This interview argues that setting big goals can actually prevent great discoveries, because the stepping stones to breakthroughs rarely look like the final destination. Computer scientist Ken Stanley shows with experiments that over 99% of final images passed through a completely different intermediate form—like a car's wheel coming from an alien face's eye. Instead of chasing goals, he says to follow what's 'interesting' and 'novel,' and keep things that seem useless but intriguing. The iPhone's success wasn't Steve Jobs' vision, but him recognizing all the stepping stones were in place. No specific stocks or funds are discussed.
Professor Ken Stanley, in his appearance on the Invest Like the Best podcast, presents a counterintuitive argument: setting grand goals actually reduces the likelihood of achieving great accomplishments. The core evidence is that nearly all major inventions had key precursor steps that were not crea
Ken Stanley is a computer science professor and pioneer in the field of neuroevolution, author of Why Greatness Cannot Be Planned. The core argument of this episode: Setting grand goals actually reduces the probability of achieving greatness — because the stepping stones to major discoveries almost never resemble the final objective, and goal-oriented mechanisms systematically prune away these critical stepping stones.
Ken Stanley argues that nearly all major inventions had key precursor steps that were not created with those inventions in mind. He cites the vacuum tube and the computer as an example: when the vacuum tube was studied in the 18th century, it had nothing to do with computing, yet it became a core component of the first computer (ENIAC, 1940s). If someone in 1850 had pulled vacuum tube researchers aside to "build a computer," both the vacuum tube and the computer would have been destroyed — because the stepping stone would have been removed.
This insight comes from Stanley's Picbreeder experiment: users iteratively explored an image space through a "select-and-breed" process. After two years, a retrospective analysis revealed that over 99% of the final images (such as cars, butterflies, and skulls) had passed through an intermediate form that looked completely different from the end result. For example, the wheels of a car originated from the eyes of an alien face — and the person who selected that alien face was certainly not looking for a car at the time.
> "The only way to discover any of these things was to not be trying to discover them."
Stanley attributes this to the "like-causes-like fallacy" — society assumes that the path to a goal will resemble the goal itself, but the true structure of the complex world is the opposite.
When voting/consensus mechanisms are introduced, generational progress is completely suppressed. The contemporaneous Living Image Project and Picbreeder used the exact same underlying encoding, with the sole difference being that each generation’s most popular images were selected by hundreds of voters. After 25,000 iterations, all results remained “rainbow-colored blurry blobs”—no meaningful images emerged.
Stanley points out that this directly mirrors real-world committee reviews (e.g., NSF grant evaluations): Committees naturally gravitate toward consensus, and consensus prunes away the radical stepping stones that only a very few can see the potential in. In Picbreeder, an image only needs to be deemed interesting by one person to serve as a starting point for subsequent exploration; in real-world systems, it must first pass majority approval from a five-person committee—which is nearly impossible.
> "A committee of five people is crazy in that case. Even if they're like experts in the fields, they just basically represent the status quo."
Stanley argues that for grand goals where "one does not know how to proceed," the only principled strategy is not randomness, but using "interestingness" as the selection criterion. This is not laissez-faire: the skulls, butterflies, and cars that emerged in Picbreeder were not random; they carried meaning and structure, and it is precisely this structure that makes them stepping stones toward other interesting things.
The specific recommendations are twofold:
1. Collect stepping stones: Actively seek out and preserve things that are "interesting," even if one does not know where they lead. This requires tolerating risk — not all interesting things will "succeed," but together they form a "portfolio of stepping stones."
2. Recognize the moment when stepping stones fall into place: When all prerequisites suddenly align, one must be able to recognize that "now it can be done." Stanley believes that the success of the iPhone was not Steve Jobs's "vision," but rather that he was the first to recognize that all the stepping stones were already in place; conversely, the massive investment in autonomous driving in 2016 was a case of "false vision" — the stepping stones were not yet in place.
