AI, Technology, and the Myth of Progress — C. G. A. Peterson
Jargon-tastic preamble: My research moved out of philosophy of physics and into artificial intelligence ethics. This piece is a sort of interdisciplinary transition piece. Philosophy of time and AI Ethics don’t really belong together, but that has never stopped me before. So I hope you enjoy this lil’ essayistic farewell.
“Such an attitude stems from a tragic misconception of time, from the strangely irrational notion that there is something in the very flow of time that will inevitably cure all ills. Actually, time itself is neutral; it can be used either destructively or constructively. More and more I feel that the people of ill will have used time much more effectively than have the people of good will.” -Martin Luther King Jr., A Letter from Birmingham Jail
Technological progress seems undeniable. With it we extend both our reach and what can reach us. Microscopes and telescopes let us see unknown worlds and times — dead pasts briefly brought to life either to be savored as moments or tabulated as raw data, their beauty dissected and digitized. Can we do both?
Each innovation extends our reach further into space and with more precision across time. The written word allowed sending letters across vast distances slowly; our Network allows the same feats in milliseconds. We can speak with our families a thousand miles away, with small enough time delays that we can interrupt each other as seamlessly as in person. But we are also growing in our ability to manipulate knowledge. Artificial intelligence promises not just a physically extended reach, but, somehow, cognitive enhancement. But does mere increased power, even increased cognitive power, itself constitute progress? If progress is moving toward the good, is technology necessarily progressive?
Maybe we would do better to call technological progress technological advancement, perhaps even technological motion. Whether increasing power constitutes progress is a matter of how we use that power. Ask Dodos if they think technology is progress. Ask the families of those we killed in the Tuskegee trials. Those in positions of power might say such control is a symbol of humanity’s increased dominion of the natural world. To those falsely sentenced to prison by algorithms, a better word is probably tyranny.
The more power we have the more easily we destroy each other. Unlocking the atom gave us nuclear power and pet scans, but it also gave us Hiroshima — Chernobyl. And this is from unlocking a simple system. Cells are alive. They replicate. As we unlock the genome, as viruses become tools, the potential devastation, the permanent species level changes that could be wrought by CRISPR make The Bomb look like a book of matches.[1]
II.
Six months ago, I met an actor who used to be a theologian. His dual interest was intriguing. We chatted, swapped numbers, and exchanged equally noncommittal intimations of “let’s get a beer sometime.” A few months passed and I thought of him. I thought I’d text and see if he still wanted to talk shop. I went to my phone, and had a novel, delightfully frustrating experience. I knew his number was in my phone, but I couldn’t remember his name. It was then that I fell in love with the conundrum of trade offs implicit in our pocket-sized prosthetic memory.
Augmented memories have been on the table since the written word, but its scale has never been so vast. The written word changed the face of society; the technologically extended mind will change individual identity. Whether this change constitutes progress depends on the values we associate with identity. Do we value the individual over the collective or vice versa? This depends on our history.
In recognizing that our individual histories determine our perception of whether technology is necessarily progressive we enter into the more nuanced perspective — a historicized view of technology.
III.
The progress of technology is called into question by the inequality of its detrimental effects on the world. As it was miners who inhaled coal dust, it is scavengers who inhale heavy metals scrounging through tech scrap today. Oppression is to be expected from human nature — from history. We forget that in all likelihood the wealthy will survive climate change.
IV.
Thousands of hours of video are uploaded daily to YouTube alone. But beyond this public cloud, we are constantly amassing our private stashes. Emails kept in nested folders, text messages saved, photos of meals, of events. We feel these must be kept safe, even from time. We make copies elsewhere. As if redundancy in data management could stop death.
V.
Let the retrieval problem be defined as the inability of humans to pull necessary or desired information from our ever-growing archive. Like the lost theological-actor’s number, prosthetic memory is of no use if we cannot meaningfully access it. Far from making us god-like, an infinite archive only sheds light on our cognitive finitude. Even if we record all of the past, we can only access a fraction of it.
A possible solution to the retrieval problem is AI. If we can’t sift through the data, we can build a machine to do it for us. A “smart” archive. But like progress, memory is both contextual and laden with value. Who do we choose to remember? What aspects of them? It seems AI could solve the problem, but another problem emerges: If we use AI to determine which fraction we see, we leave ourselves victim to the values of AI. Thus, we resolve the retrieval problem and inherit the control problem.
VI.
What values would a computer have except the ones we give it? Chatbots learn the language of the humans who speak with them, even when that language contains acts of linguistic violence. As we archive and algorithmically augment our memory, we may gain a virtuous cycle. However, we may also amplify our vices. Would AI be as homophobic as we? If it learned our laws around mandatory minimums would it retrieve more favorable memories for different races? How does a machine know what we ought to remember?
