Tuesday, July 12, 2016

Meet the New Boss

I've always felt that the superhero policy of not killing anybody sacrifices pragmatism for nobility. Suspiciously, there is no capital punishment in their world either. If they don’t want to play god then that’s fine but surely they could invite Thor over and let him decide. Character transformation is wistful thinking; Poison Ivy hasn't turned over a new leaf, the idea of a labile Lex Luthor is laughable, and the lip on the Hobgoblin isn’t wobbling. Even basic imprisonment would have kept them in check if they could just stop trying to escape all the time.




In such circumstances, the only scenario which promises eternal peace would involve our superheroes riding a never-ending streak of victories over evil. Streaks, in my own observations and also corroborated by any person who has ever been rudely escorted off a sporting pitch after entering it in the nude, are never never-ending. To put never ending in perspective, some of the things that aren’t never ending include time, space, and the girth on yo mama. Safe to say, it’s extremely unlikely that this streak will outdo all of those. Therefore, by playing along, our heroes aren't really saving the world but much like the owner of the bar where a lucrative The Doors fans meet up is happening, just delaying the end.

At this point, the wise might try to change the course of this conversation and suggest that the streak doesn’t matter. If the streak ends we would probably all end along with it and assuming our posthumous concerns and priorities are unrelated to our prethumous ones, why bother. “Does an exploding planet make a sound if there’s nobody there to hear it?” they might ask to drive home their point. To that particular question, you should respond with a firm no. Sound needs matter to propagate and outer space is almost all vacuum so no sound even if someone was there. It would be a great visual spectacle though, visible from a galaxy far far away, long after.

In return, ask them if there were a million
 physics text books in the world, but none
 of them within your reach from your arm
 chair, would the laws of physics still hold?

Image source:  Wikipedia

Ignore the philosopher and look at this issue again with your economics hat on. If you can ignore how silly you look for a few minutes, then, unlike in economics, a believable, testable, theory emerges. Consider the situation where the superheroes finish off all evil and the only complaints the police get any more are from elderly people reporting loud music after 10 pm. The superheroes would be jumping ship from the dying Saving The World industry. They would have plenty of opportunities - Flash could become a delivery boy executive, Wolverine could be the guy who puts the tears in fashionable jeans, Spiderman a coconut picker, Green Lantern (the one from the Justice League cartoon, not Ryan Reynolds) a protest poster boy because he was a victim of police brutality for singing along to Janie's Got a Gun loudly after 10 pm, and so on, but these professions are neither as fun nor as glamorous as their previous jobs. Now consider another scenario where they collectively agree to let the evil live and get their PR team to put out the word that killing is unethical which is why they are sparing the villains. The simple economic truth driving their actions is that the demand for superheroes falls if the supply of evil is weak. If you think this is bad, imagine what will happen when the villains figure this out for themselves.

Captain Planet, fighting pollution and then telecasting
it on what are most likely coal powered TV sets

Image source: Cartoon Brew

Outside of ComicCon, equality, rights, freedoms, free markets, and all their sidekicks, rip-offs, and franchisees were the superheroes supposed to keep us safe and prosperous. In some parts of the world you might still get to hear the, “Is it a bird? Is it a plane? It’s democracy!” routine. These systems helped vanquish some communism, some autocrats, some walls, some extremism, and made for some great movies and books in the process. What it didn’t do is wipe them out completely and it also made for lots of really bad movies. If you actually tally them up, you will find that like an inefficient Game of Thrones character we have eradicated more things that we didn’t want to eradicate, such as biodiversity and cultures, than the ones we wanted to, I can only think of smallpox.

One can't blame the people. Apart from the fact that matters of politics and economy are complex, all the grown up stuff is placed deep inside the newspapers where we can never get to them. Experts and thought leaders and other occupants of the featured speakers page, who are paid anywhere between vast sums of money and enormous amounts of money to make sense of these things for us, get it wrong almost all the time without any loss of reputation, except in the rare case where the mistake could be traced back to an Excel miscalculation and not the economists themselves. Globalisation has brought Haagen Dazs to the country and I hope to be able to afford it one day but it has also brought about a multifold increase in complexity to most geo-political phenomena. In the current age, most people probably don’t even understand the implications of election manifestos as a result of which they default to voting by caste or religion, an act which would be considered illegal if it were any other job vacancy being filled.

The world moves so fast nowadays that between the time the
 newspaper is printed and  you get to spread it in the comfort
 of your toilet, 58% of the news is already outdated.*

Image source: The Guardian

The bitter truth is that we are much farther along the streak than our superheroes and our institutions are now falling. There were some minor cracks here and there to begin with, the sort of thing that Berlusconi manages to slip through. We should have fixed it then itself but we didn't bother. All the neglect over time has had them decayand hijacked to the point where their sole purposes seems to be to provide covers for Narendra Modis, Rahul Gandhis, Drumpfs, all those other extreme right politicians in Europe, too-big-too-fail banks, reliance on fossil fuels, and anything else you think is wrong with the world. Technology, the golden haired, saviour child, is already being called a bubble, no disrespect to the Powerpuff girl. There are no institutions left to save us. They either died a hero or lived long enough to become the villain.

