CEO and Co-Founder at Affine, an AI evangelist, business builder, and entrepreneur at heart. Ensuring growth by transformative solutions.
By 2030, AI will likely no longer be getting adopted with simple scenarios and applications. It will be expected to detect life-threatening diseases in the nascent stage, predict weather conditions of a large area over several months and become a digital collaborator to the human race. These are just a few possibilities of the potential impact of AI on life and work in the coming years. The pace of change has been unprecedented in the sector, and it promises to continue in the same vein in the years to come.
With rapid learning and adoption, AI is no longer a crystal ball technology but something that humans now interact with in nearly every sphere of life. In fact, the transformation led by AI has been so pervasive that it is deeply influencing user experience and how humans interact with brands and technologies. The way things are trending, AI will soon become an undeniable part of human life and society.
This widespread adoption and a variety of new use cases will come from the rapidly evolving nature of AI. It is already achieving faster computation, higher accuracy and lower computation and infrastructure costs. Today, AI is evolving across all three dimensions — compute, data and algorithm — which sets the context for its adoption across all realms of life and work by 2030. Here is the direction that I see AI moving within these categories.
Out of all the principal factors driving the evolution of AI, compute is the easiest to quantify. In the coming decade, computing is going to witness a major transformation. Graphical processing units (GPU) are making way for Application-Specific Integrated Circuits (ASIC) and Field Programmable Gate Array (FPGA). This is because both ASIC and FPGA showcase better performance than GPU.
ASIC will make use of multicore processing for carrying out complex AI functions, thus consuming less power. In fact, ASIC is becoming so pervasive that Google has invested in building the tensor processing unit, an ASIC developed for the cloud.
FPGA will go one step further, as it will allow designers to reconfigure designer blocks. AWS’ investment in FPGA through AWS Inferential is an indication of FPGA truly transforming the compute component of AI in the coming decade. Another area of transformation will be the IPUs (intelligent processing units), which focus on massive parallelization of complex high dimensional models and support high compute density. These are all signs of an ongoing deep transformation in the compute dimension of AI, now and in the coming decade.
The data component of AI will transform in terms of multiple sources, level of detail and mode of processing. More sources from IoT, more detail when data is recorded every millisecond and multi-modal data intake by DL techniques mean that more complex interactions will be processed in the future. Data forms an integral part of AI evolution as data scientists need to access datasets cost-effectively and teach their analysis to deep learning (DL) models.
A revolution of sorts is already taking place in terms of how AI is utilizing data to make accurate predictions. Sensors /IoT devices are generating digital dust. Logs from millisecond-level impact systems are generating data at the life span of a quantum time. Users are generating data from traditional touchpoints to systems generating zettabytes of data in basic physical processes (e.g. a chemical reaction), which are together set to transform how AI infers data. Inferring data from non-obvious sources is becoming the new normal.
These signs indicate a revolution in the data dimension of AI, and this revolution is going to be unstoppable in the coming decade.
With further advancements in artificial neural networks (ANN), the way AI thinks about a situation would not be too far from how humans perceive the same situation. This is important because we would then be able to create DL models, which make precise analysis even with limited data.
Every single day new algorithms are being developed that are focused on handling complex data, speed, parallel computation, cost, accuracy and precision. For instance, few shots learning is focused on learning more and deeper with the lesser size of labeled data sets. Distributed DL is creating a set of algorithms to facilitate parallelizing of tensor processing to make computation faster. GPT3 is virtually solving every possible NLP task with the highest accuracy. Experts are leveraging the concept of transformers in computer vision to make algorithms more context-aware, thus reducing the effort of training images through all possible orientations. Variational auto-encoders are being used for unsupervised domain-free anomaly detection.
There is also a greater emphasis on reinforcement learning, with inductive learning approaches like model-free learning coming into play. A parallel training framework is enabling better learning in a multi-agent system. This will lead to the creation of truly impactful co-bots systems or collaborative robots. All of these developments will go into the complete transformation of the algorithm component.
Major technology behemoths such as Google, Facebook and Microsoft are already investing in the scalability of AI in diverse business sectors. By 2030, I predict there will not be any major vertical left untouched by AI. The technology is well on its way to deliver wider scope, be much faster and so inexpensive that it will become a part not just of large organizations but the daily lives of every common man. AI will be new mobile technology, all-pervasive and truly powerful. We will live in an AI-driven world, and I believe that organizations that start preparing for this transformation today are the ones that will thrive in the next decade.
The decade will truly belong to those who understand the importance of data, algorithms and computational architectures and can use the transformations in these spaces in truly effective ways. AI will transform various sectors and how industry leaders can prepare their enterprises for these innovations. Watch this segment.