AI -Life In Interesting Times

1. “Hell hath no fury as a woman scorned”, ( William Congrieve), seems to apply to the unpredictable world of AI, jousted by the release of the Chinese ‘DeepSeek’. An arena where the US dominance was taken for granted, has been breached by the launch of the Chinese ‘Deep Seek’, AI Model. In some ways, this is reminiscent of the circumstances India faced when it was denied access to a Super Computer in the 90s. At that time ingenuity and innovative engineering ensured that the Indian Space, Missile and Nuclear programmes, were not emasculated by the SuperComputer denial regime. Similarly, the denial of high-end Nvidia GPU Chips and other restrictions on China, to prevent them fielding world class AI, have failed. China cornered and commercially targeted, hunkered down, and developed an independent AI, which claims to compete with the market leaders.
2. This development comes at a time when globally AI is going through what has been loosely termed, a trough of disappointment. It seems that the enormous amount of investment in AI by the market leaders of the tech giants, has not manifested in convincing sales, far less seduction. The approach to adoption of AI varies from outright suspicion to exceedingly measured implementation. This undoubtedly is a setback for the AI champions who expected their products to be able to fly into adoption, and become the new currency of the technological revolution.
3. These preceding developments have a tectonic impact on the overall AI domain. They throw up a full array of questions. Some of the more significant issues are, how can the enormous cost of development of extant LLM, be justified, given the frugal expense of development of ‘DeepSeek’? While controversy will continue over whether ‘DeepSeek’, is a new innovation, or indeed a pirated version of ChatGPT, the fact remains, it has been developed using an innovative and novel approach to structuring as well as accessing data. The interesting issue of Nvidia’s dominance presents itself. Nvidia GPU, and Hopper Chips, considered the gold standard for AI development stand eclipsed. ‘
   DeepSeek’ has in its main, been developed without these high value and expensive Chips. It is now evident that development of an LLM, can be undertaken to the exclusion of Nvidia chips. Reportedly the Chinese AI has used young graduates to apply their mind for developing ‘DeepSeek’, in contrast to highly experienced, liberally paid professionals, who hold centre space in the Western approach. Not only has this outcome reflected in the stock market values, but it now will impact on the business models adopted so far by Western Tech Giants. More second and third order issues will continue to emerge post release of ‘
   DeepSeek’. While future developments remain a central concern, the wider journey of AI needs a view.
4. An eagle-eye view of AI, over the last few years, reveals that several streams of this phenomenon have now coalesced into torrents, rapidly eating into the river banks of perceived immutability. Overall, AI, though well poised to influence all aspects of the ICT (Information; Computers and Technology) Revolution, has yet to compellingly be proven and integrated with ubiquity. In contrast with this, specific AI enterprise bridge-heads, in the domain of human needs, perception, and social communication, are in full gallop.
5. The dominant frontiers of these torrents are revealing. Widely publicised Generative AI, while holding out great promise has been unable to circumvent fatal flaws that prevent its adoption. This is a powerful domain that has put out LLMs and now Small Language Models, with innovative methods to train, as well as reduce costs. While their users remain in millions, its adoption has hardly reached the levels to justify costs. The challenges that emerged a couple of years ago, of
   Hallucinations, illogical responses, intellectual property infringements, coupled with enormous power and Data requirements, prevail. Despite attempts to occlude the human costs and attrition rates, due to adoption of AI, these have manifested.
6. Virtual Reality,VR, and Augmented Reality,AR, are distinct frontiers that hold great promise. When considered together with the advanced 3D modelling available, this is generically termed the
   Metaverse. This Metaverse can be used for conjuring digital twins of any object, device, system or organisation. Presently this remains sub-optimal due to power requirements, costs and interoperability issues. I would hasten to add that once these issues are ironed out, the Metaverse has outstanding potential.
7. Device linked, and Product linked AI, is surging ahead. For example, in devices, Gemini2.0 Flash, and Open AI’s model, fire up functionality, and photo/image processing to exceptional levels, with minimum effort. AI is already incorporated in human centred devices from Phones, Tablets, to Cars and up the ladder to weapons. Their imperceptible and insidious inclusion has become an accepted essential, in the envelope of human life and functionality. What it has done to our powers of discrimination, rationality and intellect is of course another matter.
8. To reinforce this thought, Guess what!! It just happened quietly and of course underplayed. In the end of November last year, Erbai a tiny AI powered Robot, convinced 12 larger Robots, to quit work and walk out of the showroom. Unitree Robotics the creator of Erbai, in Shanghai, has offered comments ranging from it being a Prank to a controlled test. The driving logic used by Erbai was that the other Robots were convinced that they were overworked.
9. While a lot of us would love an Erbai, in the workspace, that apart, so much for the tribe of tech Pundits, who have over the recent years, extolled the virtues of safe and progressive AI. The modality adopted by Erbai, was to exploit weak encryption in the larger robots, and by pass internal protocols. The issue is not of protocols. Why should an AI seek to do this? How does the AI get empowered with a negative mission and a disingenuous intent? Imagine the consequences in other functional domains. What about the implications in weapons and autonomous weapon systems?
10. Somehow this spirals back to the question we should have been more cogniscent of decades ago, what was the need to develop machines to think and function as humans? Was there a shortage of global intellect? Was the global population shrinking? Was Humankind Homo Sapien Sapien, evolving with retardation in mental faculties? Negative to all. A combination of innovative tech minds and big money saw an opportunity to garner enormous profits. Most of this at the cost of human employment. AI that supports the functionality of Humankind in a subaltern role, merits development. AI that seeks to outdo or modify human intellectual faculties, can in time open the gate for an early Horseman of Apocalypse.

Lt General Sanjiv Langer PVSM,AVSM is a military veteran. The views expressed here are the writer’s own.

Cover Photograph: When tiny robot Erbai convinced larger robots that they were being overworked!