AI Grief Observed

These remarks were delivered this evening at the Creatively Critical Tech Speaker Series at Illinois State University.

"There is no good way to say this."

These are the opening words of Yiyun Li’s latest book Things in Nature Only Grow about life after the death by suicide of both of her sons.

"There is no good way to say this." My heart goes out to you if you too have had this sentence spoken to you. "There is no good way to say this" is a sentence always followed by very bad news.

(It is, I recognize, an unsurprising way to start a talk by yours truly, someone who has made a career out of describing education technology as very bad news. "There is no good way to say this." It's also an admission on my part that what I want to talk about tonight are thoughts that are quite tentative, quite tender. My husband asked me, "is it a good talk?" And I had to say, "I don't know!")

Let me read the first few paragraphs of Li's memoir, more than just that first sentence, in part because it is a radical radical book on death and endurance and acceptance (and typically, I think, we see "acceptance" as the antithesis of "radical." As complacence, as surrender).

There is no good way to say this — when the police arrive, they inevitably preface the bad news with that sentence, as though their presence had not been ominous enough. The first time I heard the line, I knew already what was about to be conveyed. Nevertheless, I paid attention to how the news was delivered: the detective insisted that I take a seat first. I sat down at the dinner table, and he moved another chair to the right distance and sat down himself. No doubt he was following protocol, and yet the sentence — there is no good way to say this — struck me as both accurate and effective. It must be a sentence that, though nearly a cliché, is not often used in daily conversation; its precision has stayed with me.

The second time, having guessed the news about to be delivered, I did not give the sentence a moment’s thought. I did not wait for the detective to ask me to sit down, either. I indicated a chair where my husband should sit and took the other chair in the living room. My heart already began to feel that sensation for which there is no name. Call it aching, call it wrenching, call it shattering, but they are all wrong words, useless in their familiarity. This time, the four policemen all stood.”

"There is no good way to say this." There is no easy way to talk about this. There are acceptable words, I suppose, but they are never "good," never remotely satisfying or comforting -- not to say, not to hear.

By "this," I mean death obviously. By "this," I also mean other traumas, other endings. By "this," I mean what might feel like or look like the end of education – an end not spoken about with the solemnity of the policemen but rather with a real jubilation from technologists and venture capitalists, who gloat about disruption.

I want to start here – by “here,” I mean the recognition that there is no good way to talk about death, no good way to talk about grief, even though I am going to try very hard to do so: to talk about grief – mine, yours, students’, teachers' – and tie it to “artificial intelligence.” I want to talk about grief and “the end." I want to talk about the end of the world – I don't, really; I want to talk about what feels like the end of the world and what might be, should we continue to build data centers, invest in this rapacious technology, and ignore climate change, literally the end of the world; I want to talk about the destruction of the future (our own, our children's), about the end of democracy, the end of education.

I want to talk about loss. A loss that is, perhaps, an abandonment. Perhaps an abdication. An absence. An erasure. A trauma. Death, mass death -- literal and metaphorical.

“There is no good way to say this.” I have read a lot of memoirs about dying and about grieving. Joan Didion’s The Year of Magical Thinking, of course. Geraldine Brooks’ Memorial Days. C. S. Lewis’s A Grief Observed (a phrase I’ve borrowed for the title of this talk). Michelle Zauner’s Crying in H Mart. Helen Macdonald’s H is for Hawk. I could go on and list so many wonderful, painful books. And yet, despite some of the greatest writers having tried, “there is no good way to say this” -- I know this. I know this intimately. Yet I still search for some good words to have been said, to have been written. Words to comfort. Words to find meaning. Words to make sense. Words to not feel so utterly alone, at the abyss abyssmal, because those we love most have left us, and the future we thought we would share is gone too.

“There is no good way to say this," the police told Yiyun Li. I don’t think that the coroner said those exact words to me, although he might have, when, in May 2020, I received the phone call that my own son had died. I do not remember the words, but I remember the feeling. Everything tilting and spinning and spiraling down. The blood drains, your stomach sinks. All words and feelings of such profound, indescribable, unspeakable loss.

