First, a clarification. I visit events such as the CogX events to be stimulated, to have new thoughts and to have the neurons in my brain fired in new ways. I go to learn, not to network. So I agonise over sessions to attend, and importantly the sessions I miss. Of which there is an overwhelming majority, as this event was running 5–7 conference tracks at any time, as well as a lot of other small stage events. By and large though it’s a weirdly monastic experience, surrounded by people, but very much alone in my head, to the point where I’m actually a little bit annoyed when somebody wants to talk to me! This then is the list of things that made me think.
- If there was one session that made attending the event worthwhile for me, it was Zavain Dar’s session on the New Radical Empiricism (NRE). His argument is that the traditional scientific method is based on certain rational assumptions — which are now challenged. In the classic method, you would hypothesise that the earth was round, find the right experiments to run, collect data and prove/ disprove your hypothesis. This runs into trouble when the computational models are too complex and / or changing too often — such as gene sequencing or macroeconomic data. Also this is not efficient when the range of options is vast and we don’t know what data might be relevant — e.g. curing cancer. The traditional methods may yield results, but it might take a lifetime of research and work to get there. What Dar calls the NRE is the opposite — a data driven view which allows machine learning to build hypotheses based on patterns it finds in the data. So in the NRE world, rather than starting with whether the earth is round, you would share a lot of relevant astronomical data and ask the machine to discover the shape of the world. This approach works best in areas where we have a data explosion such as genomics and computational biology. Or where there is plenty of data but is shackled by traditional hypotheses based methods, such as macroeconomics. An additional problem that NRE solves is where the problem space is simply to complex for human minds to compute — both the examples above are instances of this complexity. You may know that Radical Empiricism is by itself a construct from the late 19th century by William James — which eschews intuition and insists on physical experiences and concrete evidence to support cause and effect. Its worth noting that there are plenty of examples of environments where quantifiable data is not yet abundant, where experts still follow the traditional method driven by hypotheses. VC investing, ironically, is such an area!
- This also led to a discussion on Deeptech led by Azeem Azhar of Exponential View and panelists from Lux, Kindred Capital and Episode1 Ventures. Deeptech is defined from an investment perspective as companies and start ups who are building products which involve technical risk. Not using existing tech to solve new problems. Usually involving products and ideas which a few years ago would have to subsist on research grants and be housed by academic institutions.
- Jurgen Schmidhuber’s session on LSTM was another highlight. Schmidhuber’s PhD thesis on LSTM (Long Short Term Memory), in 1997 was a foundation of the AI advancement which was used by a number of technology products and subsequent development. Schmidhuber presented an excellent timeline of the evolution of AI in the past 20 years and ended with a long view where he explored the role of AI and ML in helping us reach resources that were not on earth but scattered across the solar system, the galaxy and beyond. And how we might perceive today’s technology and advancement in a few thousand years.
- One of Schmidhuber’s other points was around curiosity driven learning. Mimicking the way an infant learns, by exploring his or her universe. This is the idea that a machine can learn through observation and curiosity, about it’s environments.
- Joshua Gans, the author of Prediction Machines, and professor of Economics and Tech Innovation, talked about AI doing to prediction what computers did to arithmetic. Essentially they dramatically reduced the cost of complex arithmetical operations. AI does the same for prediction or inference. Which is essentially making deductions about the unknown based on the known. And bringing down the cost of prediction has a massive impact on decision making because that’s what we’re doing 80% of the time, at work, as managers.
- Moya Green, the CEO of Royal Mail talked about the transformation that Royal Mail went through — including an increase in technology team size from 60 to over 550 people. She also made the comment that most managers still under-appreciate the value of tech, and overestimate their organisations capability to change, and absorb new tech.
- Deep Nishar of Softbank used an excellent illustrative example of how AI is being used to provide personalised cover art for albums by digital streaming and media providers, based on users choices and preferences.
- Jim Mellon, long time investor and current proselytiser of life-extending tech suggested that Genomics would be a bigger breakthrough than semiconductors. He was joined by the chief data officer for Zymergen, which works on bio-manufactured products, based on platforms which work with microbial and genetic information.
