If you’ve searched the markets for the next big thing which could bring huge savings or more customers to you, you might have noticed the nearly unlimited amount of technologies, products and startups. Also showing an actual business case with good ROI to top management is very often almost impossible since tech is seen as expensive and too risky to take into use. Everyone is talking about it, very few are actually doing it. Just look at the Gartner Hype Cycle.
Finding the solutions that would actually be useful and mature enough for your business is totally different question. Many “smart” devices are actually pretty stupid (“Hello, I’m a smart egg tray”) or dangerous because of their poor security.
Also promising startups and even larger companies might go bankrupt and take the whole proprietary cloud dependent smart devices with them. Some of the bigger names that have exercised this kind of behaviour are the infamous, overly engineered Juicero juice press (that originally retailed for $699 to press some juice) and smart watch manufacturer Pebble (which was bought by Fitbit). Smart devices and IoT platforms are often walled gardens that make it impossible to use the devices through some other vendors services. When the cloud is unplugged, everything is gone.
Why in the world would someone take these kind of risks and invest possibly millions to include smart technology in their products or processes only to find the technology unusable some months later?
On top of startups and clouds going down, market is filled with technology that is still so immature that - considering the majority of potential users - there's really no point investing into it any time soon. Companies are luring people to invest into the next big things and the actual value is nowhere to be found.
We’ve been putting together a list of hot IoT and other related and as much hyped technologies which are relevantly safe to invest in at the moment, as long as you consider also the drawbacks.
In part 1, we will cover Chatbots (the working kind) and Face recognition. Later we dive into technologies such as Edge Computing and AI.
Conversational UIs are definitely one of the most popular new tech taken into use recently. Textual communication is used on daily basis via different tools both on and off work. Natural language could be natural for a human being but it has been difficult for a machine to interpret it. During the last few years it has become more viable as machine learning technologies have made it possible to parse language into structured data. Since everyone else already has a chatbot, should you?
If your customers have simple repetitive tasks they have to do on their digital devices, like reserve a meeting room or check for the lunch menu of nearest restaurants, a chatbot can serve them well. This kind of limited scope can work well in a chat context either by using predefined keywords or maybe even with a simple selection of natural language interpretation. This kind of chatbot can free the user from navigating through multiple apps or views and achieve their goals with a few words on their favorite chat app.
For more complex contexts, a chatbot could be reasonable solution when used as a tool to complement a human. Implementing a full conversational UI requires a narrow enough scope and plenty of existing conversation data for achieving a proper intent recognition. For simple questions, like “when are you closing today” or “how long does the delivery take”, a chatbot can provide a good enough answer. A perfect solution will make processes more efficient and can save time for the humans.
Some contexts can benefit from a chatbot even if there is no reason to chat with the bot. Using a one-way chatbot to send notifications and alerts can make it possible to reduce the amount of applications to follow. Most of the group chatting tools already used in many organisations are have support for bots.
Even if you could try to fool a user into thinking they are chatting with a real person, the technology is not that far yet. Trying to cover too wide range of different conversations and contexts will lead to uncertainty. A wrong interpretation of the user intent or even just “I’m sorry I didn’t understand you” can be a frustrating experience for the user in need. A proper scope will also make it easier to define the text used in responses. Textual content has to be handwritten still, because natural language generation is not really feasible yet.
Face recognition technology has come a long way from its early days. With modern machine learning, it is quite easy to setup basic face recognition. Most of the effort can be used on the actual solution built on top of it. There are already plenty of libraries capable of calculating facial landmarks, such as dlib and openface. Also cloud-based solutions are available from major vendors.
Identifying the user by face can be a nice experience for situations where the user may not want to use their mobile device or does not bother to install an app or open a website. Facial recognition can be used for personalizing content or tuning settings for personal preferences. Unlike traditional authentication, face recognition can be done in a seamless way without usernames or passwords.
Finding faces from images is a trivial task and calculating features from these facial images is already quite accurate. These calculated features can be compared to a database of facial features to give that face a name. There are naturally several variables affecting the accuracy, but in general facial recognition works really well.
One major advantage in face recognition is that a person can be identified without explicit user actions. In this kind of smart environment, the environment can be customized based on the identified user. Showing the next meeting on person’s calendar on an info screen or sending a message about arrived guest to the host can make a big difference compared to traditional device-centered experience.
Face recognition might be a valid method for identifying people, but the identifying factor is barely a secret. One rarely leaves the house without her face, so taking copies of the face is quite easy. Using photos or video of another person, an impostor might try to fool the recognition device. That can be avoided by using liveliness detection or use IR cameras for preventing the use of fraudulent content. In many cases requiring high reliability, like opening locked doors or handling of money, some way of multimodal authentication has to be considered. Face recognition is also currently a really sensitive topic regarding privacy, so you might consider whether you should ask or at least inform the users about the ongoing recognition.
We’ll be back with part 2 as quickly as we humanly can! Mikko Pohja & Lauri Anttila