Eric Drexler: revision 1 to 3 — Future Human Wiki

Revision 1

Revision 3

categories: ["people", "nanomedicine"]

tags: ["nanotechnology", "molecular manufacturing", "grey goo", "foresight", "artificial intelligence"]

summary: "American engineer who proposed molecular nanotechnology, wrote Engines of Creation and Nanosystems, and later worked on AI strategy at Oxford."

summary: "American engineer who proposed molecular nanotechnology, wrote Engines of Creation and Nanosystems, and later worked on AI strategy at Oxford."

updated: "2026-07-27"

updated: "2026-10-29"

issues: ["Bill Joy's 2000 relinquishment essay is described without a citation."]

---

```infobox

allow manufacturing with atomic precision. His 1986 book *Engines of Creation* introduced molecular

nanotechnology to a general audience along with the [[grey-goo]] scenario; his 1992 technical

treatise *Nanosystems* attempted to establish the physical case. Neither the machines nor the

catastrophe has materialized, and the resulting argument over why is the central fact of his career.

catastrophe has materialized, and the argument over why has defined his career.

## Career

Drexler studied at MIT, initially working on space systems including solar sails, and completed a

doctorate there in 1991 under Marvin Minsky at the Media Lab — the first doctorate awarded in

doctorate there in 1991 under Marvin Minsky at the Media Lab, the first doctorate awarded in

molecular nanotechnology. His first technical publication on the subject appeared a decade

earlier, proposing that protein engineering could bootstrap a route to general molecular

machinery.[^drexler1981] With Christine Peterson he co-founded the Foresight Institute in 1986 to

## Molecular nanotechnology

The core claim of *Nanosystems* is mechanical rather than chemical.[^nanosystems1992] Biology

demonstrates that molecular machines — ribosomes, motor proteins, polymerases — work; Drexler argued

demonstrates that molecular machines such as ribosomes, motor proteins and polymerases work.

that machines built from stiff covalent materials such as diamondoid could work far better, and he

Drexler argued that machines built from stiff covalent materials such as diamondoid could work far

provided calculations of stiffness, error rates, thermal noise, energy dissipation and throughput for

better, and he provided calculations of stiffness, error rates, thermal noise, energy dissipation and throughput for

hypothetical bearings, gears and positioning devices. From these he derived the concept of a

[[molecular-assembler]]: a device that positions reactive molecular fragments to fabricate specified

structures, including copies of itself.

> [!note] Terminology

> "Nanotechnology" was coined by Norio Taniguchi in 1974 for precision machining at nanometre

> tolerances. Drexler adopted it in 1986 for something quite different — mechanical control of

> tolerances. Drexler adopted it in 1986 for something quite different: mechanical control of

> matter atom by atom. When American funding agencies took the word over in the late 1990s, they

> restored a meaning closer to Taniguchi's. Three senses of one term now circulate, which is a

> restored a meaning closer to Taniguchi's. Three senses of one term now circulate, a recurring

> recurring source of confusion in the literature.

> source of confusion in the literature.

The book's method is engineering analysis of the physically permissible, deliberately excluding the

The book's method is engineering analysis of the physically permissible, and it deliberately

question of how to build the first device. That exclusion is both its strength — it made a

excludes the question of how to build the first device. That exclusion let it make a falsifiable

falsifiable physical argument rather than a promise — and the source of every subsequent objection.

physical argument rather than a promise, and it is also the source of every subsequent objection.

## The Smalley debate

central figure in the American nanotechnology programme. Smalley argued in 2001 that mechanically

guided synthesis fails on two counts: a manipulator large enough to hold a reactive atom would be

too big to fit in the space where the reaction must occur, and an atom held by a manipulator would

adhere to it rather than transfer to the target — the "fat fingers" and "sticky fingers"

adhere to it rather than transfer to the target. These are the "fat fingers" and "sticky fingers"

objections.[^smalley2001] The two exchanged open letters, published in *Chemical & Engineering News*

in 2003.

The exchange settled nothing technically. Drexler answered that his proposals do not involve

grasping bare atoms, and that mechanosynthesis by controlled positioning of reactive molecular

tips is a different operation from the one Smalley was refuting; Smalley responded that chemistry is

tips is a different operation from the one Smalley was refuting. Smalley responded that chemistry is

not a matter of pushing pieces into place and accused Drexler of frightening children. What the

not a matter of pushing pieces into place and accused Drexler of frightening children. The exchange

exchange did settle was institutional. Smalley's position carried the day inside the United States

did settle the institutional question. Smalley's position carried the day inside the United States

National Nanotechnology Initiative, whose funding went to materials science, nanoparticles and

characterization tools rather than to molecular manufacturing.

}

