Tag Archives: data

Qwertyman No. 65: Who’s Afraid of Big Bad AI?

Posted on by
Reply

Qwertyman for Monday, October 30, 2023

I NO LONGER attend writers’ conferences and festivals that often, believing that younger writers would benefit more from each other’s companionship and encouragement, but I made an exception last week for the 66th Congress of the Philippine PEN, as a gesture of solidarity with that organization which has bravely fought to defend freedom of speech where it is threatened all over the world.

I was richly rewarded for my effort by listening to one of the most enlightening discussions of artificial intelligence (AI) that I’ve come across—not that there have been that many, considering that ChatGPT—widely regarded today as either God’s gift to humanity or the destroyer of civilizations—has been around for just a year. 

Of course, AI has been around for much longer than that. In pop culture, which has a deep memory for these things, we can’t help but think of HAL, the insubordinate computer in 2001: A Space Odyssey (which actually came out in 1968), said to be a clever play on “IBM,” just one letter to the right. Indeed the fear of technology—what some would call unbridled knowledge—has been around since Faust made his pact with Mephistopheles, reiterated in literature, film, and pop culture all the way to Dr. Strangelove and Spiderman’s Doc Ock. 

Not surprisingly, the panel on “The Filipino Writer and AI”—composed of Dominic Ligot, Clarissa Militante, Joselito D. Delos Reyes, and Aimee Morales, and moderated by Jenny Ortuoste—expressed many of the anxieties brought on by the entry of AI into the classroom, the workplace, and everyday life: plagiarism and the loss of originality, the loss of jobs, indeterminate authorship, and the lack of liability for AI-produced work. With Filipinos being the world’s top users of social media, AI’s centrality in our digital future can only be assured, like it or not, and for better or for worse.

So new has AI been to most people—and so rapidly pervasive—that most institutions from governments to universities have yet to formulate policies and regulations covering its use and abuse (the University of the Philippines has adopted an AI policy, mandating among others that all members of the academic community should be AI-literate, but it has yet to provide concrete guidelines on, say, evaluating and grading AI-assisted work).

Most revealing and thought-provoking were the remarks of Dominic Ligot, a data analyst, software developer, and data ethicist who brought up talking points that many of us miss in our usually dread-driven discussions of AI. I didn’t tape the session, but so sharp were Dominic’s observations that I can recall and share some of them here (employing an endangered resource in this human, memory).

Let’s not forget, Ligot said, that all AI works with (in the literary, journalistic, and academic sphere) is words. It may have a scary ability to amass, analyze, and re-integrate these words, but it lacks the other elements that contribute vitally to creativity: emotion, inspiration, insight (and, may I add, the power of abstraction, the kind of conceptual leap that, say, stimulated modernism in art and literature from the rubble of the First World War). 

Also, the more AI amasses, the more mediocre it tends to be, because it works with averages, and averaging averages degrades the quality of the end-product. Since the products AI spews out from the prompts it receives are plowed back into the system, this spitback reduces the originality of material even further. If it keeps feeding on itself, AI gets even dumber.

Ligot also reminded us that the suspicion and denigration with which AI has been met in the creative community is understandable, given our earlier responses to technology. When photography was invented, he said, painters didn’t see it as a potential competitor; it wasn’t art. 

What separates human creation from AI, he emphasized, could be summed up in three things: agency, inspiration, and liability. Humans make sense of their experience and act on their thoughts and feelings accordingly; they draw on their experience to imagine new possibilities. And lastly, we are liable for what we make, do, and say, for which we have both legal and moral responsibility, while an AI-driven car does not, even if kills someone on the road.

Dominic also noted the inevitable tensions between disciplines and industries that consider AI a helpful if not indispensable tool in, say, achieving more efficiency and economy, and those, like the creative arts, that may feel threatened by the options AI offers.

Overall, the discussion could offer no assurances either way that AI would leave us better or worse, but what it left me with was a desire to continue writing the way I do, while I can, before the algorithms take over everything.

And just out of curiosity, to see what AI can do now for and to the Filipino writer, I gave ChatGPT this prompt: “Write me a sentence in the style of Nick Joaquin about a man grieving over the loss of the woman he loved.” This is what the program came up with:

“In the solemn silence that engulfed his weary soul, he encountered the bitter truth that bereavement lingered like an everlasting shadow, forever haunting his heart with memories of a love that once bloomed in the garden of his existence, now reduced to a withered rose, its petals scattered across the desolate landscape of his grief-stricken days.”

“Solemn silence?” “Weary soul?” “Desolate landscape?” Maybe not bad for a freshman yet to learn about unlearning clichés, but definitely not Nick Joaquin. Of course, out of vanity, I had to ask ChatGPT to do the same thing “in the style of Jose Dalisay,” and it offered this:

“In the depths of his longing, his heart bled like fragmented brushstrokes on a forlorn canvas, for she was the vibrant palette that infused life into his desolate existence, now forever vanished like whispers carried away by a tormented wind.”

Just slightly better than Mang Nick’s AI ghost but still so not me. (Boy, does AI love adjectives!) All this says is that AI’s big mouth and stomach haven’t found and digested us yet, probably because we are not worthy. Cold comfort, but I’ll take it.