- A very good data ethics panel pondered the appropriate metaphors for data. We’ve all heard the phrase data is the new oil. Yet that may be an inadequate descriptor. Experts on the panel posited metaphors such as ‘hazardous material’, ‘environment’, ’social good’ etc. because each of these definitions are useful in understanding how we should treat data. Traditional property based definitions are limited and it was mentioned that US history has plenty of examples of trying to correct social injustice via the property route (reservations for native Americans), which have not worked out. Hence we need these alternative metaphors. For example, the after-effects of data use is often misunderstood, and sometimes it needs to be quarantined or even destroyed, like hazardous material, according to Ravi Naik of ITN Solicitors.
- Michael Veale of the UCL suggested that ancient Greeks used to make engineers sleep under the bridges they built. This principle of responsibility for data products needs to be adopted for some of the complex products being built today by data engineers. Data use is very hard to control today, so rather than try and control it’s capture and exploitation, the focus perhaps should be on accountability and responsibility.
- Stephanie Hare made the excellent point that biometric data can’t be reset. You can reset your password or change your email, phone number, or even get a completely new ID. But you can’t get new biometrics (yet). This apparent permanence of of biometrics should give us pause to think even harder about how we collect and use it for identification, for example in the Aadhaar cards in India.
- Because of the inherently global flows of data and the internet, the environmental model is a good metaphor as well. Data is a shared resource. The lines of ownership are not always clear. Who owns the data generated by you driving a hired car on a work trip? You? Your employer? The car company? The transport system? Clearly a more collective approach is needed and much like social goods, such as the environment, these models need to validate the shared ownership of data and it’s joint stewardship by all players in the ecosystem.
- Stephanie Hare, who is French Historian by education provided the chilling example of how the original use vs ultimate use of data can have disastrous consequences. France had a very sophisticated census system and for reasons to do with it’s muslim immigrants from North Africa captured the religion of census correspondents. Yet, this information was used to round up all the jewish population and hand them over to the Nazis because that’s what the regime at the time felt justified in doing.
- On a much more current and hopeful note, I saw some great presentations by companies like Mapillary and SenSat, and Teralytics which focus on mapping cities with new cognitive tools. Especially for cities which are of less interest to tech giants, and using crowdsourced information and data, which may include mobile phone and wifi usage, or street level photographs all used with permission, for example.
- At a broader level, the smart cities discussions, strongly represented by London (Theo Blackwell) and TFL (Lauren Sager Weinstein) shows the transition from connected to smart is an important one. Very good examples by TFL on using permission based wifi tracking at platforms to give Line Managers for each of the tube lines much more sophisticated data on the movement of people, to make decisions about trains, schedules and crowd management, over and above the traditional ways which include CCTV footage or human observation on platforms.
- At a policy level, a point made by Rajiv Misra, CEO of Softbank Investment Advisors (aka the Vision Fund) is that while Europe leads in a lot of the academic and scientific work being done in AI, it lags in the commercial value derived by AI, notably to China and the US. A point echoed by the House of Lords report on AI which talks about the investments and commitment needed to sustain the lead the UK enjoys in AI, currently. Schmidhuber’s very specific solution was to mimic the Chinese model — i.e. identify a city and create an investment fund of $2bn to put into AI.
- I also sat through a few sessions on Chatbots and my takeaway is that chatbots are largely very much in the world of hype machines. There is very little ‘intelligence’ that it currently delivers. Most platforms rely on capturing all possible utterances and coding them into the responses. Even NLP is still at a very basic stage. This makes chatbots basically a design innovation — where instead of finding information yourself, you have a ‘single window’ through which to request all sorts of information. Perhaps its a good thing that the design challenges are getting fixed early, so that when intelligence does arrive, we won’t stumble around trying to design it.
- Within the current bot landscape, one useful model that I heard is ‘Treat a bot like a new intern that doesn’t know much’ and let it have a similar personality so that it provides responses that are appropriate and also sets expectations accordingly. It might just start with a ‘hello, I’m new so bear with me if I don’t have all the answers’, for example.