It is worth noting that these algorithms are not without hidden conflicts of interest. Many are written to keep us clicking, watching, and posting. Big Tech wants our attention. This is not science fiction. Facebook already prompts users of “memories.” So do Apple and Google’s private photo stashes. Twitter provides collective “moments.”
Healthy memory allows us to occasionally dwell on losses to process them properly. But if presenting past pain causes us to click away, we can be certain that Silicon Valley will not want us remembering it. A good friend can suffer with us, give us the courage to feel our own pain,[2] but the arms of Big Tech do not console. They distract. Numb. As our memories become digitized, and as we retrieve them through digital filters, we cede part of our agency to the predictably callous forces of the free market.
VII.
So, we teach machines value. But how can we teach the right values when we cannot figure them out ourselves? How still when our recent history questions the concept of “right values” at all. How in this time of deep political divide can we even dream of settling on consensus in AI? How when, from the right, science itself seems under attack, as well as such basic concepts as facts and fair reporting.[3] How when from the left, the concept of morality itself is called into question.
If we really wanted to know if AI counted as progressive, we could if we could settle on which values we thought were correct. But this debate depends on another — one which over centuries remains unresolved. In the meta-morass of postmodern uncertainty, we are not even allowed to start asking if we have found the right values. We haven’t finished debating if there is any such thing as values in the first place.
Aren’t values relative to humans? Don’t we create them? Do we create them with will and power? Can’t we create them however we wish? What are the limits on our normative generativity? Is what we bind on earth bound forever in heaven?[4] Or are they revealed to us by God? Are they objectively valid? Even if they are objectively valid, are they not understood by finite, historically bound humans?
VIII.
Could science help with values? Science is always arriving at new truths. As a method, it doesn’t need to lead us to Ultimate Salvation to be leading us somewhere real. The convergence toward truth in science over time is one of the greatest pieces of evidence that there is objectivity to the cosmos — and that we have at least some kind of access to that regularity. It comes at a cost. Science has to keep discarding the past. We no longer believe in flogiston. We no longer use phrenology. Maybe we will be eternally discarding. Tossing out Newton for Einstien. Einstein for Hawking. Never actually landing on something substantive.
Niels Bohr understood this. He knew that Einstein’s theories allow for time to bend and that Newton’s did not allow for this. But he didn’t believe this means we have no solid ground to stand on. But Einstein’s (and Lorentz’s) insights don’t mean Newton is wrong, per se. Nor does it mean forces and velocities don’t exist. It merely means our description of them must be modified. Applying Bohr’s correspondence principle, we can see that far from disproving him, Einstein predicts Newton. This is convergence not deconstruction. Diet Einstein is Newton. So, science is always arriving, converging. It may never arrive, but the more it arrives, the greater sense we have of, seemingly, the world as it really is. And beyond reifying the old, beyond better defining the limitations of the old, the new theory does us a greater service: It expands the edges of science — closer to the things-themselves.
IX.
So, the more time passes, the greater a sense of the objective world we have. Does this not constitute progress? We have newer knowledge. Our newer knowledge verifies the older knowledge and stretches the boundaries of what we know, implying a cycle of increased objective regularity.
And technology is science made pragmatic, science in the world. So, if science makes progress, and technology actualizes it into the world, how can progress too not exist in the technologically augmented world? What’s more, if science’s progress is objective in nature and does not rely on political perspective, this translates to technology too. How can one argue with cochlear implants? With penicillin?
X.
“The current amazement that the things we are experiencing are ‘still’ possible in the twentieth century is not philosophical. This amazement is not the beginning of knowledge — unless it is the knowledge that the view of history which gives rise to it is untenable.”
- Walter Benjamin, Theses on the Philosophy of History[5]
The indignation American democrats feel at the rise of Trump’s neo-nationalism, though perhaps righteous, often takes the form of an anger at “old” problems. I first felt it when I heard that then campaigning Trump refused to denounce the head of the KKK as a supporter. Perhaps we can have a debate about healthcare, I thought, or maybe even marriage equality. Those are the issues of our time, but how could we still be dealing with the KKK? Aren’t those old problems? How are we making progress if we are stuck dealing with “old problems?”
From a certain perspective, this is the wisdom of post-modern deconstruction. Human nature can never be overcome. Epistemically, morally, hermeneutically — short of divinization we cannot overcome our own finitude. If the problem is the Pythagorean theorem, yes. We have overcome it. Like the progression from Emancipation to Suffrage to Gay Rights, we have triumphed, guided by an enterprising Hegelian spirit, forward into a better future, to solve greater and better problems. But if the problem is our finitude, the limitations which constrain us are baked into our nature. It is no mystery, then, that we fight “old” battles.
