brief about machine learning and new trend of human-computer interaction

Google I/O attracted lots of attention recently, and many people are talking about bot assistant, machine learning, smart device, and so forth. Google’s 2016 presentation is more down to the earth and back to the basics, focusing on making developers’ work easier and developing more powerful machine learning algorithms, instead of showing off those fancy but immature products, e.g. Google Glass.

Google is not the only company that has been exerting efforts and resources into machine learning. In recent years, gurus like Facebook and Amazon are also getting into this area. Yann LeCun, known as the founding father of convolutional nets, became the first director of Facebook AI Research in New York in 2013. Amazon gave its machine learning capabilities to developers as an AWS offering in 2015. Big companies are not the only ones in this arena. Clarifai, providing visual recognition API and service, MetaMind, doing automatic image recognition, Premise Data, indexing and analyzing millions of observations captured globally and figuring out connections that impact global decisions, are in the battle. Just name a few. Besides keeping themselves in the trend, acquiring more users and engaging users are the most important reasons why so many companies join the ride. As more users use the product more, more data will be gathered, and the algorithm will get smarter and provide better user experience. The better user experience in turns will make users more sticky and the monetization easier. Though personally I don’t think “Winner takes all” is what will happen, obviously users have limited attention and time, so grasping users is crucial.

Machine learning gets more powerful these years because of the fast development of large-scale deep learning. Researchers and scientists are making machine learning algorithms easier to use for developers, and companies like Google and OpenAI, newly founded by Elon Musk and Sam Altman, even open the source to the public. We are seeing more and more real usages of machine learning in our everyday life, e.g. image recognition, translation, voice assistant, shopping recommendation, advertising targeting.

Technology as basic is not the only reason why machine learning grows so fast, users’ needs can also explain why. We are in an information bomb era: too much information, too many apps/software, and too many devices. Human-beings’ attention have never got so distracted. Our brain hasn’t got the time to evolve the ability for us to restore and get the information accurately and quickly when we need it. Machine learning can greatly help users to get the information efficiently, and even predict that before users ask for it. Meanwhile, emotionally, we need to feel we are in control and privileged, which is challenging, as users are becoming more and more like data sets instead of human beings with thoughts and emotions. Together with those new machine learning usages come the new way of human-computer interaction and new product ideas based on that. It is interesting to see those changes.

Before, we purely used text to send messages, Google to do search, and apps on iOS or Android to finish various tasks. Traditional way makes information discrete. Users have to install so many apps on their phones and switch between browsers and apps constantly, and their communication with other human beings is limited by the functions of texting.

​An app centralizing all the information and communication would be very helpful. I haven’t seen much in US market, but in China, WeChat has been popular for five years. Many people say that WeChat is the Chinese version of Whatsapp. This is not that correct. Whatsapp’s main function is for fast direct communication, while WeChat includes more elements, e.g. media, user stories/moments, services (taxi, food delivery) and payments. Enterprises can register a public account on WeChat, doing marketing and PR by generating contents and also providing services through WeChat’s platform. By communicating with those enterprise accounts directly, users don’t have to open another app to finish certain tasks.

​Something similar is Google’s Instant Apps (in a couple of mouth), aiming to make the web and apps one. Users no longer have to download any app that only handles a small task with very low frequency.

Above image is obtained from "B&H Photo (via Google Search)"

​Another one is the support of quick reply notification. Users can reply a message right on the locked screen on smart phone.

​Last but not the least, Gboard, released by Google before Google I/O, a new app for iPhone that let users search and send information, GIFs, emojis, and more right from the keyboard.

Above image is obtained from Official Google Blog

WeChat, Instant Apps, quick reply notification, and Gboard centralize the access of information and interaction, but also bring the challenges to the industry: Will only a few companies be better off from this? What should the apps do to increase user acquisition and engagement then (many say that the end of apps is coming)? This is definitely a new challenge, but I believe new dynamics among companies will emerge to move things forward.

All above are just about the basic communication and interaction needs. A more advanced one is to have a virtual assistant who can do things for you when or before you ask for, and even make you feel understood (to some extent). There are quite many to choose from: Apple's Siri, Amazon's Alexa, Microsoft’s Cortana, Facebook’s M, and the newly announced Allo and Google Assistant (advanced version of Google Now), and Slack bot. They are different—some are voice assistant and some are chat assistant, whichever way, they are trying to make users life easier while grabbing as much data as they can when users are using them. Machine is good at looking for data and doing calculation, while though human beings are not that efficient and rational, their emotions are powerful to build the connections. If we let machine and human beings focus on what they are good at respectively, this might create a win-win situation between companies and users.

However, will the win-win situation really happen? If so, will it sustain? What will the future be like? This is such a big topic, and even another blog post couldn’t talk about it clearly. I just have two thoughts here: Since machine can predict what we do so as to react accordingly, will our life and the world become very boring? In addition, will human beings stop thinking and acting proactively and one day we become obsolete?

Even talking about now, there are some concerns I can think of:

• Human nature: One of human natures is to form relationship with each other. Surely, if the situations in movies like <Ex Machina> and <Her> happen in real life, we might have difficulties distinguishing humans from machine. Would this be really possible, even in the far future? There are enough arguments regarding what is possible to happen to artificial intelligence and the timeline. Personally, I think we are still very far from the movie scenes if it's ever possible. For now, the experience chatting to a bot and chatting to a human is different. So for the above mentioned "privileged feeling" users need (see I got an assistant who can do those tedious and repetitive tasks for me), will what bot provides a double-edged sword? On one hand users do feel the privilege, and on the other hand, their emotions are not empathized.
  
• Data security: Assistant will greatly increase the degree of our interaction with the devices. With more and more data going to certain companies, data security becomes a serious issue. The concern is not only about if the data will be leaked out, but also about what the company or algorithm will do with those data. Google doesn't want to be evil. Let's assume it's true, but will any other company be? Let's assume no. But will company really have a control over its algorithms? Virtual assistant relies heavily on machine learning. Paradoxically, the more advanced machine learning algorithm is, the more the neural network is a black-box to us—not only for users, but also for developers.
• Predictive ability: How much will a machine understand human beings? Some research papers show that data can tell us more than we know about ourselves. I’m not quite sure about it. So predictive analysis can tell us what we really want? Let alone there are so many decisions that are not black and white, which require judgment calls, we change our mind (unpredictable from our past behavior) quite often. So how much can data really tell?

​Machine learning is an interesting topic. It is profound that only hundreds of people on earth are extremely good at machine learning research and moving this technology forward. At the same time, it has been permeating into our life, as the usage of machine learning becomes increasingly pervading and even affects how human beings interact with each other. It deserves the attention and discussion, and hopefully it will bring human beings to a better future.