- Dr Julia Short, who has built Spot — a chatbot to handle workplace harassment provided a very interesting insight about the style of questions such a bot might ask. A police person’s questions on the one hand are all about capturing in detail exactly what happened and making sure that the respondent is clear and lucid about events, incidents, and the detail. A therapists questions and line of discussion on the other hand is all about helping a victim get over some the details and get on with their lives. This suggests that you need to be clear whether your bot is an extension of the law enforcement or a counselling body. It also suggests that you might want to do the former before the latter.
- A really important question that will not leave us is: what do we do if the data is biased? If we are conscious of certain biases which are to do with gender, race or age, then we can guard against them either at the data level or at the algorithmic level, but we also need to be able to detect biases. For example, the example which I’ve now read in a few places of how the leniency of sentences handed out by judges in juvenile courts in the US vary inversely with the time since the last meal of the judge.
Clearly all of this really represents under 20% of the great discussions over the 2 days. Please do add your own comments, takeaways and thoughts.
Hail Mary (Meeker)
On the 29th of May, Mary Meeker released her annual compendium of the digital state of the world – the KPCB Internet Trends. For those who may not remember, Mary Meeker was a veteran who survived the dot.com crash and also the financial crisis of 2008, as the head of tech research for Morgan Stanley. She was named as among the 10 smartest people in Tech. She now serves as a partner at KPCB (Kleiner Perkins Caufield and Byers) and has been publishing her annual opus for a few years now.
The problem is that when you’re Mary Meeker, you can get away by putting out a deck with 294 slides. For us mere mortals, reading and absorbing this encyclopaedia of information is a challenge by itself. Every year I get this and carefully save the deck to read in detail and of course, it never happens. So this year, with the benefit of a relatively free weekend, I thought I would do a first pass and pull out some of the most interesting things that I found in the report. So here are my top 10 interesting things to take away from the Mary Meeker report – some of them confirm what we know, while others are what we didn’t know, or are truly counter-intuitive.
What I knew or suspected.
1. The devices story mobile device shipments growth has shrunk to zero. This confirms what we’ve known for a while – device evolution has stalled since Steve Jobs. And since Samsung, the largest manufacturer has a ‘follow Apple’ strategy. Will we see a new device redefine growth or will the we see a decline in shipment numbers next year? HMD – are you watching? (Slide 6)
2. The decline in desktop use despite overall growth. While mobile internet growth is expected, it’s the ‘other devices’ that is interesting. This will presumably include netbooks, etc. but also smart things. I expect in future this category will be broken out to reflect the detail on Internet of Things. (Slide 11)
3. The privacy paradox will be one to watch – after all data is how every single provider improves their services, while keeping prices low, which leads to user spending more time and sharing more data. Versus the regulators needs to protect consumers and protect data use. This will be a key axis of debate going forward and will determine the balance between innovation and protection. Unfortunately Meeker’s slides don’t carry too much insight on this by way of data. (Slides 31-36)
What I didn’t know (I’m intentionally using the singular, as you may well be aware of this)
1. While we’re aware that big tech now dominates the market cap list, what should worry the rest of the pack is how they dominate the R&D spending list, which points to a continuation of their dominance at the top. The top 15 R&D investors list is dominated by 6 technology firms, with 2 each from automotive, petroleum, telcos, Pharma), with GE as the only conglomerate. The top 5 in the list are Amazon, Alphabet, Intel, Apple, and Microsoft. Also, tech firms report the highest growth in R&D, with 9% CAGR and 18% YoY growth. (Slides 40-41)
2. We know that image recognition is an area where AI has now passed the human levels of accuracy leading to all kind of applications across scan analysis in healthcare, and more controversial applications such as face recognition. Now, voice-based natural language recognition is another areas as also demonstrated recently by Google. This should drive a revolution in customer contact centres and in human-computer interfaces in general. (Slide 25)
