Everyone of us know about Machine Learning – AI for a long time now. I have highlighted some of the day-to-day applications which we use often where ML/AI are applied. Also, how these use cases have improved over the time. Let’s have a look.
Text/Image based CAPTCHATO AI based NoCaptcha/ReCaptcha from Google
CAPTCHA was a common item in any login based websites. I remember entering some juggled-strike out-dancing alphanumeric characters. Then some came with Q&A model of solving CAPTCHA. But now, looking at NoCaptcha/ReCaptcha, AI has changed the way CAPTCHA’s are used now. One of the biggest problem with increasing security layers is the increasing complexity in usage resulting in a very bad user experience. With this new solution – It’s clear that technology should be soft on humans but hard on hacks/bots. Earlier both the bots and the real user were treated the same. I like this AI solution which increases the complexity for bots but gives the user an experience like there is NoCaptcha.
Kiosks/IVRTO trained answering machines
Kiosks were always designed to provide a visual representation of data. They used to be static with no context to it. Interaction used to be always one way. Now with solutions like Wipro HOLMES or IBM Watson – interactions are both way. Rather than we ask questions and then get answers – with the machine learning capabilities bundled with the kiosks they can guide us based on the context as well. Kiosks are now context based interactive kiosks
IVR’s used to be waiting on the response from the user to proceed further. Now based on the initial input that we give IVR rank the priorities. If we take our usual bill payments they prompt you if there is overdue. If we call pharmacy they remind you of refill or suggest visiting doctor based on your health parameters and then do a refill.
FAQ’s/Need some help sectionsTO Chatbot’s assistance
Remember the days of looking at the hundreds of Q&A in help section to resolve the issues. Even though the help section used to provide lots of data, it used to be time-consuming to pick the right information that we are looking for. Think about an assistant which can help you with what exactly you are looking for. This is a greater opportunity for the learning algorithms to understand to the customer needs and respond back. Looking at some of the solutions offered by Avaamo on this area – It’s pretty much challenging to resolve the issues by giving the user higher availability on support windows and a richer user experience.
PlaylistTO personalized autopay in YouTube
YouTube simply used to suggest some group of videos at the end of the playing item. Every time a manual intervention was needed for the next video to play. Auto Play is not so new – It was tested long time before on selected users. Currently working with one of the leading satellite TV provider of US. I don’t see this as just a small technical feature. I see it big more competitive in many ways. Google’s solution to TV watchers. Auto play kind of connects the dots and makes YouTube as a continuously running channel. And the good thing is that the learning engine customizes the channel uniquely for everyone. A simple question – “Why do we think of switching between channels?”. A simple answer – “To watch the content of my interest”. Think of a scenario – “What if I can watch Game of thrones in HBO and Impractical Jokers in Tru TV on the same channel without even knowing what HBO or Tru TV is” – I feel YouTube with features like Auto play, offline support has almost done on the future customer expectation and needs.
Individual website sign inTO Direct browser sign in
It’s month end and you must pay for your electricity, telephone, Internet, home rent, insurance, etc. You open your browser and then you try to sign in to every individual site to finish your payments. Really – Do we still want to do this in this digital era. Even in this digital era – there are service providers who doesn’t have a mobile app. So, when browser is the only way out to pay your bills, wouldn’t it be great if someone auto signs you or remembers your credentials and lets you sign in to every individual site so fast and seamless. I see signing into Chrome or safari as a feature that does this for me. From wherever you try to launch chrome – be it your mobile or laptop or on your car, chrome auto signs you in all the website that you have marked to be remembered. So, simple. However, I see this feature as short-lived. But, this will live till the last day of the browsers.
RecommendationsTO Personalized experiences in Amazon and Netflix –
Few years back, amazon’s recommendation engine – “Item to Item Collaborative Filtering” was considered as the killer feature. It drew some criticism on rooms of improvement citing much more can be achieved through the engine. Now looks like amazon’s investment on learning algorithms are paying back. Previously, if you are looking for storage cart, we would have got recommendations with 2,3,4,6, etc. – drawers. But now with the improvements on “Learning to rank” we might get storage cart with 4 drawers only based on the varieties of ranking elements considered. Also with the DSSTNE (pronounced as “destiny”) – latest open source AI recommendation engine of amazon – anyone can provide hyper-personalized-customer-experience
Netflix, on the other hand, understood that personalized recommendations would increase the amount of time viewers watch content on Netflix reducing the churn rate as low as possible. With its recent expansion to different regions globally (for instance Netflix is launched in India recently) – Netflix also have moved from territory based recommendations to personalized recommendations.
Simple voice dictationsTO contextual experience with Siri or Google Now/Google Assistant or Alexa –
Robots were a big BUZZ word when I grew up. But to me the there was always an unanswered question. Why do we depict robots as always something that looks like us? To me, anything understanding the context and deciding the action is a robot. Let’s take a simple delegation to any machine – “Pay bill”. If the action out this sentence Is going to be a question back to me asking whether its telephone bill or internet bill. Then we are 5 years behind today’s technology. Machines should be able to understand the context of which bill was meant to be paid. This can be based on priorities, ranking algorithms, etc. Anything that guides the machine to understand the context and come out with one unique result. With different layers in deep neural networks companies like google, apple and amazon are trying to teach machines with the help of machines by investing huge in this space. Today’s technology is just a start towards this area. Cloud, 5G and IoT together can give greater results on this area.
Traditional authenticationTO biometric or cloud based identity management systems –
I had no answer on – Why we should just hide passwords with asterisk and why not username? while we enter username and password in any websites. Even usernames in any banking websites are not meant to be shared. In cases like Gmail or Yahoo mail – I understand that your username is your identity and we share it to get communicated with each other. In terms of banking we use account numbers for any transactions. But still we don’t use account numbers as username for login. To me, it looks like it has become a standard to use username with visible characters and password with asterisk no matter what’s the purpose or which domain does it address. Now there is answers to all this weirdness in the form of Fingerprint, facial recognition, trusted places, On-Body detection, Trusted devices, trusted voice, Identity management systems like – Google authenticator, Okta, etc. Now all the above-mentioned systems provide unique way of authentication experience to the customer.
Add to shopping cartTO Dash buttons in amazon or Smart trash bins –
DRS (Dash Replenishment Service) from amazon changes the shopping experience. Basically, it automates the shopping cycle. There is no need to pick the items every month if you know that we are going to get the items again and again. You can call my explanation so far as just ‘Subscription’ and What is new about this. This is an age-old technique followed in many other industries. What is so new about this? Amazon Dash Replenishment Service allows manufacturers to add a physical button or auto-detection capability to their devices to reorder supplies from Amazon when necessary. So, this behaves as an ‘Intelligent Subscription’. Some other folks thought how this problem can be solved in much more simpler way. In simple terms, they added the shopping cart module trash bins. Whenever you throw an item in trash simply scan and then throw it. So, simple. Your items will be delivered by drones. (My Idea – Why not the shopping module in the trash bin find the cheapest price among different sellers and place the order with the cheapest seller. Wow)
Text searchTO Image search in Google –
Products like google search, Google photos use extensive image processing and recognition to improve search. Image search is truly considered as an internationalized search. We search to know what it is about. Even to ask about what we are trying to ask we need to know little basic. But with image search simply take a photo and then leave the rest to google. We get varieties of matched results on the searched image. Google photos matches and groups the photos based on location, person, occasion, etc. Google photos adds context to the photos.