Penman No. 245: The Devil Is in the Data

Posted on by
Reply

movie-mistakes-5-the-matrix-1091436-twobyone

Penman for Monday, April 3, 2017

 

 

WE’RE OFTEN told that the devil is in the details, but today, you’ll likely find it in the data—that amorphous, often opaque stream of 1’s and 0’s whose infinite permutations define, describe, and direct our daily lives from our social-security pensions to our children’s grades.

Forty years ago, all you might have needed was an ID card, a driver’s license, and a passbook for bank deposits and withdrawals. You didn’t have a mess of numbers and certainly no passwords to memorize (and to lose). A signature was something you scribbled on paper with a pen, and an ID picture—invariably ugly and smelling of the sourish chemical bath it came out of—was a black-and-white mug shot the size of a postage stamp.

Speaking of which, if someone wanted to steal your “data,” they first had to steal an envelope with your name and address on it, and (if they wanted to keep the theft a secret) open the envelope gently with a cloud of steam. Postmen had to be waylaid, dogs euthanized, windows broken into, brass locks opened, and filing cabinets pried loose. Then, and only then, could the robbers get their gloved hands on your stock certificates, your paramour’s love letters, your UPCAT score, and your secret recipe for Lola Oryang’s Sinigang.

These thoughts crossed my mind last week as I sat in a committee meeting that pondered the impositions and implications of the Freedom of Information Act on government bureaucracies like our university’s. As old-guard civil libertarians, we all marched at one time or another for freedom and democracy, and have prized freedom of expression above all others, so it was probably safe to assume that we were all in support of freedom of information—the notion that citizens have the right of access to information from their governments, toward greater transparency and accountability, which presumably leads to better governance. (I’m not a lawyer, so this is my pedestrian’s appreciation of it.)

So the citizen asks a question, and the government provides the answer, right? Not so fast.  Like many good things, FOI isn’t as simple as it sounds, especially when you get down to the nitty-gritty of its implementation. This could have been why, despite calling for an FOI law for many years, Filipinos didn’t get one even under Noynoy Aquino, who promised it when he ran for president in 2009. It took President Duterte—in what I’ll hazard to be his most popular move among his opponents—to enact an FOI measure through Executive Order No. 2 shortly after he took office last year, since Congress, the keeper of many secrets, couldn’t pass an FOI bill.

The order mandated that “Every Filipino shall have access to information, official records, public records and to documents and papers pertaining to official acts, transactions or decisions, as well as to government research data used as basis for policy development,” with information referring to “records, documents, papers, reports, letters, contracts, minutes and transcripts of official meetings, maps, books, photographs, data, research materials, films, sound and video recording, magnetic or other tapes, electronic data, computer stored data, any other like or similar data or materials recorded, stored or archived in whatever format, whether offline or online, which are made, received, or kept in or under the control and custody of any government office pursuant to law, executive order, and rules and regulations or in connection with the performance or transaction of official business by any government office.”

FOI is supposed to work hand in glove with another initiative called Open Data, meant to be a repository of all the data requested from and proactively disclosed by the government, working through an eFOI platform where requests can be posted and processed.

That all sounds good if you’re going after an important or interesting document like a politician’s financial records, but again it’s not that simple. EO2 is accompanied by a list of nine major classes of exemptions, including sensitive personal information, trade secrets, and certain proceedings and investigations. (For all these documents, visit www.foi.gov.ph.)

In other words, and contrary to what some people might have naively expected, FOI is by no means an absolute right, bounded as it is by considerations such as those delineated by the Data Privacy Act of 2012. The DPA was designed to bring the Philippines and its booming BPO industry, which handles loads of personal data, in line with international privacy standards.

Between freedom of information and data privacy is a wide gray swath that’s still being demarcated by lawyers, ethicists, and the administrators who have to implement whatever the final rules are going to be.

In the university, we routinely receive scores of requests from, say, alumni looking for their old classmates and asking for their current addresses; unfortunately, even if we had them (which we often don’t), we can’t and won’t give them out. We can give you your transcript of records, but not someone else’s. These are the easier questions to deal with.

But what do you do when someone asks for the results of a publicly funded research project that could have commercial applications? How do you respond to a patently frivolous request for information (which, if properly framed and presented, the law still requires you to answer in some form within 15 days), like “How many blades of grass are there in UP Diliman?”

It’ll take a while to sort these issues out, which lie on the periphery of a larger and thornier ideological debate. We authors may believe our copyright to be sacrosanct, but there’s a whole “freedom of knowledge” school of thought out there (and even within academia) opposed to the idea of intellectual property and copyright, supported by groups that also advocate electronic civil disobedience.

“Is ‘We don’t know’ a valid answer to an FOI request?” I had to ask the people at our meeting, but I couldn’t get a categorical yes or no. Strangely enough, I found that ambiguity comforting, because it assured me that we still had a lot of thinking to do. Data is one thing, information another, and wisdom beyond legislation to establish.

freedomofinformation

(Image from techcrunch.com)