AI: deep learning vs shallow thinking

In his essay “Don’t believe the hype: the media are unwittingly selling us an AI fantasy” which appeared in the January 13, 2019 issue of the Guardian, Cambridge academician John Naughton writes that much of the narrative surrounding the benefits of ‘machine learning’ (one particular manifestation of the Artificial Intelligence technology) is pure nonsense, artfully crafted by tech giants that own and control the technology but faithfully parroted by mainstream media.

Naughton’s trenchant observations are on point when applied to the Indian context as can be seen from the spate of articles promoting AI that have regularly appeared in local newspapers penned by leading academic organisations such as the IITs and business/technology think tanks. Typically, these articles have focused on two buzz words — innovation and entrepreneurship — and conveniently ignored the many challenges posed by AI.

While some of these challenges are in the areas of democratic governance, security, privacy and discrimination, the most important issues that need to be addressed have much to do with employment and the ever-widening gap between the haves and the have-nots.

The word capitalism which traces its origin to the French word ‘capitalisme’ (the condition
of one who is rich) is conventionally defined as a socio-economic system based on private ownership of resources or capital with little or no oversight. To which we might add the exploitation of public and private spaces of citizens involving their physical possessions, private thoughts and modes of behaviour, and economic vulnerabilities.

The phrase ‘machine learning’ refers to computer programmes that enable computer networks to discover patterns hidden in huge amounts of text, audio and video data (so-called ‘Big Data’). Much of this data, of course, is provided gratis by web and smartphone users to IT behemoths such as Google, Facebook and Amazon who use the data for marketing and advertising purposes and generate huge profits, none of which is shared with the user community.

The latest entrants to this club of IT giants, Uber, the networked transportation company and Airbnb, the networked accommodation company, don’t own any vehicles or real estate but nevertheless profit on every transaction involving their subscribers’ vehicles and homes.

In the European Union, the General Data Protection Regulation (GDPR) gives people the right to know on what basis computer programmes make decisions that affect their future. In India, while there has been a justifiable emphasis on the right to privacy, especially in the web environment, little attention has been paid to the AI software whose harmful effects on employment, economy and body politic are rarely touted.

Take, for example, the impact on auto-rickshaw drivers by online wireless communication-
enabled transportation and delivery networks. That these drivers now must share a portion of their already meagre incomes with faceless companies such as Uber, Ola and Zomato or the fact that they can be unilaterally dropped from their subscriber base for a variety of discriminatory reasons such as political, caste or religious affiliations is never mentioned by the media.

Much of the recent advances in AI employ ‘deep learning’ algorithms which analyse vast amounts of real-time cognitive and emotional data to generate plausible connections, make predictions on the future behaviour of humans and machines alike. These algorithms take decisions without human intervention and provide no explanation whatsoever for why or how any decision was reached.

More importantly, given the highly dynamic nature of both the algorithms and the data being analysed, is any explanation even possible if something does go wrong? Especially in a smart city environment where smart homes and smart transportation systems coexist with gullible citizens, politicians and regulators hampered by little knowledge on how smart technology works or how global technology companies operate.

By dangling huge investments and the promise of jobs, tech companies have succeeded in co-opting legislators into granting them extremely favourable tax concessions and choice real estate to set up operations while, at the same time, defanging any meaningful regulations on issues such as data privacy, monopoly power, unionisation and job security.

The recent episode involving Amazon’s backing out of a pledge to locate a very large warehouse with significant employment opportunities in New York City even after being offered over two billion dollars in tax concessions should serve as an eye-opener. Not so well publicised is the fact that Amazon is at the forefront of rapid introduction of AI-based robotic and drone technologies in product warehousing and delivery operations. Perhaps there is a ‘deep’ at the end of the AI tunnel?

(The writer is a computer science and IT specialist)