Scaling AI for businesses: Challenges and potential solutions

“Implementation of AI in businesses requires re-thinking in terms of how people are experiencing AI and how much organisations are willing to research and invest in the complexities and potential solutions surrounding AI. For AI to become more self-reliant and capable, more human intervention in terms of understanding the complications of language and mathematical calculations is required.”

For people familiar with technological advances, AI invokes excitement, possibilities and caution. Intelligence, artificial or not, must be precise and useful in its application. For years, AI has been experimented upon and studied in detail. As enterprises around the world adopt technologies to strengthen their business motives, experimentations with AI technologies continue to ease business operations and fuel growth customer bases. But integrating AI in businesses is not free of problems and for sounder real-life implementation of AI, resolution of these problems is essential. And to do that we need to scale very part of the technology stack.

At Facebook, we use AI to understand and decode content, especially in terms of Facebook’s policies and whether any content, in any form, is violating the organization’s policy standards. The potential of AI use in Facebook is profound. We are working towards integrating AI into Facebook to make the feature of ‘Recommendations’ more popular, especially for cases such as an emergency requirement of blood, where swift and accurate responses are essential.

Implementation of AI in businesses requires re-thinking in terms of how people are experiencing AI and how much organisations are willing to research and invest in the complexities and potential solutions surrounding AI. For AI to become more self-reliant and capable, more human intervention in terms of understanding the complications of language and mathematical calculations is required.

Businesses need to explore the relationship between AI and human intelligence with context. Human interventions in the form of teaching AI systems properly will ensure greater accuracy of AI functioning. Present AI systems require a lot of human supervision and examples to emulate actions. The goal for the future is to reduce AI’s dependency on the human tutor and make it self-taught intelligence, capable of performing certain functions on its own.

Companies which have large amounts of data should work towards feeding AI systems with all the data and allow the AI mechanism to learn in a self-supervised manner. The importance of language in AI usage and implementation cannot be stressed enough. For example, in Facebook unlike earlier models of unsupervised translations of language by AI, we now use AI models which can map different sentences in different languages and ensure coherent translation.

The scope of AI is massive and thus a lot is expected from AI in business organisations. To meet those expectations, an ideal AI system should be self-learning and able to represent geographical diversity, thereby helping organisations to expand their services worldwide. To that end, businesses need to accelerate digitalization of the organization and acquire or train a competent workforce.

[box type=”shadow” align=”” class=”” width=””]Insights for the above article has been taken from ET Tech Summit 2019 – Session titled, ‘Scaling AI: Pushing the frontiers on every part of the stack’ by Manohar Paluri, Director (Artificial Intelligence) at Facebook.[/box]