3. The extent to whichAmazon & Google are getting to dominate the enterprise AI race. To be honest, we know instinctively that the AI race will be one by players with the largest data stockpile. But the range of services being offered for enterprise customers is still an eye-opener. We’ve just started playing around with Google’s Dialogflow, but they also have Tensor (cloud-based H/w), the recently announced AutoML (machine learning), and Vision API (Image recognition), while Amazon has AWS based tools such as Rekognition (image recognition), Comprehend (NLP), Sagemaker (ML framework), and of course their AWS GPU clusters. (Slide 198 – 200)
4. The growth of Fortnight and Twitch on the gaming front – pushes forward what we saw with Pokemon Go. The sweet spot between the hardcore platform based gamers and the casual gamers and kids where millions of people get just a little bit more involved about game, that does not need a special platform – is the story behind Fortnight (Slide 24)
What I didn’t expect
1. The highest increase in spending in enterprise IT is in networking equipment. This is a surprise. I haven’t found the data on this yet, and while the 2nd and 3rd place results don’t surprise me, (AI and hyper-converged infrastructure), my curiosity is definitely piqued by why companies are spending more on networking equipment – connecting to cloud environments from the enterprise perhaps? More connected devices and environments?
2. I’m seeing a lot more confirmation of the models of lifelong learning. This is repeated by Meeker, but her really interesting insight is around how much more learning freelancers invest in compared to their presumably complacent employee counterparts. Perhaps unsurprisingly the top courses sought include AI & related subjects, cryptocurrency, maths and English. (Slides 236 and 233)
3. Meeker makes a great point but Slack and dropbox and I wouldn’t have picked these 2 companies as the flagbearers of consumer-grade technology in the enterprise. But clearly, they are among the most penetrated consumer style tools in the corporate environment. (Slides 264-268)
Meeker has a big section on the Job market, on-demand jobs and future jobs. She also makes the same point others have made about how all technologies so far has created net new jobs. While I agree with this backlog, history is not always the best predictor of the future. And the fact that there will be net new jobs tends to gloss over the significant short-term and geographical disruption in livelihoods that is likely to occur. Think Detroit or Sheffield. There may be more automotive and steel manufacturing jobs today than in 1980 but they are in China, not in Detroit or Sheffield. And so of not much solace to the unemployed factory worker and his / her family in these towns. This may well be the story of AI – but potentially at a larger scale and possibly in a shorter time frame. (See slides 147-163).
There are also useful slides on the gig economy and on-demand jobs now being a scaled phenomenon. (Slides 164-175)
There are also entire sections on China, Immigration and Advertising – which I’ve not delved into as they are currently of less interest to me personally. The E-commerce section also didn’t have anything that jumped out at me as noteworthy. Happy to be corrected!
Seven for 7: Alexa sends the wrong message; does GDPR take us backwards? Uber crash – design flaw; future gazing with Michio Kaku; AI Winners; Ocado transformation and Energy Industry Updates.
(1) Amazon Echo: message in a bottle
The technology story of the week is undoubtedly the one about Amazon Echo and the message it inadvertently sent. ICMYI, a couple in Oregon had a call from an acquaintance to say that Alexa had sent them a recording of a private conversation of the couple, without their permission, or even knowledge. Amazon’s explanation is that this is the rare combination scenarios where a normal conversation between the couple somehow triggered all the keywords and responses that made Alexa record, validate and send the conversation to the acquaintance. This feels like the equivalent of the money, typewriter and Shakespeare problem, only, it’s not an infinite amount of time.
Here’s Amazon’s explanation: https://www.recode.net/2018/5/24/17391480/amazon-alexa-woman-secret-recording-echo-explanation
(2) GDPR – impact on marketing and innovation.
