- >>
- OPINION
Artificial Intelligence technology is apparently changing by the week. But a dispassionate view indicates it mutates by the day. The apex forces at work are profit, market capture and that which competitive business does best, hollowing out or killing competition. Unlike other products and offerings, AI is focused on the Human brain, intellect and all dimensions of sentient behaviour.
AI is gradually coalescing into a quad of generative AI, social media-centred, product and domain-specific, and human health-centred development. The first, while claiming to put at the service of humanity the wisdom of centuries, in effect replaces or seeks to enhance human intellectual functionality.
The second, while powering all forms of interactive social media, has successfully lent itself to disinformation as well as, influencing opinion. The third segment is part of a mushrooming world of specific AI, that intends to take functionality of products into a distinctive domain, at warp speed.
From mobile phones to driverless cars, kitchen appliances to friendly robots and positively unfriendly weapon systems, it is an interminably expanding domain. AI in human health, while devoting effort to the needs of the impaired and injured, as well as qualitatively upgrading the medical milieu, is also stealthily progressing the frontiers of ‘immortality’.
Manifesting as it does today, AI is a result of the pursuit of designing an electronic human-modelled brain. The designing of Neural Nets commenced in the 1940s. The present Revolution however is a consequence of the Deep Learning breakthrough, of 2012.
It was discovered that the most critical Training of AI Models could be achieved with a different chip. These are the Graphic Processing Unit chips, GPUs earlier used only for Gaming. Deep is an expression that signifies that the learning is structured in several layers of the electronic brain.
In 2012 Alex Net, came out with a model of 8 layers, by 2015, it was convincingly overtaken by Image Net, with 100 layers. These models gave the ability for AI to be trained on enormous volumes of data as well as its own experience, Machine Learning.
A human brain averages about 100 billion neurons with each neuron interconnected to between 1000 to 10,000 neurons. The biological brain apart from thousands of years of evolution, also has a unique quality of neuroplasty.
The AI brain on the contrary is a mathematical entity, be it algorithms or circuits/neural nets. All its responses are mathematical representations. There is a detailed set of criteria, weights, probabilities, probabilistic optimization and randomness, that are put into play to produce high-order AI.
This, of course, is most evident in Generative AI products such as Chat GPT, Gemini, Claude, Llama, and others. Unable to match the qualitative genius of the Human Brain, AI seeks to surpass biological brains quantitatively. Gen AI can store much more info than a human brain and can respond at up to trillions of bits per second, surpassing any brain.
Couple with this the factors of human fatigue, inaccuracies, speeds of realisation, as well as plain laziness. It is consequently unsurprising that AI can attempt to work around higher human functions.
Before entering into the dizzy world of big tech and AI financial value, we must be clear on its business model. Gen AI and Social Media, which are trendsetters, are based on an ‘engagement-based revenue structure.’ The more popular and potentially addictive a Service is, the more money it can make.
Engagement and length of Engagement are both financial drivers. Attention is the currency that drives their economic model. Commencing at the apex, the five biggest global tech companies collectively have about 570 billion $ in cash.
In perspective, those that have global AI ambitions including some of the above, have about 630 billion $ in cash and investments. Present spending on AI is set to cross $ 500 Billion in the next three years, 30% of which will be Gen AI. The aggressive push for AI in its full spectrum, is consequently poised to dominate markets.
A view on the chip that made it all possible is imperative. Nvidia, the company that has just roared past its 2 trillion $ mark, has provided the heartbeat of the AI revolution up until now. Its H100 GPUs called Hoppers priced at 30 to 45,000 $ each have been irreplaceable for training AI Models, a function critical to the market fielding of these models.
Nvidia also provides a software platform, where customers can tailor these chips to their needs. As competitors feverishly race to develop alternate cheaper Chips and procedures, Nvidia has now fielded Blackwell B-200 GPUs with 4 times faster outcomes, and 40 -60% higher cost.
There seems to be a period of serious competition ahead for Nvidia since Google and Anthropic have developed their in-house chips, and companies like Ace and Mistral are very close to internally optimised AI options.
The entire AI ecosystem likes to use the term Stack. The Stack is representative of the entire domain from the Chip up to the vast Data Centers euphemistically called Cloud. In slow but definitive moves, it seems evident that the big Tech Companies and the AI giants are ideally poised to take over all elements of the Stack. All this seems to present a clear trajectory. Ironically this is not so.
Gen AI, the most talked about and touted of all the AI segments, is far from ubiquitous. Despite the enormous hype and tremendous hard sell, the majority of the business world is tentative in its approach and cautious in its evaluation.
There has also been a notable backlash to the possible future frontiers of AI. While there is a flurry of judicial pronouncements on AI, the EU AI Act is the first definitive regulation on AI.
The Indian and US initiatives in this direction are well underway, and in the interim India has also clearly specified an advisory for AI. Article 5 of the EU AI Act, is definitive in that it prohibits AI systems that deploy subliminal technologies beyond a person’s consciousness, with the objective of materially distorting the behaviour of a person or a group of persons, impairing their ability to make an informed decision.
It is distinctly apparent, that despite its powerful push, AI, significantly Generative AI and Social Media will have to confront the realities of National lawmakers and their sensitivities to the emerging contours of AI.
Product-related AI needs to negotiate the specific working domain of their functionality. Here there is a far more direct relationship between the client needs and what AI can offer. Undoubtedly driven by comfort, convenience and pragmatic needs, they also pry on us based on their seduction.
In the medical domain, a lot can be said in praise of advanced robotic surgical procedures, cutting-edge diagnostics and finely titrated drug development that are benefiting from AI. The recent Neuralink wireless chip cranial implant, which is successfully giving wireless internet access to a physically impaired patient, is commendable.
It comes of course with its package of concerns, on hacking, and privacy. The reported use of AI in the formulation of advanced medications promises to substantially enhance pharmaceutical options.
On balance, while it is a reality that Lawmakers across the globe are wary and exceedingly conscious of the potential of AI, they face a daunting challenge. AI due to its rate of development and mutation is exceedingly difficult to define, ring fence, and hold accountable.
There is an enormous draw towards, as well as a huge curated appeal to the populace. Laws and advisories run the risk of stifling an industry that is financially significant, as well as infringing privacy and personal rights.
AI captains are well aware of this and have mastered the ability to position products in ‘deep grey seas and amid swirling currents’. Ultimately, it is the individual sentient and rational intellect which will be the final frontier for AI.
Lt Gen Sanjiv Langer PVSM, AVSM, is a former Member Armed Forces Tribunal, and former Deputy Chief Integrated Defence Staff. Views expressed here are the writer's own.
