NEW DELHI: This year the Nobel Prize for Physics has gone to two pioneers of Artificial Intelligence (AI) -- John Hopfield and Geoffrey Hinton, respectively of Princeton and Toronto universities.
The prize has been awarded for the help rendered in the process of creating the building blocks of Machine Learning (ML) that would revolutionise the way work would be done in the future.
The Royal Swedish Academy of Sciences explained how the interconnected neural networks on which the two scientists worked tried to emulate the neurons in the human brain and how the experiment was already yielding results in improving productivity in our daily lives in general and advancing the cause of medicine, in particular.
Hinton, interestingly, also warned that AI -- which he compared with 'another Industrial Revolution' -- could produce unforeseen consequences creating a situation where 'things could get out of control'.
Hinton had an unconventional background as a psychologist who was always curious about knowing how the human mind worked. He had headed the AI division of Google in Toronto and worked on "training machines about how to learn by fine-tuning errors until they disappeared".
Hopfield on his part, once addressed the Princeton News Conference and shared concerns that "AI could create an apocalypse".
The two Nobel Prize winners thus highlighted the risks posed by the new level of technological advance that AI represented.
Incidentally, AI essentially rested on Computer Technology and the award of a traditional science prize to AI pioneers showed how boundaries between disciplines were fast disappearing.
Hopfield, however, did clarify that neural networks borrowed heavily from "condensed matter physics".
Hinton said while accepting the Nobel Prize that "whenever I want to know the answer to anything I just go and ask GPT-4" but he remarked at the same time that "I do not totally trust it because it can hallucinate" -- saying also with a spin that "almost on everything it is not a very good expert which is very useful".
Significantly, Hopfield is known for creating an associative memory that can store and reconstruct images and other types of patterns in the data.
The AI operations are thus governed by the ruthlessly fundamental principle of "garbage in garbage out" and this is a sobering thought in running them.
It is clear that three basic concepts are applied to all AI activities. First, the data on which they were based must be absolutely reliable. This can be deemed to be a thing common between the real-life working of an Intelligence agency and the sphere of Artificial Intelligence.
Secondly, the outcome of an AI exercise rests on 'Data Analytics' which unlike in the case of an Intelligence agency, does not have the benefit of the application of the human mind.
AI works on detecting 'patterns' through neural nets and language models and this sometimes can be the basis for producing a limited forecast about how an entity would behave in the future. This can be said to be emulating the 'modus operandi' method adopted by the Police in evaluating the next move of a criminal.
The third point that adds to the importance of the accuracy of data fed into the AI system is that it should minimise the possibilities of 'wild' readings being made in the future.
Reliable data will better enable Machine Learning to detect a 'mismatch' and seek to 'correct' it through the application of possibilities until the known correct solution is reached. At the end of it, however, this still highlights a 'limitation' rather than an assured 'gain' of AI.
An established use of AI for the advancement of business is its application in conjunction with Machine Learning -- in evolving Global Customer Centric solutions (GCCs). Here also data quality is emphasised and the promotion of 'data democracy' is talked about.
It requires a balancing of global and local priorities. Success here will be determined by the degree to which human readings on say cultural nuances were integrated with advanced technology that could be pressed into use for reading the evolving business trends and customer preferences through data analytics.
Tailoring the offerings to 'hyper personalisation', virtual try-on experiences for products, and multimodal customer interactions combining text, speech, and video, are the three major tasks that could be entrusted to AI and ML.
Data augmentation and analysis are the basic determinants of GCCs. It is a matter of great national satisfaction that India stands second in adopting Generative AI platforms after the US, excelling in content editing and productivity tools.
It is now established that AI is already helping businesses and other organisations to diversify their activities and enhance productivity.
It is also universally accepted that a combination of machine and human intelligence is the key to ensuring customer satisfaction and loyalty.
Understanding data -- not merely accessing it for analysis -- requires a deep understanding of human needs, cross-cultural subtleties, ethical considerations, regional factors, and timely cognisance of the evolving social and market trends.
We are in the age of knowledge-based decision-making and therefore information should be shared within the organisation subject to the needs of data security and role-based access. This will help the cause of data democracy and promote innovation and the generation of insights.
The importance of data safety and security cannot be overestimated, however, considering the reality of lawbreakers stepping up their criminal activities through manipulation of data, misinformation, and cyber frauds.
What has made the criminalisation of cyberspace a matter of national and international concern is that the dark net which is a hidden segment of the internet -- accessed only through special software and configurations -- has become a platform for illicit activities like the narcotics trade.
The promises and perils of AI are being discussed today at the same level of seriousness -- this buttresses the importance of human oversight over any kind of 'machine delivery'.