I’m sure you’ve all received hundreds of emails in the past week exhorting you to stay in touch and re-sign up for all the emails you’ve been getting from people you didn’t know were sending you emails. But now that the moment has come, how will marketing work in a GDPR world? In one way this will take marketing backwards – as there is now a ban on algorithmic decision making based on behavioural data. It’s a moot point whether advertising falls into this category but companies may want to play it safe and in any case, the confusion will create a speed breaker in the short term. We may now be back in the world where if you’re watching or reading about champions league football you will see a beer ad irrespective of who you are. Not just marketing – a lot of innovation will also come under fire – both because of safety first practices, but also because some organisations will use GDPR as a shield for enabling innovation stifling practices, as highlighted by John Battelle of NewCo Shift. He argues that the regulation favours ‘at scale first parties’ – large tech platforms that provide you with a direct service such as Netflix, Facebook, or Uber – where users are likely to still give consent for data use more readily than to smaller, upcoming or relatively new and unproven services.
Dipayan Ghosh in the HBR – GDPR & advertising: https://hbr.org/2018/05/how-gdpr-will-transform-digital-marketing
John Battelle on GDPR & Innovation: https://shift.newco.co/how-gdpr-kills-the-innovation-economy-844570b70a7a
(3) Driverless / Uber/ Analysis
The analysis of Uber’s recent driverless crash has now thrown some light on what went wrong. And the answers aren’t great for Uber. In a nutshell, the problem is design and not malfunction. Which means that all the components did exactly what they were designed to do. Nothing performed differently and no components were at fault for failing to do their job. But as a collective, the design itself was flawed. The car had 6 seconds and 378 feet of distance to do something about the pedestrian crossing the street with her cycle. But it was confused about what the object was. The human in the car only engaged the steering 1 second before the crash and started breaking 1 second after the collision. The car was not designed to warn the human driver about any possible threats. A lot of the inbuilt safety systems in the Volvo vehicle including tracking driver alertness, emergency braking and collision avoidance, were disabled in the autonomous mode. In a nutshell, the responsibility lies with Uber’s design of autonomous cars. Uber has stopped testing in Arizona but has now started exploring flying taxis. Not a project that might fill you with confidence!
Uber crash analysis: https://sf.curbed.com/2018/5/25/17395196/uber-report-preliminary-arizona-crash-fatal
(4) A glimpse of the future: Michio Kaku & Jack Ma
The robotics industry will take over the automobile industry. Your car will become a robot – you will argue with it. Then your brain will also be robotised and brain net will allow emotions and feelings to be uploaded. You will be able to mentally communicate with things around you. Biotech allows us to create skin, bone, cartilage and organs. Alcoholics may be able to replace their livers with artificial ones. You may be able to scan store goods with a contact lens and see the profit margin on goods. The first 7 mins of this video tells you all of this through the eyes and experience of futurist Michio Kaku. Jack Ma (14 mins in) also talks about trusting the next generation. And how we are transitioning from the industrial era where we made people behave like machines, to a world where we are making machines behave like people. Believe the future before you see it, to be a leader, according to Ma.
(5) Who’s Winning The AI Game?
With the whole world hurtling towards an AI future, this piece looks at who exactly wins the AI game – across 7 different layers. It won’t surprise you to know that China is making amazing gains as a nation – their face recognition can pick out a wanted man in a crowd of 50,000. But it might surprise you to note that Nvidia’s stock is up 1500% in the past 2 years on the back of the success of their GPU chips. Meanwhile, Google is giving away Tensorflow free. All this points to a $3.9 tn market for enterprise AI in 2022. Are you ready for the challenge?
Who wins AI, across 7 layers
(6) Ocado – digitally transformed.
When Ocado launched in 2000, it was on the heels of Webvan, a category of providers who set up to focus on eCommerce fulfilment, as an arm of Waitrose. Cut to 2018, and Ocado is a story of successful digital transformation. Ocado is today a provider of robotic technology for warehouse automation. Having become profitable in 2014, it now has a valuation of $5.3 bn and is set to become a part of the FTSE 100.
(7) Understanding statistics: What medical research reports miss
When a drug is tested and the outcome suggests a 5% chance of a possible side effect, this does not mean that you have a 5% chance of being impacted if the drug is administered to you. It means there is a 5% chance that you will have the condition which leads to you having a 100% likelihood of being impacted. This is a subtle but very important distinction in how we interpret the data. But continuing down this stream of thought, it points to the lack of personalisation of medicine, not just the misinterpretation of data.
