by Ashutosh Jogalekar
Like most people, I have been baffled, mystified, unimpressed and fascinated by chatGPT, the new AI engine that has taken the world by storm over the last few months. I cannot remember any time that a new AI development got so much attention, led to so much mockery and caused so much alarm. As the writer Ted Chiang wrote, chatGPT – and what would inevitably be its subsequent versions and spinoffs – represent a blurry, lossy, version of the Internet, containing all the unique strengths and gory flaws of that medium. There is little doubt that chatGPT is not intelligent the way humans are based on the complete lack of nuance and and elementary factual errors it still makes as it cobbles together patterns from data without high-level understanding. But there is also no doubt that it creates an illusion of human intelligence that is beguiling, endlessly fascinating, addictive and frankly disturbing and potentially dangerous. Dangerous not because the machine is intelligent but because humans will no longer be able to distinguish between intelligence and the illusion of intelligence.
Some people think that AI engines like chatGPT herald a global threat of artificial general intelligence (AGI) because of their unprecedented ability to generate misinformation and the illusion of intelligence. AGI is AI that escapes from its narrowly defined applications to become all-encompassing. We should grapple with the challenges that AGI would pose, even if the probability of a true AGI remains slim. But as the parent of a 2-year-old, I feel an even more urgent question looming over the horizon: what are chatGPT or AGI going to do to our children’s generation? How can we try to ensure that they can handle the challenges posed by these unprecedented technological developments?
We don’t know how to answer this question yet since the impact of technology on society is inherently unpredictable – who would have predicted that social media would drive us apart by cocooning us in our echo chambers instead of triggering a global open exchange of ideas – but some lessons from history combined with old-fashioned, boring common sense might help us make sense of the brave new AI world that chatGPT represents. When the new becomes unpredictable and risky, the old-fashioned and boring may not look too bad.
Even if we don’t know exactly how such AI might impact our children’s lives, there are safe guesses that we might venture. Because of its easy, lazy availability, chatGPT and its cousins will be extensively used by both teachers and students in schools. Students will of course use them to answer their homework questions (and predictably to cheat), but teachers might also use them to create tests and to set up challenges for their students. In an age where there are problems with both teacher shortages and teacher quality, such AI might be seen to provide welcome help. One can imagine a time when AI is used to both create and evaluate standardized tests. And as the capacity of AI to solve math problems is being demonstrated more and more, we could well see it being used a tutor or intelligent assistant even in graduate study and research.
These developments might be salutary in so far as the AI captures the essence of what it means to ask intelligent questions and answers. But we aren’t there yet. Many people have captured examples of chatGPT generating absurd and potentially dangerous answers, like figuring out a use for asbestos in baby food when asked for it or creating a fictional meeting between J. P. Morgan and Mark Twain that never happened. In my own case, I asked chatGPT to give me a proof of a well-known fact in number theory, that the number e (the base of the natural logarithms) is transcendental. A transcendental number is a number that is not the root of a polynomial equation with rational coefficients. I had probably studied the proof sometime in college but had forgotten it. chatGPT came up with a sequence of steps that looked plausible and logical to me. But expert mathematician friends of mine quickly identified an assumption in one of the steps that was nonsensical. The same goes for the meeting between Morgan and Twain: they lived during the same time span in history and were both influential in New York social circles, so it’s not implausible that they may have met. But only a historian will be able to tell whether they actually did.
And that’s what I am calling the illusion of intelligence. AI is capable of generating stories, facts and situations that taken together may be complete nonsense, but neither your average student nor your average teacher would be able to tell if they are; only experts will. And the nonsense might self-perpetuate if not recognized on time, especially since some of these models reinforce themselves through what they perceive as successful results.
But to me as a parent, the illusion of intelligence is nowhere as concerning as the illusion of humanity. This point was driven home by a recent news article written by a reporter who had an extended conversation with an AI that is part of Microsoft’s search engine, Bing. The reporter had a conversation with the AI for several hours and what followed was a bizarre set of passive-aggressive comments culminating in an obsessive declaration of love by the AI akin to that displayed by the lead character in the film “The Crush”. It is hard to know what turns of logic the AI is taking that is enabling it to craft such responses but it is clear that this behavior is unexpected and startling, albeit deterministic.
This behavior raises some genuine concerns for my daughter’s generation. Telling accurate information from almost-accurate information will be hard enough, but the emotional repercussions created by humanlike AI responses will be more serious. What if teenagers form fictitious relationships with AI that make them fall obsessively in love with these disembodied intelligences? This does not seem unlikely in an age when many teenagers lack human connection to begin with and are instead surrounded by their screens? What if emotional manipulation by these AIs at an impressionable age leads young people to harm, either to themselves or to others? One question is whether such concerns are similar to the ones voiced by every parent’s generation; before the advent of radio, TV and the internet. However, it is hard to dispute the qualitative change that these AI systems embody. Radio, TV and the internet were all propaganda weapons, but they were still distinct, disembodied entities that could be distinguished from human beings. With the new generation of AI tools, distinguishing the deed from the doer, the tool from the maker, is becoming increasingly difficult.