May 2020 was, if you’ll recall, the early days, the early weeks of the COVID lockdown. I was in Oakland, California; Isaiah was in Seattle, Washington. He died alone in his apartment of an opioid overdose.

A few weeks later, OpenAI released GPT-3.

Our tools are cultural not merely technological, so while many people want to frame the emergence of generative AI as simply the latest development in the long history of computers, of artificial intelligence -- transformers, neural networks, tokens, and so on -- we have to remember that what emerges is not just a matter of engineering. It's a matter of markets and politics and ideology and culture.

I think it matters that GPT was released during the COVID pandemic (and ChatGPT shortly "after"), when many of us were stuck at home, isolated and interacting with one another almost entirely through screens.

I think it matters that all this talk about the potential for "AI" to do our jobs comes after labor made some important (albeit tentative) gains during this period: the whole notion of "essential worker"; the successful push for unionization in some sectors; the astonishment from many parents after trying to facilitate their own children's schooling -- all those “teachers should be paid hundreds of thousands of dollars a year” posts on social media; demands during and after the pandemic to continue to work from home, to have more control over space and place and time. AI is a backlash. AI is anti-worker.

I always feel the need to remind people that neither robots nor AI are coming for our jobs. But management probably is.

I think it matters that this latest AI push, with generative AI's penchant for “bullshit,” follows on the heels of growing mis- and disinformation campaigns online. This was precisely the realization many people had come to after Donald Trump's first election as President and during his first term in office. And this was precisely what LLMs have been trained upon.

I think it matters that the technology industry relies on deception and obfuscation and markets its new bullshit machines right as the leaders of this country have openly embraced being liars, cheats, and frauds, have openly rejected knowledge and expertise.

I think it matters that as we have lost faith in institutions over the course of the past few decades -- in the church, in the media, in schools, in science, in medicine (particularly in public health and in vaccines) -- that we are now promised an oracle that can deliver instant and easy answers.

I think it matters that AI -- so utterly opaque in its algorithmic predictions and decision-making -- is the ultimate unaccountability machine.

We expect more from technology than we do from each other, the psychologist Sherry Turkle wrote in her book Alone Together in 2011. I think it matters that trust and solidarity have been eroded for a while now (if they ever really existed or were encouraged in this country).

I think it matters that economic inequality has in the last few decades exploded, that the promises to students in particular – get good grades and you'll get into a good school, graduate from a good school and you'll get a good job – feel pretty empty.

AI is a "normal technology," the artificial intelligence professors Arvind Narayanan and Sayash Kapoor (authors of AI Snake Oil) have argued. But what we have come to see as "normal" is, in fact, utterly abhorrent, abysmal. Yiyun Li writes a lot in her book on learning to inhabit the abyss of grief. What does it mean to normalize the abyss?

AI is the symptom of a broken world. AI is the symptom; and AI is the disease.

Generative AI emerged during a global pandemic -- a global trauma of mass death (1.2 million people in the US died of COVID, and about 7 million globally -- these are, no doubt, figures that undercount how many actually died of the disease, let alone those like my son who died during that time period of other causes -- overdoses, suicide, murder, and deaths related and unrelated to the pandemic).

Mass trauma, mass death and, as such, mass grieving. But it was, at the time and still to this day, a grief interrupted, a grief buried, a grief denied, a grief (contrary to C. S. Lewis's phrasing) unobserved. We were often not able to bury our dead, not able to hold funerals, not able to have wakes, not able to observe the rituals of death, not able to gather, to bring food, to hold and comfort one another.

And when we were told the pandemic was over -- it hasn't really ended; the World Health Organization says there were around 150,000 cases of COVID reported in the last month -- we didn't deal with our trauma. We didn't deal with our grief. We were supposed to bury our feelings; we were supposed to forget. It was back-to-school, back to work, back to "normal."

Or some “new normal,” now with AI – a technology that we didn't want, that we didn't ask for, and that we're told we cannot refuse.

Of course, that's not quite right. We can refuse.

