Why Greatness Cannot Be Planned

Ch 1. Myth of the Objective. (Love a book that starts with a quote from Moby Dick, ‘My means are sane, my motive and my object mad’). What if you didn’t live your life trying to achieve objectives? EG Find a date, make a sale, win the game. Could you achieve things without objectives? How does having an objective limit our ability to explore? Objectives lead to measurement, and invention of a progress gradient. That makes sense when steps are knowable. But does it make sense when objectives are ambitious? You have to explore to find an ambitious objective, thru possibility space / research space. Can you ever really know how to achieve an ambitious objective? OK then, map out the steps to inventing; a new kind of music. Time travel. Holographic TV. World-changing inventions are not one-shots but rather step-by-step exploration and improvements (the ENIAC computer didn’t launch us to the iPhone or AI. But it was a step to get there.) Maybe a big achievements is less likely when you are working toward an objective. The world of really new, really ambitious inventions are too contingent.

Ch 2. Victory for the Aimless. Lots of stories of unplanned success, serendipity, the right to follow a passion, etc.

Ch 3. The Art of Breeding Art. Authors were/are AI researchers and had an early project called PicBreeder. They think of it as ‘Genetic Art’ because users generate a series of images and then pic one or more to ‘breed’ forward. Lots of examples. It starts out with random shapes/colors/gradients. Users ‘discover’ novel images (a butterfly, a car, a face) over time. But you can’t AIM for breeding a car, it’s too contingent on chance. You can’t plan out your stepping stones in advance.

Ch 4. False compass. Lots of stepping stones can’t be predicted, or they turn into a Chinese finger trap (eg, you gotta back out of what you were doing to get free). Matt thinks: guys just explain the concept of a local maximum, it’s more clear than this. Evolution has ’survive and reproduce’ as its stepping stone criteria. Matt says again: guys stop assigning a GOAL to evolution. “Almost no prerequisite to a new invention was invented with that invention in mind.”. Almost catchy. Unambitious objectives, which are only a couple steps away, are fine, good, but that’s only when you can see the pathway ahead.

Ch 5. The Interesting and The Novel. So if you can’t plan steps to achieve a goal, what can you do? A good stepping stone is Novel, which is a clue about its interestingness. Novelty search is a kind of non-objective search (I guess? Isn’t novelty search its own objective? Not really say the authors, because you can never be done being novel, as soon as you think you’ve achieved it it’s not novel anymore. Harrumph says matt). However, serendipitous discoveries don’t happen randomly, they’re more likely to happen to people who are alert, intelligent, and prepared.

Ch 6. Long Live the Treasure Hunter. Not everything is possible. Non-objective search is a powerful treasure hunter… just not for any particular treasure. To do this with people, don’t group them into teams and committees. Just let them work independently. They give some examples about furniture that adapts to your use that makes my industrial design brain queasy.

Ch 7. Unshackling Education. Guess what, objective thinking hurts education. Campbell’s law => the more you use a metric for decision-making, the more opportunity for corruption. (Eg, if you reward a measurement, it will no longer be a good measurement). Eg ‘teaching to the test’ instead of testing what students know. Providing incentives for outcomes reduces the diversity of outcomes and remember diversity of outcomes is part of ‘novelty’ (I guess?) Here they give many more examples from Picbreeder and here is where I lost my patience with this, they have ONE metaphor they keep throwing at this and it’s just not great or very convincing.

Ch 8. Unchaining Innovation. Funding scientific breakthroughs is important, because of the incredible expected value of a true innovation. But assessing which research proposals are likely to lead to an ambitious break thru is tough. Once we could look at the map and say ‘what about there, we haven’t explored there yet’, but conceptual science and invention doesn’t work that way. Funding grants is done by a consensus of peers, a surefire way to NOT fund truly original work. Perhaps a better way would be anti-consensus, eg proposals that get strong reactions in both directions. Then some boring pic breeder examples.

Ch. 9 Farewell to the Mirage. Control over Objectives (at least ambitious ones) was always a mirage. So you’re not giving up your objectives. You just need to realize they were never attainable in that way anyway.

Ch. 10. Case Study 1. Natural evolution. Guess what, evolution/natural selection doesn’t work toward an objective but still achieves incredible results. Sure, OK, whatever, but Matt says: remember teleology is not allowed when you talk evolution!

Ch 11. Cast study 2. AI. More pic breeder / etc. not interesting.

Noted on January 14, 2025

Hate to say it, but should've been a blog post.

Noted on January 14, 2025

A rare request from the BNB Book Report series! Still reading it now, will transfer in my notes here.

Noted on December 21, 2024