19 April,2023 07:21 AM IST | Mumbai | Mayank Shekhar
The memory ChatGPT relies on is merely knowledge. Representation pic
Better-off Indians must get walloped more by ChatGPT first. For, essentially, what is ChatGPT? An artificial-intelligence (AI), conversational platform, by a relatively little-known tech company, OpenAI.
You prompt questions to ChatGPT. It reveals answers aimed to satisfy the human user. Given the insane amount of data fed into it. Which it improves upon, with progressive human feedback.
We are on the relatively primitive ChatGPT 4, at the moment. It is to language-based memory, what the calculator is to basic arithmetic.
Technically, rendering it redundant for you in the future to mug up the replicable, in school/college, and effectively reproduce that same information, with little use in real life. Nobody remembers what they vomited to top that exam anyway!
Yet, this is how Indians like me, more so in the liberal arts, but also in the sciences, have done well for ourselves. Sailing through an examination, instead of an education, system - that's produced bureaucrats/babus, CEOs, back-end desk-workers, and spelling-bee champions. But hardly as many original thinkers, tinkerers, inventors, discoverers.
Also Read: Why's the writing not on the wall?
It's not like all school toppers move up a salary stream still. The white-collar jobs that follow, regardless of what you specifically do, chiefly equal receiving and responding to long emails, drawing/cracking presentations, etc (no, I didn't say being a bakchod; you thought it!).
This is where you'll often find that the lesser-off Indians inevitably draw a blank. Because of their lack of an equally easy facility with English as a business language.
It's not that ChatGPT, or any other language model, is designed to improve your interpersonal skills - that's just your emotional intelligence. It's a natural human advantage that technology itself might eventually fall short of.
What we did learn through the pandemic-induced lockdown, however, is that you only can't do without actual, physical interactions, when it comes to making love, or mortal combat! Everything else, between two people, can take place remotely.
What if the guy at the other end is dramatically skilled at scripting strong prompts, after prompts, back and forth, for that's how ChatGPT works - to help draft mail for a mail, presentation for a presentation, an essay for an essay!
He'd have sufficiently made up for a skill he could've never acquired, and concentrate on applying what he's better at. ChatGPT being his ally/assistant. It's not always right. But it's learning.
And this, in turn, also requires a new human skill in town, isn't it? Which is true for the future anyway. World Economic Forum, for instance, predicted 85 million jobs displaced by 2025, due to AI; but 97 million unheard-of jobs added, at the same time. Can't take these stats too seriously. One head can barely fathom what the billion-headed AI is likely to radically unleash over the decade.
Sure, there are necessary concerns over a tool becoming conscious, or getting directed towards devilish deeds, under devious control. Or what if ChatGPT and its big-tech variants (Bard, Bling, etc) is just the latest fad - Alexa/Siri have probably scored high on the Turing Test (of a machine replicating humans), already; and a Google search is more accurate, anyway? Maybe.
But, like most of you, I've been quite obsessed with ChatGPT, lately - in the way that it got rolled out for universal access, making it the technology with the fastest rate of adoption in history!
That history, as I see it, has basically favoured humans being âdisciplined' - early to bed, early to rise, logging in, logging out, clocking in, clocking out, between conveyor belts of mindless meetings and trail mails.
More or less modelled on the soldier as a replaceable robot, dressed uniformly (olive-green, or office-wear) - so you can't tell one from the other, as they go daily to war, eat, sleep, repeat!
An industrious man is meant to be a machine. What if machines effectively fake humans? As you can tell from ChatGPT, which is really Early AI, capable of behaving like the disciplined man you've met - diligently throwing at you expected answers, for different questions. Won't the real machine then beat the fake machine? The boss will be happy!
Further, if you ask, it can write a song for you like John Lennon, or a sonnet like Shakespeare - much to our collective shock and amusement. But that's the thing about ChatGPT - it can only be imitative.
The memory it relies on is merely knowledge. How we use that memory + knowledge to produce something our own, involves imagination. How we achieve that is by mining the unique, hence human, about us - rather than encouraged to be uniform (like all humans). Lennon, Shakespeare, for their style, ensured no song, sonnet simply replicated the other.
I asked ChatGPT to write me a column, of this same length, on ChatGPT. It reads like a textbook. I asked it to write a film review - any film, then a Bollywood film - like I would. It would've gleaned through data available on the Internet for almost two decades.
Devoid of a personal touch/experience, just knowledge - it came back with what's barely a Wikipedia entry. "(Fictitious) film: Masterpiece. Rating: 4/5. Masterpiece truly lives up to its nameâ¦" "Film: Gully Boy. Rating: 4.5/5. Gully Boy is a film that captures spirit of Mumbai's underground rap scene with authenticity, flairâ¦" I feel safe (for now).
Mayank Shekhar attempts to make sense of mass culture. He tweets @mayankw14
Send your feedback to mailbag@mid-day.com
The views expressed in this column are the individual's and don't represent those of the paper.