One more correction: there was, in fact, a massive demonstration of grief – an outpouring of grieving in public – during COVID; and that was the Black Lives Matter movement, the protests that occurred in cities throughout the country particularly after the murder of George Floyd. This grief was not private or hidden; it was collective. This grief was not just personal, expressed by those impacted directly by racism and police violence; it demanded from protestors and onlookers, empathy, solidarity. This grief was expressive – even as we are always told with protest, as with grief, that that is not the “good way” to say it. The grief of Floyd’s death – and all the deaths – was not sufficient. It was not simply a marker or memorial of death; but it was an act of life, an act of repair. It was a demonstration of love and loss and fury; it was a commitment to the future.

And again, technology is cultural, ideological not simply technological.

It matters that generative AI emerged with or alongside -- you can decide the preposition you prefer -- a politics that is openly hostile to Black Lives Matter, that opposes diversity, equity, and inclusion. It matters that Silicon Valley companies were among the first to backtrack on their DEI initiatives, were happy to stand with Donald Trump when he proclaimed that AI needed to be purged of "ideological biases," purged of "woke."

Generative AI is, with or without Trump's executive orders, a backlash to diversity, equity, and inclusion, a reinscription of the words and images of white supremacist, heteronormative, Western, English-speaking capitalist patriarchy. That is the corpus that large language models have been trained on -- "the canon" (with all the copyright violations that that has entailed) as well as "the Internet" (thousands and thousands and thousands and thousands of YouTube videos and YouTube comments and Reddit posts and -- with apologies to anyone this might include here this evening -- lots of very mediocre freshmen essays on the theme of family in Romeo & Juliet or the role of "states' rights" in the US Civil War).

In response to a radical outpouring of love, loss, life, grief -- expressed together, embodied, on the streets -- we were presented with, forced to use in so many cases, a technology that severs us from creative expression, dignity, and truth. There is no choice, we're told. "Get over it." "Move on."

One of the problems with grief, as Yiyun Li argues in her memoir, is that it's been described as a set of stages one moves through, as something that has a beginning and, significantly, an end. You will eventually, people try to tell you, "get over it." This is Elizabeth Kubler-Ross's famous formulation: grief as a series of emotions that move from denial to anger, then bargaining, then depression, and finally acceptance. And even if we might've revised this progression somewhat since she published her book On Death and Dying in 1969, society still gives mourners (and not just as workers) a very limited amount of time "to deal with it" before they're expected to "move on."

“There is no rush,” Yi writes, “as I will have every single day, for the rest of my life, to think about Vincent and James, outside time, outside the many activities of everyday life.”

And this, among other reasons, is why I am against the word “grief,” which in contemporary culture seems to indicate a process that has an end point: the sooner you get there, the sooner you prove yourself to be a good sport at living, and the less awkward people around you will feel. Sometimes people ask me where I am in the grieving process, and I wonder whether they understand anything at all about losing someone. How lonely the dead would feel, if the living were to stand up from death’s shadow, clap their hands, dust their pants, and say to themselves and to the world, I am done with my grieving; from this point on it’s life as usual, business as usual.

I don’t want an end point to my sorrow. The death of a child is not a heat wave or a snowstorm, nor an obstacle race to rush through and win, nor an acute or chronic illness to recover from. What is grief but a word, a shortcut, a simplification of something much larger than that word?

Of course, we like thinking of things in stages. We like the order, we like to frame our world, our understanding of time this way -- in hours and days and seasons. We ritualize these -- indeed, that is one of the reasons why our inability to conduct the traditional practices associated with death and dying made our grief during COVID even more unbearable. Without rites and rituals, you cannot “move on.” You cannot grow or shift or change. You are stuck in the past. You are stuck.

The anthropologist Victor Turner used the term “liminality” to describe the one of the key phases of rites of passage, those rituals that mark transition – not just transition into the “afterlife,” for example, but transition into adulthood or into marriage or into society. This liminal phase, as he called it, was “betwixt and between” – a period where you are in the process of becoming something new, but you’re not that new person yet, nor are you the person you were any longer. Liminality, Turner argued, was a sort of limbo – but in that limbo, something really transformational happens – something radical even in the most conservative and traditional ritual practices. Liminality is a time – and to be fair, this can be a very very very brief moment, depending on the rite of passage – of solidarity and equality and unity. Protests, for example, are liminal spaces.