So what can we do to make sure that our children’s generation is not swept away by a mass of deceptive disinformation and emotional manipulation? The first thing to realize is that there is no foolproof or satisfactory solution, that every generation until now has had to grope its way toward making its own kind of peace with technology that’s speeding leaps and bounds ahead of humanity’s emotional and social understanding of it. But there are things we can do, things that have worked reliably for centuries. For this my lodestar is, curiously enough, Harry Truman.
To me, Harry Truman is a good role model for how humans can grapple successfully with sweeping, startling technological change. Truman was a simple Missouri farmer who remains the only president to not go to college. But beneath this seemingly rough, uneducated exterior lay a deep reserve of good old-fashioned common sense and great determination combined with an autodidact’s ability to read and to learn fast. Truman was faced with an utterly alien, unprecedented and immensely destructive technological force in the form of the atomic bomb, much like my daughter’s generation will be faced with an alien and potentially destructive force in the form of chatGPT and other forms of AI. Just like my daughter’s generation will be thrust suddenly into this brave new world, Truman was thrust suddenly into the presidency after Roosevelt’s death; he did not even know about the Manhattan Project until after he became president.
The way Truman dealt with these forces and with other challenges was with what his biographer David McCullough called ‘the old values’: “work hard, do your best, speak the truth, assume no airs, trust in God, have no fear.” This armamentarium of values stood Truman in good stead and enabled him to face the unexpected and immense political and technological headwinds that suddenly came his way – nuclear weapons, the Cold War, the Korean War and the rise of communism, a destroyed Europe in desperate need of recovery.
I see a similar list of what I too would call boring ‘old values’ that might equip the next generation to deal with the unexpected and massive forces of AI.
The first value – which Truman shared – is to read authoritative books. This value precedes the coming of AI and applies to the internet. Unfortunately we have become too accustomed to reaching for Google or Siri or Alexa when we have a question. As the examples noted above indicate, when we reach for chatGPT or other similar AI the results will be problematic. Instead, we can fall back on those tried and tested sources of truths that have educated and enlightened people since medieval times. Read E. O. Wilson to understand ants and biology, William Shirer and Hannah Arendt to understand totalitarianism, Enrico Fermi to understand physics, Barbara Tuchman for the First World War, William James for psychology, and of course, if you want to explore the deeper side of humanity, there’s always Dante, Homer, Confucius and the Bhagavad Gita. Unlike chatGPT, these sources have stood the test of time, often for centuries, and been vetted over and over again as sources of truth and authority. And while we are at it, let’s also discover the pleasures of the public library and our local bookstore. What chatGPT will serve to obfuscate, good books will illuminate.
The second value is to rediscover the value of human interaction. Psychologist and writer Steven Pinker was recently interviewed about chatGPT and one of the interesting things he said was that he doesn’t worry about the new AI taking over our lives because we still value authenticity. For this, look no further than vintage collectors like myself who are often willing to pay a small fortune for signed book editions or photos or coins and stamps from a particular era. No matter how convincingly digital technology reproduces these sources, we will not prefer them to the originals. As human we still crave authenticity, and there is nothing more authentic than other human beings. The most important bulwark for our children’s generation against the deceptive arts of AI will be other people, and so it will become more important than ever now for us to make sure that our children get adequate social interactions.
The third value might be the most important of all. Let us teach our children to think critically, to have their ‘baloney detection kit‘ always available, to always be ready to ruthlessly question both others’ and their own assumptions, to know – in case of things like chatGPT – both the train of logic and the provenance of an assumed source of truth. Inculcating critical thinking is not easy since both adults and children are prone to jumping to conclusions and to take the intellectually lazy way out. But critical thinking can in fact be taught by parents if it becomes part of their shared experience with their children, if it’s taught during dinnertime conversations, during road trips, as bedtime stories. Critical thinking should become second nature for our children in the age of AI, and for that it is up to us as parents to seamlessly integrate it into every single day’s activities and conversations. When all is lost, much can still be gained if we have our minds’ critical faculties.
Whatever its limitations and ultimate influence, there is no doubt now that a wave of AI is upon us and even more upon our children’s generation. It is up to us as parents to equip the next generation with the tools necessary to weather and even profit from this wave. To deal effectively with it, the best thing our children can do is to reinforce their humanness in all its intellectual, social and personal glory. Sometimes the best response to new values is in fact the old ones.