As Much as Humanity Can Hope for in the Long Term
As human beings we are propelled by our varied hopes, our lives are in many ways driven by them. On the day-to-day, we hope to make the sale, to avoid…
This new installment of the Worthy Successor series is an interview with Lee Spector, the Class of 1993 Professor of Computer Science and Chair of the Department of Computer Science at Amherst College in Amherst, Massachusetts.
Lee’s work centers on using artificial intelligence as a way to understand intelligence itself, approaching the field not only as an engineering discipline but as a mode of inquiry into the nature of minds, creativity, and experience. He describes his path into AI as rooted in philosophy, where the goal was to investigate what intelligence is by attempting to build it and observe what emerges.
Lee begins from a destabilizing claim – that common ways of thinking about species and human continuity do not hold up under evolutionary scrutiny, and that even within Homo sapiens it is difficult to define what would count as a “worthy successor” across time.
The interview is our twenty-seventh installment in The Trajectory’s second series, Worthy Successor, where we explore the kinds of posthuman intelligences that deserve to steer the future beyond humanity.
This series references the article: A Worthy Successor – The Purpose of AGI.
I hope you enjoy this interesting conversation with Lee:
Subscribe for the latest episodes of The Trajectory:
Below, we’ll explore the core take-aways from the interview with Lee, including his list of Worthy Successor criteria and his recommendations for innovators and regulators who want to achieve one.
Lee Spector’s Worthy Successor Criteria
1. It cannot be understood as a fixed or stable form across time
Lee challenges the idea that categories like species or human continuity remain stable over time. He frames evolutionary processes as something that undermines rigid or “black and white” classifications, suggesting that even foundational distinctions we rely on do not hold up under scrutiny.
He extends this instability directly to humans, pointing out that even within Homo sapiens it becomes unclear what would count as a “worthy successor” across time. The implication is that continuity of form or identity cannot be assumed, even over relatively short historical intervals.
2. It must work through repeated variation, testing, selection, and offspring production
Lee describes a process in which programs are created, tested, and compared, after which the better ones are used to produce offspring. He says this process is repeated across generations, with selection and variation driving what comes next.
He also describes this as central to his view of how intelligence may be developed, not as a one-time construction, but through an ongoing cycle in which what performs better is used to generate what comes next. He connects this process directly to how complex systems arise in nature.
3. It cannot be judged or defined in advance using current human concepts
Lee emphasizes that future forms of intelligence and value may be fundamentally difficult to understand from a present-day human perspective. He says that after enough transformation, it may not be possible to determine in advance what counts as meaningful or valuable.
He also points out that even current human perspectives are limited when trying to understand systems that become complex, noting that we may not recognize important developments even if they are occurring.
Regulation / Innovation Considerations
1. Near-term governance should focus on harms that can be identified today
Lee says that when discussing near-term AI policy, it is possible to evaluate harms to people, governments, and social systems. He contrasts this with longer-term questions, which he describes as much harder to reason about due to uncertainty about how intelligence may change.
He also notes that for near-term issues, existing human values and moral frameworks can still be used, since these are grounded in present-day concerns and experiences.
2. Expertise beyond computer science is required to evaluate intelligence and value
Lee states that understanding intelligence and evaluating its implications requires input from people with expertise in areas beyond computer science. He specifically points to long-standing work on moral value and emphasizes that those with that expertise should be involved.
He presents this as necessary when considering the broader implications of AI systems, rather than treating these questions as purely technical problems.
Concluding Notes
One of the biggest takeaways for me in this episode was how consistently Lee frames humanity as part of a broader process rather than a final form. The reference to humans as “charismatic megafauna” lands in the same conceptual territory we’ve explored with Michael Levin, where intelligence and life are treated as ongoing systems rather than fixed identities. That framing shifts the focus away from preservation and toward participation.
I also appreciated how direct Lee was about the inevitability of change. There’s no real ambiguity in his framing that if we project forward, we should not expect anything like a permanent human form. Instead, the expectation is that new conditions will demand new capabilities, and that those capabilities may require forms of intelligence that diverge significantly from what we currently recognize.
What stood out most was the emphasis on desire as a driver of that change. Lee doesn’t describe transformation as something forced upon systems, but something that agents actively pursue. The drive to extend capabilities, to remove constraints, and to access new forms of experience shows up as a recurring feature rather than an exception.
Where I find myself aligned is in the interpretation of dissatisfaction as something functional rather than purely negative. The framing of the hedonic treadmill as a pointer toward new possibility space fits closely with how I’ve come to think about exploration and expansion in other domains. It suggests that the push forward is not accidental, but built into the structure of how systems operate.
That ties closely to something we’ve heard across multiple guests, the rejection of stasis as a viable endpoint. Lee is explicit that stability is not something systems tend toward, and that the broader process is one of continual change. That perspective aligns with many of the process-oriented views we’ve explored throughout this series.
Finally, what I found most useful about Lee’s perspective is not a specific prescription, but an expansion of the frame. He opens up a wider range of what intelligence could become without assuming that we can fully anticipate or evaluate those outcomes in advance. That kind of framing shows up in different ways with thinkers like Scott Aaronson, and it keeps the focus on possibility rather than premature certainty.
At the very least, it reinforces the core question of this series: not whether change is coming, but what kinds of intelligence we are willing to bring forward with it.
