Hot Water

July 30

1h 17m

Episode Description

S2 E11 Hot Water by Julius Knapp. Interviewed by Brian Smyers

The Summer Of 2016

The summer of 2016 was hot. I know because I was outside baking in Green Ridge State Forest on a camping trip with three other idio-…uhm, friends, including the one I would eventually start a podcast with. I had been camping with this group several times before, in fact. Fortunately, we had all learned enough lessons to know that a camp shower would be valuable in this situation. One by one, we each cleaned off the summer sweat and grime that had accumulated throughout the day.

The camp shower was a small pop-up shelter, just barely large enough for a single person. Hooked onto the top lip of the shelter, was a small shower head connected by a hose to a small pump. The pump sat in a 5-gallon reservoir of heated water. Yes, despite the summer heat, we still enjoyed the simple pleasures of a warm-to-hot shower. The water took some time- maybe 15 minutes- to heat over a single propane stove.

With a limited supply of hot water, showers would have to be kept short. Through the first three people, only 1-½ to 2 gallons of water had been used. Now, my daily showers are typically only 5ish minutes long. I use two-in-one shampoo and conditioner and then body-wash. Quick and efficient. On the day in question, however, somehow I managed to use almost the entire remainder of the reservoir. I can only assume that a leak in the line had manifested somewhere or that a fitting had come loose after the three previous uses. Regardless, the event has become the subject of repeated ribbing over the years. I do my best to roll my eyes, grin, and bear it, but I’ll admit it’s difficult given that I know the truth: some mechanical failure was leaking the water as I was using it.

Please William, May I Have Some More?

This event has been on my mind recently and before I get into it, I promise, dear reader, that I will bring this back around. In-home hot water heaters are a relatively recent addition to modern homes. Consulting Copilot and the relevant sources suggests that installation and use of residential hot water heaters occurred primarily throughout the 1940s and 1950s, following World War II.

Despite invention of the primary components almost a half-century earlier, the infrastructure- public water utilities, indoor plumbing, gas utility supply, etc.- was not readily available in a residential setting until that period. This is to say nothing of the cost, which was prohibitive for many people. Early in-home water heaters were expensive to purchase and expensive to maintain. They were inefficient and broke often. Our modern convenience has truly come a long way. It’s doubtful that we today would even recognize some of those initial models.

Nonetheless, the initial users were chastised. Water usage exploded during this time. It might come as no surprise to some, but cheap and readily available hot-water, once a luxury, eventually became seen as a necessity. The economic efficiency of hot-water production actually increased water usage. Despite being able to produce hot-water with far fewer resources, meaning fewer resources would be needed per-capita, overall water usage increased on net.

The observation of this relationship has occurred often enough that it even has a name in economics: Jevons Paradox. William Stanley Jevons first described this relationship after he observed that, despite the burning of coal becoming more efficient and individuals needing less coal to accomplish the same tasks, consumption of coal increased as adoption within the community could spread more easily. Just as with individuals finding more ways to use coal, families with in-home hot water heaters could now take daily showers, automatic dishwashers could become a thing, and laundry could take far less time to complete.

Fast forward to today, residential hot-water heaters continue to become more efficient, and adoption continues to expand. Heck, hot-water heaters have become so common that it’s possible to take a hot shower in the woods, far from any plumbing or utilities. It’s so common that what would have been a tragedy a century ago, the exhaustion of the entire hot-water reserve, can now be the subject of some light teasing. But we are certainly in a new phase, we’ll call it late-stage Jevons. Hot-water usage, while it continues to increase, has nearly leveled off. Just like the towel after my shower this morning, as a society, our hot-water adoption has been basically saturated. Hot-water heaters are installed in every new home. It would be exceptionally odd to purchase a new home or even to find an old home where one wasn’t previously installed.

