By Anusha Rammohan, Co-chair. AI Working Group, IET Future Tech Panel
The recent advances in generative AI have been heralded by a mixture of awe and trepidation. While some believe in its transformative potential, where generative AI is the stepping stone towards general Artificial Intelligence – the holy grail of AI research, others have issued dire warnings about AI-induced job losses, model bias and inaccuracies as well as security and copyright concerns. But the one undisputed fact remains that generative AI is here to stay and in the short term, enterprises have started to consider the inclusion of generative AI in their overall digital and AI strategies as a foregone conclusion.
Over the last year, generative AI has become one of the core areas of focus for not just start-
ups but large technology companies as well. India with its strong base in services as well as
technology products is particularly well positioned in this space. In fact, India as a country is already ranked 6th in terms of investment in AI as of 2022. With ample access to great AI talent and the ability to generate large amounts of data from a relatively diverse population, India can definitely aspire to lead in the generative AI technology race. But when it comes to cutting edge digital technologies, generative AI is as expensive as it can get and, in more ways than one!
Whether it is the resources involved in training a model or the time it takes to build a complete generative AI solution, enterprises may find it hard to justify the ROI. Hence, they may be left wondering whether to buy off-the-shelf solutions or build it themselves.
While the conundrum of build vs buy may not be new to most organizations, the arguments on both sides for generative AI are full of nuances specific to this technology. For one, a generative AI model involves multiple training steps. The pretrained foundational model is perhaps the most cumbersome and is both resource and time-consuming to build from scratch. The second step which is fine-tuning this model for specific applications is easier but still requires some resources and time. Most AI solutions also require an application layer that interacts with say other digital systems within an enterprise. Lastly, any AI solution requires periodic monitoring, maintenance and updates as needed. Enterprises have the choice at each of these steps to build in-house or outsource to AI consultants.
The parameters for making the choice depend on the access to talent, ROI, and capital investment for in-house deployment vs operational costs for outsourced efforts. But the most important factor is what an enterprise’s overall short-term and long-term AI strategy is and how generative AI fits into this strategy. Having a clear strategy involves having a coherent plan around how, when and where generative AI can and will be used within an enterprise. As with any new technology and especially one such as generative AI with all its hype, enterprises often find themselves struggling to formulate an approach that can be expected to guarantee concrete outcomes. Whether projects are done in-house or outsourced, they can only be successful if tied to outcomes.