Generative AI, a frontier technology capable of creating fresh content from existing data, is poised to disrupt various industry landscapes by leveraging machine learning models to analyse patterns, understand contexts, and generate new content or solutions, it can optimise processes, enhance creativity, and drive innovation.
Nvidia, a company that is famed for video game hardware, saw their valuation touch $1 Trillion. This is on the back of growing interest in AI. India, however, stands at a crucial juncture. On the one hand, there is a vibrant startup ecosystem exploring Generative AI, and on the other hand, there is a push towards becoming a manufacturing hub. It is expected that AI will add $967 Bn to the Indian economy by 2035 and $450–500 Bn to India’s GDP by 2025.
AI would add $967 Bn to the Indian economy by 2035 and $450–500 Bn to India’s GDP by 2025
Indian generative AI start-up ecosystem is at a nascent stage, with around 60 registered start-ups majorly offering generative AI-native solutions built in-house. However, the space is rapidly growing. Out of the total funding of over $590 Mn received so far by Indian generative AI start-ups, over $475 Mn was received during 2021-23. However, the road to leveraging Generative AI is paved with hefty hardware requisites and procurement challenges.
Central Processing Unit (CPU): The heart of computation, a high-speed CPU, is crucial for navigating the complex mathematical labyrinth inherent to Generative AI. Examples: Intel Xeon Platinum or AMD EPYC processors
Random Access Memory (RAM): A minimum of 128GB RAM is mandatory to harbour vast training datasets and the generated content.
Graphics Processing Unit (GPU): Known for their parallel processing prowess, GPUs expedite the training and operation of Generative AI algorithms. Examples: NVIDIA’s A100 or A40 GPUs
Use and Training: Deploying Generative AI models, like the colossal GPT-3, necessitates a hardware arsenal robust enough to withstand intense computational onslaughts. The journey from training to using such models is a hardware-intensive endeavour, with training being the more demanding phase due to the iterative nature of learning algorithms.
Procurement Challenges: The quest for suitable hardware can be a daunting expedition. The high costs and limited availability of potent CPUs, GPUs, and ample RAM are significant hurdles. For instance, a single NVIDIA A100 GPU could cost over $10,000. Yet, rays of hope emanate from cloud computing platforms like Google Cloud or Amazon Web Services, which offer on-demand access to formidable AI hardware. Moreover, the process of importing electronic components can be complex and time-consuming, involving a lot of paperwork and compliance requirements.
Next-Generation Hardware Avenues: What the Future of Hardware Holds
Quantum Computing: By harnessing quantum mechanical phenomena, quantum computing could significantly accelerate Generative AI training, opening doors to faster optimisation and simulation of complex data distributions.
Neuromorphic Computing: Emulating the human brain's architecture, neuromorphic computing promises efficiency and adaptability, potentially lowering power consumption and fostering better learning capabilities in Generative AI.
In 2021, a significant progress was made by Google AI towards building fully error-corrected quantum bit (qubit) prototype. Google AI achieved 100x suppression of logical errors as the code size from 5 to 21 qubits was increased.
In 2022, Intel's unveiling of Loihi 2 showcased the potential of neuromorphic computing in Generative AI. Neuromorphic computing refers to the design of computers and software that are inspired by the structure and function of the human brain. Intel’s Loihi 2 is a neuromorphic research chip can process neuromorphic networks up to 5000x faster than biological neurons.
A research by McKinsey estimates that the rapid advancement of the generative AI will globally unlock economic benefits in the range of $2.6 Trillion to $4.4 Trillion annually across industries. The journey of Generative AI, from a promising concept to a game-changing reality, hinges significantly on the evolution of hardware technologies. As we stand on the cusp of next-generation hardware innovations, the horizon of what's achievable with Generative AI expands, bringing us closer to a future where the synthesis of artificial and human intelligence seamlessly blends.