XI.
Perhaps then we attempt to cross the barrier between fact and value. Maybe we answer the question of values by turning to science itself. If the universe is fundamentally ordered, can we not lean on the objectivity of science in determining questions of value? Might science allow us to sidestep the postmodern deconstruction of objectivity imposed on us by our historical embodiment? Sensation and language may fail us, but what about space and time? If anything can give us an objectivity in the question of progress, these ahistorical physical realities should.
But even time and space are not themselves objective; we known this for 100 years. Even time is relativized.
In the Lorentz transformations we see that two people need not even agree on how much time has passed. A clock’s tick ∆t can change depending on our speed v. And this speed is itself not measured absolutely, but relative to others. Derrida might say there is no central clock. But, as in postmodernism, in relativity, there is no central clock because all observers are their own centers with their own proper times. And in relating many centers, strange realities can come to be. Given the right conditions, a child might become older than her father.
If 5 years for one could be 5 days for another we run into an interesting phenomenon: the order of events itself becomes relativized. Where one sees a volcano erupt, a star explodes, and then a comet crash into Jupiter, another can see these events as all happening at once. What’s worse, given the right conditions, there could be frames where these events occur in the exact opposite order. History is different in these frames. First the comet, then the star, then the volcano.
There are constraints on this time reversal, but it deals a death blow to the progress myth: How can the present be better than the past if the present does not even fully exist, if the present itself is relative to where we happen to be space-time thrown? If my present needn’t match yours, if your history could be my future, the intellectual foundations of a collective concept of progress crumble. The relativity of simultaneity calls into question the notion of progress because it questions the ontological status of the past itself.
XII.
How can we escape this aporetic impasse if we encounter a deconstruction of objectivity at all turns? We find grounding in neither science nor time. So, progress needs a new form. Objective progress is lost, the objective present is lost. If progress is to exist, it must take a form that honors the subjective turn.
But what are the limits? Could we ever be “post” postmodern?
XIII.
Claim: The ontological status of time is best understood in relation to finitude.
XIV.
Imagine a fossil. We will never see the creature who created it, but we know the limits of what would have been its reach. We can be correct both in saying we will never have full objective knowledge of the fossil itself, and we can also say we are epistemically justified in constraining our understanding of it. It is shell shaped, not fish shaped. So too there must be limits on the crumbling of the ontological status of the past.
Thankfully, in the physics picture, there are such limits on time reversal. Empirically and in theory, nothing can travel faster than the speed of light. This embodied practical limitation is often understood in what is referred to as a “light cone.” If we cannot travel faster than light, there are places in the universe that are sufficiently far away that no matter what we do we will not be able to influence them. Even if we devised history’s most powerful laser and sent a signal, they are too far away for us to influence in our finite lifetime. Events that occur at these distances are called “spacelike.” Their time order is subject relative and reversible.
(Pardon the next three paragraphs of physics lesson)
One might imagine one’s sister has moved to Mars, and she has been badly injured in an accident. She is bleeding rapidly and will die within 20 minutes. She calls home. The signal takes 14 minutes to arrive. We can send a response. But if she will be dead before it reaches her, the event of her death is an event outside the light cone of our response. Although in our history, her death is before our message is sent, since our message can never reach her, other histories have the events reversed. The universe allows for her death and our sending the message to be at the exact same moment, for her death to precede the message, or for the message to precede her death. So much for objective history.
XV.
There is possible a confusion here though. There are four events. The injury, the distress signal, our response message, and her death. Our response is far enough away from her death in space-time that they will never causally affect one another. They are spacelike events that can be seen in any order. The first two events, the injury and the distress call, however, are close enough together that they are causally connected. They are in the same light cone.
Here we see how locality can modify progress. If events are close enough together that they are causally connected, there is no world in which their temporal sequence can be reversed. There is no observer who can see your death as occurring before your birth. There is no frame of reference in which the distress signal can precede the injury. To observe this order, one would need to travel faster than light. As best we can tell, this does not occur. Space-like separated events might be considered, decoupled from each other’s histories. And we see the concept of the present is globally relative in that there are events at vast distances that can have their order reversed, but on a practical scale, on a human scale, there is an “absolute” past. Perhaps, then, relativized progress is possible, if it is constrained to local-scale progress.
XVI.
Here an interesting thought arises, the absolute past exists in a sense, but it is our causal finitude that allows for it. Relativism localized, becomes objectivism pragmatized. A general form could be: the only things that are truly relative are the things that are so far from us causally that they become pragmatically meaningless. Or conversely, if things are close enough to us to affect us, we cannot fully relativize them.
