Whenever I find myself a little “unhinged” in life, I find solace in reading. As I scanned my bookshelf and Kindle library for my next book, I realized that I need to be a little more intentional about what I’m reading.

While I have read some of the larger sci-fi series, I haven’t really covered the greats (and the works that made them great): Asimov, Heinlein, Clarke, Stephenson, LeGuin. Hence, I’m embarking on a SciFi binge for a while. One of the first books that I read is “The Moon is a Harsh Mistress” by Robert Heinlein.

Warning: Spoilers abound, though I’ll try not to give too much away of the main plot.

The hardcore Sci-Fi angle: A(G)I in all its glory

Reading this book today for the first time gives a much different experience than it would have even 4-5 years ago. Much of the plot hinges on an AI (AGI?) that essentially emerges as a consequence of unintended events.

I found it mind boggling to recognize concepts that are instrumental to birthing the actual AI world today. The book was written in 1966, and yet it effortlessly predicts so much of the real mechanisms of AI over the last 3 years.1

  • RLHF (Mannie vetting the list of jokes to teach Mike what is funny v/s not): there is literally a passage where the protagonist muses that it’s hard to tell a good joke by oneself, but easier to judge a good joke
  • the concept of “scaling” (Mike “awakens” by being augmented with components willy-nilly) - and the events at the end of the book leave one wondering what happened
  • Different personas for different users (Mike / Michelle / Adam Selene)
  • The very concept of an LLM (Mike evolves as he gathers more data and social cues )

One of the most “human” interactions was the most touching for me. It’s buried in the thick of action, but it made me pause even when so much was happening. In a situation where Mike (the AGI) is controlling a lot of moving parts (and is hence stretched thin) - he asks Mannie, “his first and most dear friend”, to go away, to avoid getting distracted. The reason? Because Mike expends more than necessary compute on having a conversation with Mannie - because he wants to make sure that he doesn’t offend his first and most dear friend”. He goes through multiple scenarios in his head of what to say and exactly how to say it.

This is such a sweet parallel to what humans (at least some of us) do - obsess over the tiniest of things to say in order to maintain relationships. It made Mike more than just a machine - he understood that relationships are paramount, and worth really investing in.

One overarching theme will stick with me though - just how much is possible with one AGI + a handful of driven humans dedicated to a cause. Heinlein explores how society itself could be vastly different with a different set of constraints.

The TANSTAAFL angle: Life on the Moon

Loonies (as they call themselves) face many contradictions and a completely different set of constraints from those on Earth. This makes the Lunar society evolve in a unique direction: their (unofficial) motto is “TANSTAAFL”: There Ain’t No Such Thing As A Free Lunch. Having studied at IIMA, this was a welcome callback. Everybody pulls their weight and there are no handouts (there simply aren’t any extra resources in such a constrained environment). Everyone has to adhere to a strict but informal system - there simply cannot be any waste, as the system itself would collapse. Anybody who doesn’t pull their own weight, or tries to extract more than their due, is simply eliminated.

Furthermore, the moon also has a scarcity of women. The whole concept of marriage and interaction of the sexes has been redesigned so that all resources are shared without jealousy, in a consensual manner, for the good of the greater collective. Marriage is like a contract - protecting rights, sharing resources and building redundancy.

The revolution starts when people articulate where TANSTAAFL breaks: Luna is exporting resources to the Earth in the form of grain, and not getting anything back. Mike calculates that Luna will essentially implode in < 7 years if the drain continues.

Then, through the revolution, Heinlein has us indulge in the fantasy - what if you got to design a completely new society, with a whole new set of constraints, distinct from what you are used to today? Prof asks the new Congress to really imagine why any laws are needed at all. Why should the government be told what to do? Why not just prescribe “what must government never intervene in”, and leave it at that?

One quote that really made me pause and think was

“Under what circumstances is it moral for a group to do that which is not moral for a member of that group to do alone?”

“MiaHM” gave me more to chew on than most books I’ve read in years. This is what old books offer: ample food for thought wrapped up in a story to get lost in. The broader narrative around constraints - whether it be in the development of intelligence (human or artificial) as well as the evolution of entire societies - is thought provoking.

Put differently, constraints lend themselves to “first-principles thinking” - a skill not easy to develop, but invaluable in today’s world. One final quote in this same vein of constraints - simple but worth sitting with

“When faced with a problem you do not understand, first do any part of it that you do understand, and then look at it again.”

Footnotes

  1. It’s remarkable that these concepts were imagined nearly 60 years before ChatGPT launched. The parallels between Mike’s development and modern LLMs—emergent behavior from scaled systems, learning from interaction, RLHF-like feedback loops—feel almost prophetic. ↩