Part One - Apprenticeship
My older son Ben had his first “serious” internship this year, including performance feedback, a mid term review and 12 work hour days. While I do want him to have a healthy attitude to work and a “Sabbath rhythm” that sets aside time to contemplate, play and rest, I am grateful that some of this time is spent on “grunt work” such as spotting typos and conforming slides to style guides. Why? Because experience has taught me that apprenticeship - learning to do simple things exceptionally well - is commonly a foundation for fruitful creative and critical thinking.
In the 2011 documentary Jiro Dreams of Sushi, sushi master Jiro Ono explains that he makes his apprentices squeeze boiling hot towels with their bare hands to develop sensitivity and feel before they are even permitted to touch fish. I had my own “Jiro” experience in law school. A well-known law tutor taught (and wrote the book on) a subject called “Conflicts of Laws”. Conflicts jurisprudence sits at a level of abstraction that - when one is asked to write an essay - cries out for the undo, copy and paste functions of a modern word processor. Not for John Collier. Hand-written essays only.
“To force you to think.”
Annoying as this rule was, JC had a point. Tasks, especially simple tasks, can be done mindlessly, or they can be done in such a way as to impart useful disciplines. Writing conflicts essays by hand, I did much more up-front planning. I paused to more carefully consider how to deliver information so a reader is receptive, how to structure an argument to be persuasive, how to choose words for effect, and how to formulate tangible takeaways. Writing by hand, you need to have thought things through to the point that you know where you will land before you can even start. Such skills are helpful for lawyers - but are also relevant whether you are writing a speech for FDR to rally a nation at war, or writing your best friend a birthday card. And though I only draft digitally nowadays, those disciplines have stayed with me and I frequently draw on them when I write for Deeply Boring.
Technology - as John Collier observed - affords convenience, but at the expense of a kind of learning that takes place at a more subliminal level. I fear AI will hyper-accelerate that trend. 30 years ago I learned how to use a Mac to write essays (but not for Conflicts!), and it already risked impairing my thinking. Today, my son Ben uses coding copilots to debug his Python scripts and PANDAS to crunch data - and I wonder what Collier would say. Tomorrow we will rely on GenAI to do our reading, writing and some (or even much) of our thinking. Is this unreservedly good - given we might learn a lot through the playful experimentation that emerges from doing simple tasks, like editing college essays and making sushi?
It seems to me - when starting out at the entry level of a new endeavor - that simplicity directs our creativity and inquisitiveness inwards, the peripheral boredom pressurizing it into an explosive force. In “Jiro Dreams” we learn how Jiro experiments with whether massaging octopus for 45 minutes rather than 30 better brings out the flavor. How many hours of mollusk massage does one clock when scaling the heights to sushi nirvana? Could one still reach the summit without? If you need more persuading: try this.
Perhaps to many of us such effort might seem pointless. But the counterargument might be that perhaps our failure to grasp this is what makes many of our lives feel pointless.
Part Two - Seeing Ghosts
Often, true authority to criticize something comes from knowing it better than everyone around you - authority which is neither cheap nor quick to come by. Good lawyers succeed in the practice of law by paying close attention and uncovering the questions that nobody else is asking (“Mr Simpson, does the glove fit”?). This is not something taught in law school. It is something learned from scanning and re-scanning what we perceive with a skeptical mind.
To illustrate the formative contribution of repetitive tasks in activating such a skillset, let’s imagine the grunt work assignment is reading and summarizing a series of emails from Steve to Bill, lifelong frenemies. In all of the emails, Steve greets Bill with the same salutation: “Hi Bill.” But in one email, Steve gets right to the point and drops the formality: “Hi.” As humans, we pick up on the change in tone immediately. We are sensitive to it. Does synthetic intelligence catch this? LLMs are good at summarizing what is there - but are they even looking out for what is not there? Because when humans analyze, we do not just look for what is seen, but also what is unseen. We recognize that life often unfolds in the gaps.
How effectively does an LLM’s text summarization capability convey such subtleties - if at all? Does assigning workaday tasks to AI result in missing such potentially important nuances altogether? I wonder what to make of a world in which we make certain tasks so easy that our younger generation essentially spam-clicks its way through foundational learning, and loses the opportunity to acquire the essential critical thinking skills that those tasks would otherwise foster.
Lateral thinking is important for a whole host of reasons, but one that is rarely discussed is how such thinking, as an expression of individual agency, is an antidote to technological conformity. Technology drives us towards standardized, prepackaged experiences. Tweets have character limits. ChatGPT produces block text for us to refine. Insta showcases life without imperfections. But it is in the messiness of the edit that life happens. I moved to Substack to escape the LinkedIn character limit. My essays offer more questions than answers. It is often in the imperfections that I find life’s most profound revelations. In all of this, I express my agency - the agency to question, to create, to challenge.
Interestingly, Large Language Models used for Generative AI also have imaginative capacity. Sometimes you ask for a factual response, and get a made up answer. Researchers call this tendency hallucination, or confabulation. This propensity to state untruths is one of the more frustrating aspects of GenAI - one that that advocates of AI Safety perseverate over.
But maybe GenAI hallucinates because the neural network of an LLM was designed to simulate the human brain, which can both reason, based on fact, and speculate, based on conjecture. We “hallucinate” (“see things”) whenever we speculate about the unknown, whenever we ask a question that we do not know the answer to. Without this speculative instinct, we can neither innovate (Could we make all the information in the world indexable? - Brin and Page, 1998) nor challenge the status quo (What would happen if I sent Henry to China? - Nixon, 1971).
Viewed this way, our propensity to speculate is inseparable from our agency, as expressed in critical thinking. We need to imagine, if we are to reimagine the world. If so, perhaps hallucination is nothing more than ChatGPT attempting - but failing - to engage in critical thinking. What should we make of this? Is it evidence of singularity? I think no - at least not yet. An important difference between machine and human hallucination is that when AI hallucinates, it presents falsity as truth. It does not realize it is imagining things (Can AI even understand what it is to imagine? Does it dream of electric sheep)?
But not us. Unlike AI, we are constantly flitting back and forth between reality and fantasy. We can get caught up in the fantasy world of a Lord of the Rings re-run on TV, parallel process the background noise of our children playing (or arguing, or fighting), and ask ourselves whether we want our children to be more like Frodo or Samwise when they grow up, all at the same time. We distrust AI because it sometimes sees ghosts. But we deeply trust each other - to work, live and play together in a world that is simultaneously real/tangible and fantasy/aspirational. The multi-tasking virtuoso that is artifactual intelligence seamlessly interweaves the fantastical suspension of disbelief, rational, logic-based thinking, and the speculative instinct that underlies critical examination.
This is a miraculous trifecta - and whether one realizes it or not, humans have designed foundational experiences to encourage all three in our younglings. We can use play to encourage the first tendency (fantasy), and schooling to promote the second (logic). What a shame if we were to undermine the disciplines of apprenticeship that nurture the third - to abdicate our potential to make miraculous things happen in our pursuit of inconvenience avoidance.