Stanley emphasizes that novelty is a core component of "interestingness," and novelty requires a vast frame of reference: to know what is new, one must first know what is not new. This transforms curiosity and exploration from "indulgence" into "principled behavior."
Constraints are not goals, but they are the key mechanism for keeping stepping stones "interesting." Stanley uses natural evolution as an example: the constraint of evolution is "survive and reproduce," which essentially requires every organism to be a "walking Xerox machine"—a body containing a complex set of mechanisms capable of replicating itself. This constraint is extremely strict, but it ensures that all products maintain a high level of complexity, thereby providing fertile ground for subsequent innovation.
> "It's basically a Rube Goldberg machine generator. Every time you vary the Rube Goldberg machine, which is like this crazy convoluted thing that makes a copy of itself... They'll basically fill the space within that constraint of every possible thing you can imagine."
For an individual, the constraint is "what you are genuinely interested in"—it must be strict enough to maintain standards, but not so strict that it becomes a goal. Stanley's own approach is to "rebel against himself": predict what others think he will do next, and then not do it.
Stanley argues that the key to allocating innovation resources is not improving review criteria, but changing the review mechanism itself. The current system's problem is that committee reviews prune away 99% of the stepping stones, whereas what is truly needed is for these stepping stones to reach the very few who can see their potential.
His suggestion: significantly widen the filter from idea to exposure, allowing more people within the organization to see the ideas being generated; then, let those with a vested interest validate them – if one person is willing to use their own resources (time, budget, reputation) to support an idea that is not their own, that idea is worth investing in. This requires only one person to experience a "lightning bolt of connection," not a majority consensus.
Stanley acknowledges this requires courage, as superiors will ask, "How do you measure the effect?" But he believes this is precisely the purpose of conversations and books – to change mindsets first, then change mechanisms.
| Position | Guest Stance | Key Data |
|---|---|---|
| None | No specific investable positions discussed | — |
(This episode is a pure conceptual discussion, involving no specific companies or position moves.)
1. Ken Stanley: The key prerequisite step for almost every major invention was not created for that invention.
Support: Vacuum tube researchers in the 18th century studied electrical properties with no connection to computing, yet vacuum tubes were the core component of the first computers in the 1940s.
2. Ken Stanley: Setting ambitious goals reduces the probability of achieving great outcomes because stepping stones do not resemble the endpoint.
Support: Over 99% of final images in Picbreeder passed through an intermediate form that looked completely different from them; the car wheel originated from the eye of an alien face.
3. Ken Stanley: Consensus mechanisms (voting/committees) systematically eliminate radical stepping stones and are enemies of innovation.
Support: The Living Image Project used the same encoding as Picbreeder but was selected by hundreds of voters; after 25,000 iterations, it still produced only blurry blobs.
4. Ken Stanley: Replace goals with "interestingness" and use "novelty" as a compass—this is not indulgence but a principled exploration strategy.
Support: The skulls, butterflies, and cars that emerged in Picbreeder were not random; they had meaning and structure, and it was precisely this structure that made them stepping stones to other interesting things.
5. Ken Stanley: The success of the iPhone was not Steve Jobs' "vision" but his being the first to recognize that all stepping stones were already in place.
Support: The massive investment in autonomous driving in 2016 was "false vision"—the stepping stones were not yet in place, leading to widespread investment failures.
6. Ken Stanley: "Survive and reproduce" in evolution is not a goal but a constraint—it requires every organism to be a "walking photocopier," thereby ensuring all products maintain high complexity.
Support: If the constraint were removed (allowing everything to reproduce), the world would degenerate into "inert blobs covering the Earth."
7. Ken Stanley: Innovation resource allocation should let "those with a vested interest" validate ideas, not committee voting.
Support: If one person is willing to use their own resources to support an idea not proposed by themselves, that idea is worth investing in—it only takes one person to experience an "electric connection."
8. Ken Stanley: Novelty requires a vast frame of reference—to know what is new, one must first know what is not new.
Support: This transforms curiosity and exploration from "indulgence" into "principled behavior," rather than "wasting time."