The Importance of Data in Artificial Intelligence
For us to allow AI to progress we need to encourage easier access to cross-border data flows
In the contemporary world, artificial intelligence is amongst our grasp. An intelligence that can process more information at speeds that weren’t previously conceivable. It is not that the idea behind AI is a new one. In Forbes’, ‘A Very Short History of Artificial Intelligence’, Gil Press traces back the origins of AI to Catalunya in 1308.
While AI has been around since centuries, it has never had the scope of application that it does today. For instance, consider a scenario where we can ask the Google Assistant to find restaurants near me. The application first detects your location, scans the nearby eating joints, collects their phone numbers, calculates your distance to them (by walk, bicycle, car, and public transport), gets a link to their websites, gathers the menu cards, and shows you how other people have rated them. It does so in a matter of seconds, ready to take on your next command.
Complex Functions that AI Does
If you were to point to how Google Assistant and other AI programs are able to perform such complex functions with ease, the answer would point to data. Humans, as a race, are recording our activities on a scale like never before. This recording is facilitated through sensors, digital activity, and data. Because our search histories, friendships, and payments are recorded, we can observe patterns in when and how we perform certain activities and make ourselves more efficient in doing them. For example, if one has to use the metro to commute to work every day, it would be handy to have the train timings on your phone/watch as you enter the station. Similarly, when getting in the car, it would be helpful to have the shortest and/or least congested route to work with you. AI today takes care of both those scenarios, giving you one less thing to think about.
Not only is AI able to simplify life by reducing stages from our patterns, but it can also help identify new patterns as well. Moreover, they may be patterns that we might not have the foresight to predict. Google’s AlphaGo, a deep learning system designed to play the board game Go. It made multiple moves that were eventually successful (beating world champion Lee Sedol 4-1). However, why it chose to make those moves is beyond human comprehension. An article by WIRED, ‘How Google’s AI viewed the move no human could understand’, called the moves ‘inhuman’.
Importance of Data
All of these predictions, moves, and insights are possible because of data. AI learns from all the data it has available. The more data it has the better developed its insights become. This brings us to the question of how this technology can benefit from cross-border data flows.
AI getting access to cross-border data flows can have two conceivable implications for the future of the technology. Firstly, the availability of transborder data flows, the volume of data for AI to process would increase exponentially. Regardless of how huge national datasets may be, they would be tiny in comparison to what they could be once data from different countries/companies supplement them. With the availability of cross-border data flows, AI would have more material to learn from and more patterns to uncover. It would speed up the development of the software that processes these insights. This brings us to the next implication, the nature of cross-border data flows.
Cross-border data flow, because of the diverse data range it can carry, is a very lucrative prospect for AI-based applications that aim to tackle international problems. The availability of international data can elevate AI from a national level to a regional one. Think of operating on not just an Indian, but a South-Asian landscape to identifying aspects of broad-based problems or opportunities.
More Practical Aspects
Consider the scenario where RyanAir (a low-cost European Airline) wants to invest in an AI algorithm to devise new routes. To get the algorithm working, it would need to take into account a certain number of variables. For instance, the current amount of movement between the two cities across different forms of transport, gauge the price the consumer is willing to pay and check if the price offered by RyanAir will be competitive in the sector. Notably, this process is scalable, and if RyanAir has enough data, it could perform the same operation for an n number of routes. However, In order to get these insights, RyanAir would need access to data sets from London and Geneva. Building on this, AI could solve for similar patterns for any or all of the routes RyanAir might be looking to operate in. Successful completion of the insights developed would require less-man hours, output more patterns than humans would, and would require a fraction of the money it would take to conduct such an analysis manually.
So, what would the free flow of all data mean to a program that feeds on data to grow and learn? It might mean everything. For the growth of AI, it is crucial that we enable cross-border data flows. The principle of the more the merrier applies. So, for us to allow AI to progress at the speed we know it can, we need to encourage easier access to cross-border data flows. In doing that, we open up the scope of identifying our patterns on a broader level to solve for bigger problems.