May 18, 2017
We at Let’s Talk Payments love to debate on potentially game-changing technologies – and artificial intelligence has certainly been one of the candidates for the aforementioned buzzword category. So we got our team together to discuss how AI has changed their lives with them being the end-user of any product or service that used AI in some way. The purpose here was to gauge where we are in terms of realizing the technology’s impact and its manifestation in our daily lives.
People who bought this also bought
You may also like
While shopping online, quite more than often we come across above lines along with a bunch of really cool products displayed below them. This is a classic example of collaborative filtering, a method in machine learning which involves making automatic predictions (filtering) about the interests of a user by collecting preferences or taste information from many users (collaborating).
One other interaction with AI in our daily life is the entire YouTube video watching experience. This ranges from curated playlists and channel suggestions to YouTube video recommendations, all of which has been derived from our activities on YouTube. Google collects data from YouTube users, uses algorithms to find similar videos, filters out noise, combines related videos with user activity, and ranks items based on several signals, system architecture, and evaluation metrics.
There are many other instances where an AI engine makes it easier for us to choose such as the address bar in Gmail which predicts which addresses we might want to keep in To and CC based on our previous mail account activity.
However, we humans have accustomed ourselves to these nuances as part of the experience and seem to have made peace with the longer term behavioral changes that stand to be brought upon by the highly addictive nature of recommendations. Now we discuss some AI implementations which we aren’t exactly fans of; barring some exceptions, many of our team members pointed out chatbots. At this stage, many members do not think of any indispensable value proposition when it comes to the present-day chatbots. The problems arise when people compare chatbots with the experience they get with a human answering their questions – in that regard, chatbots have a long way to go. These problems with chatbots, however, cannot be overstated because they try to mimic humans, who are one of the most complex creations on earth and the learning curve depends upon two crucial factors – amount of relevant dataset and level of sophistication of the learning algorithm, both of which we think are highly time-intensive. So in our opinion, a chatbot is one of the very abstract implementations of AI that still has a long way to cover on the nice-to-have to the need-to-have path.
We now seek to identify the approaches towards AI implementation that might give an edge. Let’s get back to one of the crucial factors we just discussed – Data. Data is indeed like oil, and the success of AI implementations from data-rich companies like Google, Amazon, and Facebook only touches the surface of the topic in consideration. A successful AI implementation will plug into multiple structured and unstructured data sources; it will have an almost equal focus on data sanity as it has on the learning model.
Another observation which extends from the significance of data pertains to shift in data management practices; this blog by USV discusses this in detail. We think that blockchains seek to democratize the access to data thereby leveling the playground for smaller players. Indeed, the investing theme based on convergence of complementary technologies has gained much ground over past two years, as evident by the proliferation of funds subscribing to the same.
Convergence or not, it would be interesting to see the AI revolution unfold in the next decade. Would it be a mass implementation of more targeted use-cases built on localized data sets or would the convergence of technologies create new internet businesses more valuable than existing ones?