So, paradoxically, we find finitude, causality, and locality save us from the tyranny of total relativism. There are events whose future could be my past, but by definition these events are so causally distant from me they become pragmatically vacuous. So, in a feat of cosmic irony, our finitude creates a locally absolute past around us. By constraining our causal influence to our local light cone, we might find the progress we seek. That is, like Einstein requires us to modify Newton, we must modify our progress myths. If we give up Absolute Objective Progress, we might yet find Locally Objective Progress. Or — the ontological status of time is best understood in relation to finitude.
XVII.
The borderline: There are events that are too far from us to influence. There are events too close to ever be separate from us. But what is the border? It is conceivable that there may be events that exist right on the edge. What is the deciding factor that pushes them over the edge — either into a void beyond our reach or into our life’s toolkit, our locally objective history?
Maybe this is Benjamin’s “moment of danger”[6] The precise moment that determines whether an event is to become history or be consigned to oblivion. The past reaches out to us. On the borders, whether we let it reach us is our choice and in so doing we create the boundaries of our locally objective history.
XIII.
The absolute gives way to the relative, but the relative is made absolute, locally. Absolute history does not exist, but local history has a local absoluteness. Maybe there is a God’s eye view. Maybe God is omnipresent. Maybe God could cause something in “another light cone,” but we do not inhabit this perspective, we cannot understand this type of time. We needn’t see this way, in order to understand local progress. Locally, we have the future. Locally, we have the past. Locally, there are events which flare up before us commanding our attention. Locally, we can decide if these events are to become part of our world. We give up the One Objective Reality, but we gain many realities — realities whose boundaries and regulations must be renegotiated under the condition that they become causally related. A task of increasing priority in this technologically connected, globalized world.
XIX.
So, in techno-globalization, in the construction of the Network, we return to the question of technology as progress. Technology extends our causal reach and what can reach us. In widening our causal bubble, it widens our locally objective universe. So, it would seem, as we achieve more and more, the past becomes more and more locked into our history.
Facing the past, our increasing influence creates an ever-greater archive. From our local realist perspective then, our personal and collective clouds achieve more than mere tabulation. In granting greater effective causal access to the past, our archives appear to be, somehow, creating a greater collective shared reality. Digitally overlapping light cones cement a shared past, reducing the risk of the moment of danger, in which forgetfulness, decay, and death break down the edges of our shared reality.
No wonder our tools and knowledge tempt us to think we’re gods. We are living through a revolution in technology that is creating a revolution in history. But in understanding finitude, we understand just how much this is utter fantasy.
Facing the future, we are left with a problem of access that gets exponentially worse every second. More new data is stored today than was stored yesterday. More will be stored tomorrow than today. If Moore’s law and recent history are to be trusted, we are witnessing the beginning of a digital Cambrian explosion in collective memory.[7]
But this explosion is the quickest path back to the retrieval problem. We return to the theologian-actor’s number, desired but lost in the archive. We return to the paradox of having access to more information than our finite minds can access. A digital despair of possibility.[8] So, we write AI to search through the archive for us. We have so many facts in our globalized archive of digitally interlaced light cones, that we teach the machine to sift through them for us. And, in doing so, we re-inherit the control problem. Who gets to decide what are the most important facts?
We could be tempted to lean again on the regularity of the universe as expressed through technology. We could believe that since you can’t argue with penicillin and you can’t argue with cochlear implants, you also can’t argue with the AI that we task with telling us what to remember. We could even believe that this brings us closer to the God’s eye view, and, seemingly, closer to progress.
But in giving over our powers of memory and history to machines, we do not become super men, we become cyborgs with implants unseen. And we don’t design the implants ourselves. We largely let Big Data do it. And as long as we allow advertisement and market forces to form the backbone of how we filter this information, a trend that shows no signs of stopping, we will live out our decades singing along to our Hegelian march toward earthly techno-progressive perfection. We will live thinking we march, scientific and victorious, toward the future. But, all the while, slowly and imperceptibly, we will cede more and more of our minds and histories to the lever pullers, the coders, and the corporate Silicon Valley backroom influencers, whose names we will never know and whose values may or may not be our own.
[1] And this is not even accounting for technologies we can’t even imagine yet. I imagine a swordsmith’s first encounter with a machine gun.
[2] See Brené Brown 2012
[3] The objective size of Trump’s inauguration crowd comes to mind.
[4] Matthew 16:18–19
[5] Benjamin Theses on the Philosophy of History 257
[6] Benjamin Theses on the Philosophy of History 255
[7] Hat tip Halbwachs
[8] Kierkegaard, The Sickness Unto Death
Originally published at https://cgapeterson.com on February 25, 2019.