India is fast developing into a major player in the artificial intelligence (AI) domain with close to two-thirds of all IT professionals working on actively deploying it in their enterprise. However, the question that needs to be asked now is around how smart is the machine learning process and whether it is smart enough to detect biases that could lead to discriminatory outcomes?
We had dealt with this topic just last week, while assessing the human bias that is bound to creep into initial AI programming via algorithms. Which is why we thought it would be interesting to understand what an expert at IBM feels about the situation. In an article published on India AI, Siddhesh Naik, leading AI and Automation sales at IBM South Asia shared some insights.
A sudden spurt in AI-led functions
Since the 1990s, India has been developing expertise around big data analytics, machine learning and AI, though it was only in the post-pandemic period that we saw a big boom around ideas like chatbots, digital assistants and driverless cars, says Naik adding that companies are now using AI to augment core business processes and services.
He highlights three specific trends in this industry, the first of which relates to using AI to stay ahead of competition and build efficiencies in the business processes using automation. The second is the realization that data management is crucial to AI deployment, which necessitates the right set of tools. The third one is the most important as AI users realized the potential of creating trust in the algorithms.
Naik considers IBM’s AI solutions for business to be a good option for enterprises to scale AI with capabilities that include natural language processing, automation and trust with the ability to function across hybrid or multi-cloud environments. However, he also shares a warning note that things could get murky if one slips up on the trust bit.
Watch out for these obstacles
He says enterprises need to clear certain obstacles before they can successfully realize the potential of AI adoption. Things like data complexity, the right tools or platforms to develop AI models, the limited exposure to AI skills and expertise, and the penchant to bring complex and hard-to-integrate projects into the ambit of AI-led solutioning.
This is largely due to the fact that Indian businesses haven’t exactly been data friendly, which results in multiple data sources, types, structures, environments and platforms being used. This gets further aggravated by the use of hybrid or multi-cloud architectures, leading to operational data remaining in silos resulting in possible blind sides to data analytics.
The challenges aren’t insurmountable
Naik believes that enterprises can address these challenges by a data fabric approach that allows them to use disparate data sources and storage repositories. However, even if data disparity gets sorted, enterprises need to build trustworthy and responsible AI systems that reduce bias, monitor performance variations and model drift while continuously attempting to explain AI-powered decisions to all stakeholders.
“At a time when AI-based applications are expanding into uncharted business and society realms, Indian policymakers are charting its potential for growth and social transformation. The Indian government’s policy think-tank, NITI Aayog, is actively working on the AI for All program to democratize AI usage. Additionally, the Government intends to leverage AI for healthcare, agriculture and e-government under the Digital India initiative,” he says.
Governments and enterprises must proactively prepare to embrace the growing use of AI in crucial decision-making. While enterprises should define and adopt safety standards before adopting AI to manage their operations, governments must ward off human biases around non-data driven facets of our existence.