Machine Learning, Deep Learning and Artificial Intelligence have moved into the realms of every day life
Machine Learning, Deep Learning and Artificial Intelligence have moved into the realms of every day life and are no longer restricted to the scientific research or programmers’ domains. As they occupy centre stage in influencing every aspect of business and individual’s decision making process, the predominant question that is being frequently raised is the efficacy of such decisions and the resultant actions. Such concerns are expressed even though technology has been able to add tremendous value in doing better jobs than humans in areas which are repetitive in nature like in the case of retail checkout stations, areas that have high risks associated with errors like in the case of banking transactions or could be life threatening like in the case of deep sea diving in oil and gas exploration sites. These applications and several others particularly when combined with advancements in voice recognition, sensors and natural language processing are delivering solutions that are reducing costs, offering pertinent choices to the decision makers, making the processes more efficient and removing the biases from the actions.
Despite such obvious benefits, AI and technology driven solutions often times are unable to factor emotions, intuitions, experience curves and personal values in arriving at the recommendations. There is also the danger of algorithms influencing and shaping a whole new mindset and opinion forming as has been highlighted by the recent cases of the role of social media and analytics in electoral behaviour.
Since the bots and algorithms are developed based on available data and known patterns of occurrences or a series of actions initiated in the past, marginalised sections of the society or under represented voices would stand to lose out due to their limited presence in the digital medium or partial views towards them thus leading to lopsided recommendations or actions thereof.
Algorithms also falter at times due to common sense failures or misunderstanding the intent in a given situation. When a human interact with other humans they have a common knowledge of how the world works and how to transact in any given situation which is embedded in the DNA of humans however machines do not have this advantage and can relate to only the clearly defined data and training they have undergone based on this data for predefined situations.
Therefore, before the process of building algorithms commences, it is important to address these issues in order to make the algorithms reliable, ethical and pertinent to situations. To facilitate this, human centered design of algorithms would be essential such that spontaneity and opportunity to relate to personal experiences are encouraged in order to make the automated decision pathways more robust and practical.
Organisations need to be also conscious of the biases that could arise both on account of humans as a result of personal and societal or corporate experiences as well as those could arise out of machines on account of limitations of data and evolve new framework for decision making that combines the best of both – humans and machines. Repeated training and reinforcement would be required for both the AI tools as well as for those who would be using them to help them interpret the recommendations more accurately.
It is therefore desirable that algorithm and AI tool creation are planned as a collaborative effort between the digital experts and the domain experts with the full cognisance of psychological, legal and social factors relevant to the ecosystem of the stakeholders. Algorithms with the human touch is the recipe for success!
The writer is CEO, Global Talent Track, a corporate training solutions company
No comments:
Post a Comment