In the meantime, machine learning algorithms have acquired the intelligence required to beat an Asian guy at an incredibly complex game of strategy. One can safely assume that running government would be a piece of cake. Some of you are currently jumping ahead of yourselves and drawing parallels to Seldon’s work on Psychohistory. The distinction would be that this system could actually run government instead of being some secretive oracle only useful for nicking money off of increasingly annoyed bookies. As of now, comparisons with The Matrix are also unfair.

I take this opportunity to announce that I am going to start work on this project. I’m calling it AlphaGovernment. It is a collaborative project which means that most of the work will be done by other people but I’ll be the face of it. I'm looking for co-founders who don't know much and are named along the lines of Dipanshu, Dipika, Deepak, etc. That way we can claim that Deep learning is critical to our success and garner some media attention. Nobody wants to cover a group of people who know what they are doing and solving a problem with some simple regressions.

Deepak Chopra, please do not contact me.

Image source: Wikpedia

The first hurdle would be the collection of data. Archives from the press and social media would serve handsomely. We can further enrich this corpus by adding the scripts of political dramas like West Wing, House of Cards, Yes Minister, The Thick of It, etc. Bias should get neutralised by combining multiple sources. Tracking particular words and phrases across time would provide a diverse set of circumstances-action-outcome trajectories which would form the foundation for our model. We would also need to watch out for all the false links, for instance all the reports of Salman Khan’s hit and run case over multiple years could get linked to his subsequent appointment as Olympic Ambassador because the wording might confuse the model into thinking that Salman Khan is some sort of retired boxer-athlete who was very motivated in his youth.

The second hurdle would then be to define the objective to be attained so that the model can try and optimise for it. A thorough reading and listening of all the great speeches throughout history would give us the base concepts. A word association exercise will help us connect any terms related to any of these concepts. We can then look for these terms against media coverage of events and calibrate the impact of an incident to the sentiment rating of the media coverage, and also get a relative importance of each concept. Petition response data from change.org can improve the results further. This would ensure that the objectives are dictated by actual data generated by people instead of the personal beliefs and biases of the coders. The results might state that a ban on surge pricing or enforcing net neutrality is more important than instituting poverty alleviation measures but that's democracy.

To give you an idea of the existing body of work on this topic,
a summary of the first page has three Quora links, one Yahoo
Answers link, and five other random forums. The one UN
link points to page for a relief agency for Palestine refugees.
Related searches includes 'goal of human resources'.

Every optimisation needs to work within certain constraints and that would be the third and final problem. Most constraints would be formulations of not pissing off people enough to lead to riots. If the only feasible solution did involve pissing off a group to the point of riots then the system would have to figure out a way to make it contribute to the overall objective. Maybe the system could launch a scheme which would reward rioters with loyalty points if they chose to direct their ire only towards illegal constructions and vehicles with severely overdue pollution checks. These loyalty points would be redeemable at the nearest electronic stores for old stock, which is what the rioters were probably expecting to get out of their rioting anyway. The cooperation of the electronic stores would be borne out of the gratitude of not having been looted and ransacked and now they have freed up space for new inventory and earned public goodwill for helping quell the riots. Everybody wins.

Of course, there still are some wrinkles to our grasp of the technology and there will be a teething period during which the darker sections of society get misidentified as gorillas and banana prices get assigned a high importance even outside of banana republics, or the system sends brochures about pregnancy health care to teens whose parents don’t yet know she’s pregnant. The good part is that unlike current demagogic politicians, the system will actually learn from these mistakes and make corrections to how it operates. In the words of my grandma -


General artificial intelligence,
despite all our efforts and diligence,
Will only happen after,
we refine and master,
The more tricky artificial common sense

The future of this system will strongly benefit from the Internet of Things pervading our daily lives. With this in mind, I'm also going to start a company that makes SmartPeople(c). The company will be called Speculate, Create, Use, and Update Laboratories (SCUUL). This is not a collaborative project and all profits will be mine. If you're interested in investing, please get in touch.

SmartPeople, like an Internet of Things thing, are conveniently pre-fit with a bunch of sensors capable of taking input or providing output, and all over the air. From a young age, SmartPeople(c) will be assigned to SCUUL servers to train their own models. As the SmartPerson(c) learns and is able to successfully demonstrate it, SCUUL will keep reassigning them to more advanced servers capable of handling higher volumes and more diverse sets of data. At some point the SmartPerson is expected to show a strong interest in a particular field which is when they get attached to a SCUUL servers with a higher degree of knowledge in those fields. After the SmartPerson's models have been trained by these specialised servers as well, they can be deployed to the field where they will link up with AlphaGovernment to pass on data to it and also receive data and instructions from it to act upon.

Once there are enough SmartPeople supplying data, and the model has stabilised and reached a certain maturity, the need for human moderation of the model will also be eliminated. AlphaGovernment will become an incorruptible, fully autonomous czar which we can delegate governance to while we go back to doing the things we are best at.

Like planking.

Image source: Wikipedia

* This number has probably dipped further between the time I wrote it and you read it. 

No comments:

Post a Comment