’t—technologies that require ongoing human maintenance or participation in order to work—are not supported by venture capital for precisely this reason. They are considered unscalable because they demand more paid human employees as the business grows. … When technology increases productivity, a company has a new excuse to eliminate jobs and use the savings to reward its shareholders with dividends and stock buybacks. What would’ve been lost to wages is instead turned back into capital. So the middle class hollows out, and the only ones left making money are those depending on the passive returns from their investments.”
- How to reproduce the legendary Joy Division Unknown Pleasures image in R. And a version in Python implemented using numpy and matplotlib. I was going to look at developing my own version but was somewhat put off after reading that last post. So I just ran it instead and produced this output:
A man, a plan, a canal: Panama
- In a week where the governing elite has been rocked by the revelations found in the Mossack Fonseca Panama Papers, the Irish Times highlighted how the material shone a light on an shadowy world where in grotesque parody of the democratic system, everyone is treated equally on a truly global scale, a sort of SPECTRE of our times. Except in this case we don’t know who Number 1 is, the man with the plan:
The cache of 11.5 million records shows how a global industry of law firms and big banks sells financial secrecy to politicians, fraudsters and drug traffickers as well as billionaires, celebrities and sports stars.
- Quartz focussed on how the papers starkly reveal the corrupt roots of global extremism and the extent to which many countries are being systematically looted by their own leaders:
Imagine, for a moment, how much money Middle Eastern royals, dictators, and various crony capitalists have stashed away, laundered, or simply shunted around. Consider the present condition of the Middle East, and how many billions, if not trillions, of dollars were flushed down the toilet in pursuit of pleasure, acts of hedonism, vanity projects, cults of personality and grudge matches. All money that could have been investing in the future of a region that was once one of the most dynamic in the world.
- According to Thomas Piketty, offshore tax havens are an anachronism that must be more transparently governed and monitored or the global political system risks fragmentation at the hands of irate and indignant masses everywhere. With the rise of populism and anti-establishment politics everywhere, it seems that particular genie is already out of the bottle.
- The revelations thus far are mere entrée material. However, the exclusive nature of the arrangements made with media outlets means we may never get told the full detail of what’s in the Panama Papers. Even the likes of the Guardian who are unlikely to bite the hands that feed them:
The filtering of this Mossack Fonseca information by the corporate media follows a direct western governmental agenda. There is no mention at all of use of Mossack Fonseca by massive western corporations or western billionaires – the main customers. And the Guardian is quick to reassure that “much of the leaked material will remain private.” … Expect hits at Russia, Iran and Syria and some tiny “balancing” western country like Iceland. A superannuated UK peer or two will be sacrificed – someone already with dementia.
- Given this backdrop and crucially as highlighted in the Security section above how it seems to have been conducted, it is perhaps appropriate to think the “unthinkable” namely just how eminently possible it is to hack the democratic process and win an election. Bloomberg ran a revealing profile piece on Andrés Sepúlveda who claims he “rigged elections throughout Latin America for almost a decade“. It makes Watergate seem like child’s play. We may never know for sure if there wasn’t a similar political hand in glove going on with this leak.
- Long absorbing blog post on why Trumpism is a reincarnation of the “Jacksonians”, a group that go a long way back in US politics. The article explains why he is eventually destined to fail offering similar reasons to those expanded upon last week, namely that Trump’s schtick relies upon progressive shock tactics and:
“You cannot easily maintain a geometric progression when it comes to outrageousness.”
Culture and Society
- Why Google’s notorious April 1st prank this year backfired – laughter doesn’t scale:
Well, it looks like we pranked ourselves this year. 😟 Due to a bug, the MicDrop feature inadvertently caused more headaches than laughs. We’re truly sorry. The feature has been turned off. If you are still seeing it, please reload your Gmail page.
- A video on the origin of the avocado. The fruit which coevolved with giant ground sloths (functioning much like berries do with birds) was later adopted by humans:
- Silent Circle’s CEO Mike Janke on the various ways his current job resembles his previous life as a US Navy SEAL sniper.
- In a week when we have seen arguably some of the worse that human leadership has to offer us in the form of secret offshore bank accounts, it is perhaps worth ending with a reflection on the life of the last Native American Crow tribe war chief Joseph Medicine Crow, who died at 102 and left among many other treasures these words of wisdom:
“Plains Indian warfare was not about killing. It was about intelligence, leadership and honour”