(1) Doing Agile at Scale
(2) Artificial Intelligence At Scale – for non-technology firms
(3) The Ring of Success:
What makes a new product successful? I met Jamie, the founder of Ring a couple of years ago in London and was struck by his directness and commitment. He even appears in his company’s ads. Ring.com was recently acquired by Amazon for $1bn. Here one of the backers of Ring talks about the factors which made Ring a success. In a nutshell, the list includes (1) the qualities of the founder, (2) execution focus and excellence, (3) continuous improvements, (4) having a single purpose, (5) pricing and customer value. (6) integration of hardware and software. (7) clarity about the role of the brand.
(4) Blockchain and ICO redux:
Do you know your Ethereum from your Eos or your MIATA from your Monero? This piece from the MIT Tech Review will sort you out. And for those of you who are still struggling to understand what exactly blockchain is, here’s a good primer. Of course, you could always go look at my earlier blog post on everything blockchain.
(5) X and Z – The Millennial Sandwich
X & Z: Or the millennial sandwich. All the talk in the digital revolves around millennial, but there is a generation on either side. The generation X – followed the baby boomers, and it turns out they have a better handle on traditional leadership values than millennial. This article talks about Generation X at work.
(6) Big tech validates Industry 4.0
This week, the large tech players disclosed significant earnings, beating expectations and seeing share prices surge. In a way it’s a validation of the industry 4.0 model – the abundance of capital, data, and infrastructure will enable businesses to create exponential value, despite the challenges of regulation, data stewardship issues and other problems. Amazon still has headroom because when push comes to shove, Amazon Prime, which includes all you can consume music and movies can probably increase prices still more.
(7) Defending Democracy
The US elections meets the technology arms race – this article presents experiences from a hacking bootcamp., run for the teams who manage elections. While the details are interesting, there is a larger story here – more than influencing the elections either way, the greater harm this kind of election hacking wreaks is in its ability to shake people’s faith in democracy. As always, there’s no other answer than being prepared, but that’s easier said than done!
(image credit: Pixabay)
(1) The One Thing You Should Read about AI this week:
- Industries where the number of use cases are the highest, include (1) Insurance, (2) Banking, (3) Retail, and (4) Automotive & Assembly.
- Functions with the highest number of use cases include (1) Supply Chain management and manufacturing, and (2) Marketing and Sales
- Specific domains where the impact might be the highest include (1) customer service & management (2) risk modelling (3) predictive service / intervention (4) workforce productivity and efficiency (5) analytics-driven hiring and retention, and (6) yield optimisation.
- The highest absolute impact of AI is to be found in Retail, but Travel and transport & logistics can extract the highest incremental value over other analytics techniques.
- Image data is the highest value, after structured and time series data, and ahead of text.
- Challenges and limitations: (1) labelling training data (2) obtaining large enough data sets (3) explaining the outcomes and decisions in clear enough terms – e.g. for product classification or regulatory (4) transferring of findings to adjacent use cases, and (5) risk of bias in data/ algorithms
(2) Data: Facebook is a misguided amateur compared to Palantir
(3) Design: Welcome to Generative 3D Design
(4) eCommerce and Retail – change of guard, and disruption for the economy
(5) Asset-light business models
(6) Alexa Fashion – a glimpse of the future
(7) Battery Wars
Conversational Systems aka chatbots are starting to become mainstream – here’s why you should stay ahead of the game:
The shape-shifting of the yin-yang between humans and technology is one of the hallmarks of digital technologies, but it is perhaps most pronounced and exploit in the area of Conversational Systems. But to truly appreciate conversational systems, we need to go back a few steps.