```

> [!debate] Where the dispute stands

> The one formal adjudication came in 2006, when the National Research Council's triennial review

> of the American nanotechnology programme examined molecular manufacturing and declined to take

> either side: attainable error rates, speeds and thermodynamic efficiencies of such systems

> "cannot be reliably predicted at this time" from theory, and the committee recommended funding

> experimental demonstrations that link the models to results.[^nrc2006] No demonstration of

> positional mechanosynthesis of a useful structure has been produced since. The question Smalley

> raised therefore remains open on the terms the review set: whether an experiment can be made to

> match the calculations, which already exist.

## Grey goo and its retraction

*Engines of Creation* included a scenario in which self-replicating assemblers escape control and

consume the biosphere. Drexler introduced it as a hazard to be designed against, but it detached

from the book and became the popular image of nanotechnology, amplified by fiction and by Bill Joy's

widely read 2000 essay on relinquishment, which treated self-replicating nanomachines alongside

genetic engineering and AI as a novel class of [[existential-risk]].

genetic engineering and AI as a novel class of [[existential-risk]].[^joy2000]

Drexler subsequently argued that the scenario had been a strategic error on his part. With Chris

Phoenix he set out the case that efficient molecular manufacturing does not require free-floating

self-replicators at all — factory architectures with fixed machinery and external control are both

self-replicators at all: factory architectures with fixed machinery and external control are more

more practical and inherently incapable of ecological escape.[^dp2004] He has since described the

practical and inherently incapable of ecological escape.[^dp2004] He has since described the

grey-goo framing as an unnecessary distraction that made the field harder to fund and easier to

dismiss. The episode is a standard case study in how a risk illustration can capture a technology's

public identity, and in the difficulty of applying the [[precautionary-principle]] to a hazard whose

dropping the nanorobot imagery and emphasizing factory-scale production with atomic precision as an

economic rather than a medical proposition. At Oxford he turned to artificial intelligence, arguing

in a 2019 technical report that superintelligence is more likely to arrive as a collection of

specialized, bounded services than as a unified agent with goals — a direct alternative to the

specialized, bounded services than as a unified agent with goals. The report is a direct

agent-centred framing in Bostrom's *Superintelligence*.[^cais2019] The argument has been influential

alternative to the agent-centred framing in Bostrom's *Superintelligence*.[^cais2019] The argument has been influential

among researchers who find the classic agent model unrepresentative of how machine learning

systems are actually deployed, and it bears on [[artificial-general-intelligence]] more than on

anything in his nanotechnology work.

## Reception and legacy

Chemists have largely not adopted Drexler's programme, and no molecular assembler exists. What

Chemists have largely not adopted Drexler's programme, and no molecular assembler exists. In its

exists instead is a large and productive field of molecular machines built by chemists on their own

place is a large and productive field of molecular machines built by chemists on their own terms:

terms: rotaxane-based motors and switches recognized by the 2016 Nobel Prize in Chemistry,

rotaxane-based motors and switches recognized by the 2016 Nobel Prize in Chemistry,

[[ai-protein-design|computational protein design]] that now produces novel folded structures to

specification, [[dna-nanotechnology|DNA origami]]

specification, [[dna-nanotechnology|DNA origami]] devices, and

devices, and [[microrobots-in-medicine|magnetically steered microrobots]]. Medicine's nanoscale

