Electric Dreams: The Reality that Bytes
Fan’s of the BBC series Yes Prime Minister may remember the episode titled Power to the People. It introduced the fictional concept of the Politician’s Syllogism when compelled to action, a logical fallacy that goes like this. We Must do Something. This is Something. We must do This.
In the post-Berlin Wall era, which also saw the dawn of the climate crisis Western governments, having prevailed in the cold war and become master of the world under US leadership, went on a relentless pursuit of technology, not to outcompete a rival power, but for the sake of technology itself.
New challenges were perceived for attack, and previously impossible technological solutions presented themselves as the answer.
So were born extraordinary surveillance tools, data centres of hitherto unimagined capacity recording every data bit of existent information, and software to query and crosslink all the surveillance systems and the data they collected.
Electric vehicles were reborn a century after they went extinct, to solve the new crisis of climate change and to move the world into a future where Oil rich States could no longer disturb global stability and the new power structure.
Self-driving automobiles, until now an idea in science fiction, became the desirable future, even as electronics took over virtually every interface between man and machine. Cars fitted with screens, menus, and built in sims could continually communicate with their manufacturers, report on movement and be disabled at will by the manufacturer or the authorities.
Stealth aircraft and then autonomous pilotless aircraft, or drones and other forms of “smart weapons” overwhelmingly became the focus of cutting-edge military spending.
The driving force was the Politician’s syllogism. Technology offered itself to politicians, to bureaucrats, to corporate boards, and to Venture Capital investors as the “Something” in “We must do something, This is Something, We must do this.” AI is one such manifestation
The enormity of the AI bubble can scarcely be exaggerated. The investments are staggering, the stakes for the governments of the great powers, are colossal, the commitment of businesses to the technology is a tide, the industry’s claims upon resources, from chips to energy are bordering totalitarian.
These expectations will probably not be met. Because what they’re calling AI is not AI at all. It is not intelligent. The stated goal of all AI companies, from Open AI to Microsoft, has been the development towards Artificial General Intelligence (AGI). AI, as offered to users, was touted as the stepping stone towards AGI.
AGI is the standard of machine intelligence that would approach things as a human being would without being trained for the task with all relevant data. The difference is that AI, as presently manifested isn’t really intelligence at all. It is intelligence in the sense that secretive espionage agencies refer to information. AI, as at present, is simple an incomprehensibly vast database and extremely sophisticated algorithms for querying, processing and presenting data.
There is no intelligence there. It is just dumb algorithms processing vast chunks of data. The importance of AI has been such that building the data sets to “train” the algorithms ignores all intellectual property and privacy rights. All books, all art, all films, all photos, all communications, all human creation ever that was technologically accessible and possible to be digitised was taken by the AI companies to train their algorithms.
Unfortunately, they are running out of data to feed the AI models. It turns out that constantly compounding computing power does not make AI better. The promise of AGI that was dangled in front of everyone from investors to officials, a promise which led to such extraordinary commitments of capital and by government, has been quietly dropped. The industry has in fact stopped talking about AGI and it isn’t coming any time soon.
This should have been obvious from the beginning to anyone with normal intelligence, let alone to the brilliant men at the helm of affairs in the worlds of finance, computing, and government. But it wasn’t. It echoes exactly how the promises of all electric car future were never going to be fulfilled, and how that fact should have been obvious right from the start.
Electric Vehicles serve as a good illustration of the intellectual breakdown of government and capital that is centred around technology. The most elemental, the so called “back of the envelope” calculations would have revealed 30 years ago that EVs were not going to be viable for a variety of reasons.
To begin with, a simple calculation of the quantity of gigawatt hours of electric production required to replace fossil fuel powered cars on any meaningful time scale, in comparison to extant capacity, would yield jaw dropping results. Where was all that electricity to come from? Where was the grid capacity to transmit all that power? Where would the capital investments come from? Nothing in the power production and distribution plans of any country, except China, indicated activity to gear up for EVs.
If the purpose of EV was care of the environment, how were those requirements to be met without building huge new coal, nuclear, gas, hydro and renewables capacity? The environmental costs of doing that would not be offset by the gains of ditching fossil fuels. What about the soil and water pollution from all the mining, refining, and processing toxic minerals for exotic battery chemistry? What about the huge energy demands for all the aluminium that would be needed?
Further, how would world peace and stability be helped by shifting wealth from unstable oil states to unstable mineral rich countries caught up in civil war? And what would all that oil sitting around do when it becomes dirt cheap to be picked up by those not subjecting themselves to the costs of the EV revolution? What if the daddy of technology decided to go back to fossil fuels as the US seems to be currently doing.
The automobile industry runs on the residual value of vehicles. The demand on the factories comes from people being able to apply the residual value of their vehicle towards the purchase of a new one. Automobiles are the third largest expense for most people after housing and education, and as such represents a huge portion of the consumer economy. It is this, the residual value of a vehicle, every time it changes hands, that enables the transactions to happen, and keeps assembly lines running. Banks, insurance companies, manufacturers managing production and sales, all know that a fossil fuel car takes a 25-year journey from new sale price to zero residual value.
But when it came to EVs, the world collectively forgot to think about what happens when an electric car depreciates to zero after 10 years because it needs a new battery which costs far too much to make replacement feasible.
EVs were an irresistible force for 30 years, dominating government thinking, expectations, policy purpose, investment, and “the future”. Reality began to bite in the last year or so and governments started suspending “fossil fuel cars banned by 2035” type policies. Manufacturers committed to phasing out their fossil fuel engines reversed direction in the face of catastrophic losses.
AI may turn out to be a similar chimera. Huge promises made of a future that would never be and which should have been obvious right from the start. The fundamental question that should have been asked was this: Have scientists unlocked the nature of intelligence? Have they figured out cognition? Have they understood how consciousness, emotion, judgement, lived experience, pain, pleasure function? Have they understood how the human brain accesses, retrieves, and processes information? How does it do so with such extraordinary energy efficiency?
The answer was no. It is still no, and because of that, all the big companies have is huge data sets, sophisticated querying, processing and reporting algorithms, but little more.
AI is an extraordinarily powerful tool, and extraordinarily useful. The ability to apply raw and all published learning on the data to tasks such as medical diagnosis, and legal research are potent tools. Search algorithms find result that would take humans millions of hours and find results that perhaps humans never would have. The solution to protein folding problems that won a Nobel prize two years ago is one such example.
But computers and robots that can act in our stead? That remains a fantasy.
B R Singh is a retired IAS officer who served in the J&K cadre.