Aye aye, AI

What was once a luxury for the wealthy, and then a gluttony for early adopters, is now the norm. There are many parallels to our present adoption of AI. What many forget, is that AI, at least our modern understanding of it with neural networks and back-propagated learning, has been around since at least the 1980s. Modern speech recognition and computer vision, with techniques that almost every Computer Science undergrad would recognize, was first researched around that same time. Well-funded universities and research labs have had access to this technology for a long time. Gradually and inexorably the technology became cheaper and easier to implement, and new uses were discovered.

Eventually, one use was developed that nearly anyone could use. It required nearly nil initial understanding to get started but was incredibly powerful with even a little training. Perhaps surprisingly, the use that was developed wasn’t substantively new. It was just a new way for humans to do something that we have been doing for millennia: chat.

ChatGPT was just a new implementation on a language model, GPT-3, OpenAI had developed internally and trained for many years. It had been trained on a huge set of diverse information. More specifically, ChatGPT had been trained on more conversational information and it gradually “learned” how to respond given a particular kind of input: a chat prompt.

With that, users like you, Brian, and me could open a chat and start a conversation, interacting with the language model in a way that felt natural and familiar. And with that, using AI language models had become vastly more efficient. And just as Jevons predicts and like the residential water-heaters of the past, AI adoption exploded. It was adopted across a huge array of tasks and industries. Business owners, like myself, could draft marketing copy without any previous marketing knowledge or experience. Entrepreneurs could create websites without any previous coding experience. Casual artists, authors, playwrights, and any number of other creatives, could explore their art in ways they couldn’t before. Indeed, I would argue that more art has been produced within the last two years than had been produced in the two decades prior to the release of ChatGPT.

Moving Forward by Walking Backwards

Of course, AI does have its detractors, and it has been the subject of many arguments between Brian and me. Brian believes that AI will ultimately lead to a societal collapse. Jobs will be lost, natural resources will be exhausted, and human social relationships will be replaced by artificial ones.

The thing is, I think he’s right. Jobs will be lost, natural resources will be exhausted, and real human relationships will be replaced by artificial ones. Where we disagree seems to be the scale and severity. Ask a plumber if they would suddenly have more work if residential hot water heaters were to suddenly vanish. Ask someone in the depths of their depression, anxiety, or guilt if they would happily give up the opportunity to talk about their mental health with a bot that doesn’t judge or abruptly end the session when time is up.

But I can almost hear Brian’s retort now. “How about let’s ask a Boilermaker? I won’t hold my breath waiting for you to find one.” “Have you heard some of the advice these AI psychiatrists have given? It’s sycophantic at best, arouses vanity somewhere in the middle, and incites narcissism at worst!” “Also, you said nothing about the rampant depletion of our natural resources to cool the data centers that feed this technology”.

To that I would say: jobs won’t be lost, people will adapt as they did before; we need better models, not fewer, and one specifically tailored to mental health (or whatever application); and the technology will improve, it will become more efficient and will use fewer resources. It will have to, because it’s not going away; for better or for worse, the genie is out of the bottle, and the monkey’s paw has been mummified.

But that’s not what this is really about. We’re talking past each other; implicitly agreeing with every disagreement. It’s both fascinating and exhausting, and I challenge you, dear reader, to listen- don’t speak, just listen- to two people arguing sometime. You can see it in Congress, on almost every talk show, and just about every time you see two people arguing about politics or religion.

Reviewing the previous argument, one can see every statement is a refinement or definition of terms and never a full rebuttal. Any good improve performer should notice the familiar “Yes, and…” pattern. This pattern takes some time to recognize consistently. Indeed, it’s taken most of my adult life to learn the signs. It’s as though we’re both moving forward, but one of us is walking backwards.

So, the question becomes, how do we break the pattern? Of course, despite what my wife, mother, and Brian, would say, I’m always right. But in these circumstances, they are too. I agree with every point Brian makes. And despite he or I calling it a disagreement, our goal is still the same. We want people to lead healthy and happy lives, gainfully employed or otherwise. We want our rivers, forests, and air to be clean, abundant, and full of life. We don’t have to agree on the effects of AI to agree that there are problems with AI. Whether there’s a leak in the line or not, the water is still gone.