For the longest part of the evolution of information technology, the technology has been the unwieldy and intransigent partner requiring humans to contort in order to fit. Mainframe and ERP system were largely built to defend the single version of truth and cared little for the experience. Cue hours of training, anti-intuitive interfaces, clunky experiences, and flows designed by analysts, not designers. Most of us have lived through many ages of this type of IT will have experienced this first hand. If these systems were buildings they would be warehouses and fortresses, not homes or palaces. Too bad if you didn’t like it. What’s ‘like’ got to do with it! (As Tina Turner might have sung!)
Digital technology started to change this model. Because of its roots in consumer technology rather than enterprise, design and adoption were very much the problem of the providers. This story weaves it’s way through the emergence of web, social media and culminates with the launch of the iPhone. There is no doubt – the iPhone made technology sexy. To extend the oft-quoted NASA analogy, it was the rocket in your pocket! With the emergence of the app environment and broadband internet, which was key to Web 2.0, it suddenly introduced a whole new ingredient into the technology cookbook – emotion! Steve Jobs didn’t just want technology to be likable, he wanted it to be lickable.
The balance between humans and technology has since been redressed significantly – apps and websites focus on intuitiveness, and molding the process around the user. It means that to deal with a bank, you don’t have to follow the banks’ convenience, for time and place, and follow their processes of filling a lifetime’s worth of forms. Instead, banks work hard to make it work for you. And you want it 24/7, on the train, at bus stops, in the elevator and before you get out from under your blanket in the morning. And the banks have to make that happen. The mouse has given way to the finger. Humans and technology are ever closer. This was almost a meeting of equals.
But now the pendulum is swinging the other way. Technology wants to make it even easier for humans. Why should you learn to use an iPhone or figure out how to install and manage an app? You should just ask for it the way you would, in any other situation, and technology should do your bidding. Instead of downloading, installing and launching an app, you should simply ask the question in plain English (or a language of your choice) and the bank should respond. Welcome to the world of Conversational Systems. Ask Siri, ask Alexa, or Cortana, or Google or Bixby. But wait, we’ve gotten ahead of ourselves again.
The starting point for conversational systems is a chatbot. And a chatbot is an intelligent tool. Yes, we’re talking about AI and machine learning. Conversational systems are one of the early and universal applications of artificial intelligence. But it’s not so simple as just calling it AI. There are actually multiple points of intelligence in a conversational system. How does a chatbot work? Well for a user, you just type as though you were chatting with a human and you get human-like responses back in spoken language. Your experience is no different from talking on WhatsApp or Facebook Messenger for example, with another person. The point here is that you are able to ‘speak’ in a way that you are used to and the technology bend itself around you – your words, expressions, context, dialect, questions and even your mistakes.
Let’s look at that in a little more detail. This picture from Gartner does an excellent job of describing what goes into a chatbot:
The user interface is supported by a language processing and response generation engine. This means that the system needs to understand the users’ language. And it needs to generate responses that linguistically match the language of the user, and often the be cognizant of the mood. There are language engines like Microsoft’s LUIS, or Google’s language processing tool.
Behind this, the system needs to understand the user’s intent. Is this person trying to pay a bill? Change a password? Make a complaint? Ask a question? And to be able to qualify the question or issue, understand the urgency, etc. The third key area of intelligence is the contextual awareness. A customer talking to an insurance company in a flood-hit area has a fundamentally different context from a new prospect, though they may be asking the same question ‘does this policy cover xxx’. And of course, the context needs to be maintained through the conversation. An area which Amazon Alexa is just about fixing now. So when you say ‘Alexa who was the last president of the US’ and Alexa says ‘Barack Obama’ and you say ‘how tall is he?’ – Alexa doesn’t understand who ‘he’ is, because it hasn’t retained the context of the conversation.
And finally, the system needs to connect to a load of other systems to extract or enter data. And needless to say, when something goes wrong, it needs to ‘fail gracefully’: such as “Hmm… I don’t seem to know the answer to that. Let me check…” rather than “incorrect command” or “error, file not found”. These components are the building blocks of any conversational system. Just as with any AI application, we also need the data to train the chatbot, or allow it to learn ‘on the job’. One of the challenges in the latter approach is that the chatbot is prone to the biases of the data and real-time data may well have biases, as Microsoft discovered, with a Twitter-based chatbot.