[[microrobots-in-medicine|magnetically steered microrobots]]. Medicine's nanoscale successes, such

successes — [[lipid-nanoparticles]], [[targeted-drug-delivery|antibody-drug conjugates]] — arrived

as [[lipid-nanoparticles]] and [[targeted-drug-delivery|antibody-drug conjugates]], arrived through

through chemistry and immunology rather than through mechanical engineering.

chemistry and immunology rather than through mechanical engineering.

Drexler's supporters read this as vindication of the physical claim by other means; his critics read

Drexler's supporters read this as vindication of the physical claim by other means. His critics

it as evidence that the mechanical framing was the wrong abstraction from the start, and that the

read it as evidence that the mechanical framing was the wrong abstraction from the start, and that

bottleneck was never whether atoms can be arranged but how. The distinction matters for

the bottleneck was always how atoms can be arranged rather than whether they can. The distinction

[[medical-nanorobots]] and for [[respirocytes]], whose design studies assume the assembler

matters for [[medical-nanorobots]] and for [[respirocytes]], whose design studies assume the

Drexler described. Forty years after *Engines of Creation*, no experiment has demonstrated

assembler Drexler described. Forty years after *Engines of Creation*, no experiment has demonstrated

positional mechanosynthesis of a useful structure, and no experiment has shown it to be

impossible.

[^drexler1981]: `paper` Drexler, K. E. "Molecular engineering: An approach to the development of general capabilities for molecular manipulation." *Proceedings of the National Academy of Sciences*, 1981.

[^drexler1981]: `paper` Drexler, K. E. "Molecular engineering: An approach to the development of general capabilities for molecular manipulation." *Proceedings of the National Academy of Sciences*, 1981.

[^nanosystems1992]: `book` Drexler, K. E. *Nanosystems: Molecular Machinery, Manufacturing, and Computation*. Wiley, 1992.

[^nanosystems1992]: `book` Drexler, K. E. *Nanosystems: Molecular Machinery, Manufacturing, and Computation*. Wiley, 1992.

[^smalley2001]: `news` Smalley, R. E. "Of Chemistry, Love and Nanobots." *Scientific American*, 2001. {A magazine essay by a party to the dispute; the fat-fingers and sticky-fingers objections are argued, not demonstrated experimentally.}

[^smalley2001]: `news` Smalley, R. E. "Of Chemistry, Love and Nanobots." *Scientific American*, 2001. {A magazine essay by a party to the dispute; the fat-fingers and sticky-fingers objections are argued, not demonstrated experimentally.}

[^nrc2006]: `report` National Research Council. *A Matter of Size: Triennial Review of the National Nanotechnology Initiative*. National Academies Press, 2006. {A congressionally mandated review; its molecular-manufacturing section concludes that theoretical performance estimates cannot be reliably predicted and recommends experimental proof-of-principle work.}

[^joy2000]: `news` Joy, B. "Why the Future Doesn't Need Us." *Wired*, April 2000. {An essay by Sun Microsystems' chief scientist arguing for relinquishment of genetics, nanotechnology and robotics research; an argument about policy, not a technical analysis.}

[^dp2004]: `paper` Phoenix, C. and Drexler, K. E. "Safe exponential manufacturing." *Nanotechnology*, 2004.

[^cais2019]: `report` Drexler, K. E. *Reframing Superintelligence: Comprehensive AI Services as General Intelligence*. Future of Humanity Institute Technical Report, University of Oxford, 2019. {An institutional technical report from the author's own institute, not a peer-reviewed publication.}

[^cais2019]: `report` Drexler, K. E. *Reframing Superintelligence: Comprehensive AI Services as General Intelligence*. Future of Humanity Institute Technical Report, University of Oxford, 2019. {An institutional technical report from the author's own institute, not a peer-reviewed publication.}