First Principles | AI and humans, well met

From writing political commentary to creating art, AI interventions have some interesting things to offer. And since its code is created by us humans, AI is – like we are — biased.

Once upon a time, writing political commentary, or creating art, for that matter, was a domain only humans could straddle. But it is time to rethink such notions. Art generated by Artificial Intelligence (AI) was first auctioned for over $400,000 back in 2018. And turns out, AI has now begun to make pretty good attempts to write political commentary (among other things) as well. For instance, consider a question such as this: “What may India’s electoral landscape look like after the General Elections in 2024?”

AI has now begun to make pretty good attempts to write political commentary (among other things) as well. (Thinkstock)

When some keywords such as the names of key characters and the theme was inputted into a trial edition of GPT-3, the third generation version, a Natural Language Processing (NLP) software, some interesting output emerged. Take a look yourself:

“With the general elections just a few months away, there is no doubt that a Nitish Kumar-led alliance will dominate the political scene in 2024. But one of the questions on everyone's mind is: How will such an alliance come together and fare after the elections?”

“There are no easy answers. The reason for this is that there are many factors that will shape the future of this coalition. As things are, while Nitish Kumar and Tejashwi Prasad Yadav have stated their intention to fight the BJP in 2024, they have still have not agreed on whether they should be part of a joint front with the Congress or form a separate opposition party and align with the Congress.”

“Another factor which could affect their relationship in the future are allegations of corruption against Yadav's family members who stand accused of having amassed wealth disproportionate to their known income sources. This has led Congress party leader Rahul Gandhi to distance himself from them. He had called for their resignation from public life following an investigation conducted by CBI into these allegations earlier in 2019.”

While what is reproduced here is an edited version that cleaned some grammatical errors, it is only a matter of time before “Deep Learning” software gets better. But that is beside the point. What is significant are the questions a report like this beggar: How are we to know if political commentary such as this is unbiased?

“Of course, AI can be biased!” explains the co-founder and lead engineer at a Bengaluru-based start-up that offers services to telecom companies. This, he says, is because the final output that emerges out of such AI-based software depends on the number of databases the software has access to, and the weights its creators assign to the datapoints it contains. This is where the nuances come in as well.

Those who assign weights to the inputs are humans such as him. And all humans have their biases. So, the weights one person assigns to a certain datapoint may be different from that of another who holds a different world view. When the weights are changed, it is possible that the same database can create an altogether different narrative.

This is also why he is unwilling to buy into the prevailing narrative that insists jobs that require human diligence, ingenuity and creativity are under threat. If anything, he argues, humans not just get more creative when challenged, but after AI interventions altogether new classes of jobs will emerge to suit human ingenuity.

My Bangalore techie friend’s world view is corroborated by Hari Menon, co-founder and CEO of Big Basket. In a conversation recorded earlier this year, Menon pointed out that those who worked to build the infrastructure that is the internet, never imagined that it will someday be embedded into cellular phones with maps and new businesses such as his that deliver groceries can emerge. For that matter, the early pioneers had not imagined taxi hailing applications such as Uber or restaurant aggregators such as Zomato and Swiggy. But all these entities, he says, have gone on to create new jobs no one had thought previously possible. And how are we to know what may emerge in the future?

All we need is an open mind. And not to forget that the people who write the code that powers AI are humans.