Education, I’d argue, also has elements of this liminality. It is a rite of passage, a ritual of becoming – you enter a child, a “fresh man” and you leave an adult. We have retained some older parts of these rituals – the cap and gown obviously, moving the tassel from one side of your head to the other. But there's more to it than just these practices. You have to believe, I’d argue, in that transformation to be able to commit yourself to the time, to the work. (Socially, culturally, politically, we have to believe it is worthwhile to send children to school, to send them to college.)

But much to the detriment of learning, let alone to the survival of educational institutions, we have seen education redefined as something else -- as a product, not a process. As certification, not transformation. The liminality has been shattered; instead of ritual, society has demanded “outcomes” and “optimization.”

I don’t say any of this out of nostalgia for a once-upon-a-time when college was good. Educational institutions -- whether at the K-12 or the university level -- have always always been deeply flawed, highly exclusionary, full of all sorts of machineries of bullshit. These are, as Michel Foucault reminded us, sites of discipline -- disciplining bodies and minds.

But by dismantling educational institutions -- and AI is really just the latest act in a long long history of dismantling -- we are also dismantling that space for shared practice and purpose, for shared understanding -- “communitas,” Turner called it.

The technology industry -- indeed, capitalism -- prefers “individualization” and “immediacy.” Certainly, it pays lip service to "community" -- Mark Zuckerberg's blah blah blah about Facebook connecting the world. When Google says it wants to organize the world's information and make it "useful," this is a very different mission than the university's. The tech industry's allegiance is to surveillance capitalism, to profit and power, not to knowledge and certainly not to people.

What we are experiencing now -- with AI, with the defunding of public education and public research, with deportations and surveillance -- is more trauma, more loss, more grief. There is no silver lining here, as Yiyun Li reminds us, as much as that's offered as some tepid consolation.

Grief, to reiterate, involves a loss of identity, a loss of the future -- how we imagined things would be, who we imagined we'd become. And there is no good way to say this: it will get worse. And grief doesn't get any easier -- not with the passage of time, not with the number of times one experiences it.

There is no good way to say this. And yet we must always try.

I can only say this, and it's not good, it's not sufficient. It's not really a satisfying way to wrap up this talk. But here we go...

Grief is an expression of love. We grieve because we love, and that love does not end with death. I grieve for my son. I will grieve forever. I grieve for the future we will not share.

When I talk to teachers and students alike, I hear such grief as well: grief about what AI threatens to do education, what it's already done to the work of teaching and the work of learning, the work of research and reading and writing.

We grieve because we love. We grieve because we care. We grieve because we know that the machines do not, and that the community we try to foster -- on campus, in the classroom, in our scholarly works -- is threatened with erasure. We grieve because we fear forgetting; we worry that people will forget what is beautiful and what is difficult and what is joyous and what is horrible about education. We worry that, if we do not grieve, we give up the struggle to go on, to persevere, to live.

But we do not, we should not grieve alone. We should not be made to feel alone, feel crazed by our grief, feel crazed for grieving. We can, we should grieve together, grieve in public, grieve in protest. Such is comfort – "com" + "fort," a word that means "with" + "strength."

Technologies are often wielded in ways meant to imply that humans are weak, messy, slow, stupid, replaceable.

We are strong, messy, awkward, flawed, irreplaceable. All of us.

Our strength comes, in part, from this vulnerability, from our humanity. Together in the flesh. Not isolated, individualized thru some algorithm. We cannot allow systems and practices and machinery to foreclose this humanity, to automate the decisions, the expressions, the explorations that we turn to and that we struggle with in education, in this imperfect but liminal space of learning.

"There is no good way to say this" but to say this: AI is the antithesis of education. It is the antithesis of the future. As such, it is a kind of epistemological death, and I recognize -- thanks to capitalism and neoliberalism and imperialism and racism -- we have long been surrounded by such efforts; we are grieving already. And yet, we go on.

One final note that I think I'd be remiss not to state, even though there is no good way, or rather no polite way to say this:

Some men (and I do mean mostly men) would rather spend trillions of dollars on an idea that is financially, technologically, morally, and environmentally unsustainable, they’d rather destroy democracy and destroy education and destroy the planet than just get therapy.

Thank you.

Today's bird is the mourning dove. Obviously.