We believe that chatbots are individually modular and very narrow in scope. You need to think of a network of chatbots, each doing a very small and focused task. One chatbot may just focus on verifying the customer’s information and authenticating her. Another may just do password changes. Although as far as the user is concerned, they may not know they’re communicating with many bots. The network of bots, therefore, acts as a single entity. We can even have humans and bots working in the same network with customers moving seamlessly between bots and human interactions depending on the state of the conversation. In fact, triaging the initial conversation and deciding whether a human or a bot needs to address the issue is also something a bot can be trained to do. My colleagues have built demos for bots which can walk a utility customer through a meter reading submission, for example, and also generate a bill for the customer.
Bots are by themselves individual micro-apps which are trained to perform certain tasks. You can have a meeting room bot which just helps you find and book the best available meeting room for your next meeting. Or a personal assistant bot that just manages your calendar, such as x.ai. We are building a number of these for our clients. Bots are excellent at handling multi-modal complexity – for example when the source of complexity is that there are many sources of information. The most classic case is 5 people trying to figure out the best time to meet, based on their calendars. As you well know, this is a repetitive, cyclical, time-consuming and often frustrating exercise, with dozens of emails and messages being exchanged. This is the kind of thing a bot can do very well, i.e. identify (say) the 3 best slots that fit everybody’s criteria on their calendars, keeping in mind travel and distances. Chatbots are just a special kind of bot that can also accept commands, and generate responses in natural language. Another kind of bot is a mailbot which can read an inbound email, contextualise it, and generate a response while capturing the relevant information in a data store. In our labs we have examples of mailbots which can respond to customers looking to change their address, for example.
Coming back to chatbots, if you also add a voice i.e. a speech to text engine to the interface, you get an Alexa or Siri kind of experience. Note that now we’re adding yet more intelligence that needs to recognise spoken words, often against background noises, and with a range of accents (yes, including Scottish ones). Of course, when it’s on the phone, there are many additional cues to the context of the user. The golden mean is in the space between recognising context and making appropriate suggestions, without making the user feel that their privacy is being compromised. Quite apart from the intelligence, one of the real benefits for users is often the design of the guided interface that allows a user to be walked step by step through what might be a daunting set of instructions or forms or a complex transaction – such as an insurance claim or a mortgage quote.
Gartner suggest that organisations will spend more on conversational systems in the next 3 years than they do on mobile applications. This would suggest a shift to a ‘conversation first’ interface model. There are already some excellent examples of early movers here. Babylon offers a conversational interface for providing initial medical inputs and is approved by the NHS. Quartz delivers news using a conversational model. You can also build conversational applications on Facebook to connect with customers and users. Chatbots are also being used to target online human trafficking. Needless to say, all those clunky corporate systems could well do with more conversational interfaces. Imagine just typing in “TravelBot – I need a ticket to Glasgow on Friday the 9th of February. Get me the first flight out from Heathrow and the last flight back to either Heathrow or Gatwick. The project code is 100153.” And sit back while the bot pulls up options for you, and also asks you whether you need to book conveyance.
Conversational systems will certainly make technology friendlier. It will humanise them in ways we have never experienced before. I often find myself saying please and thank you to Alexa and we will increasingly anthropomorphise technology via the nicknames we give these assistants. You may already have seen the movie “Her”. We should expect that this will bring many new great ideas, brilliant solutions and equally pose new social and psychological questions. Consider for example the chatbot that is desi§gned just for conversation – somebody to talk to when we need it. We often talk about how AI may take over the world and destroy us. But what if AI just wants to be our best friend?
My thanks to my colleagues and all the discussions which have sharpened my thinking about this – especially Anantha Sekar – who is my go-to person for all things Chatbots.
My book: Doing Digital – Connect, Quantify, Optimise – is available here, for the price of a coffee!
As with all my posts, all opinions here are my own – and not reflective of or necessarily shared by my employers.