In undergrad I took a semester of neuroethics — my favorite class. We talked about brain–computer interfaces and quickly agreed: for fairness, frontier neurotech should be for recovery, not enhancement. Neuralink has since put implantable BCIs into human trials, still aimed at paralysis, ALS, and functional restoration — not consumer enhancement. But I kept wondering: where is the real boundary of “fairness”? US inequality keeps widening; even without BCIs, some groups already win through policy, information, and capital. It feels a bit like universal values: we broadly assent to thin ideals, then fail to bind ourselves in practice.
We also covered free will (often framed in class as “science doesn’t support it” — still wide open in philosophy and neuroscience; Libet is contested), Kant vs utilitarianism, Nozick’s experience machine, animal-research ethics, and more. It felt a bit sci-fi then. Fast-forward: I think these are problems humans are about to face — or already face — while public discourse lags (if you’ve seen Don’t Look Up, you know the mood). I’ll define “posthuman” below; my personal view is that as AI accelerates, we can’t keep postponing the ethics — humans need to step into the posthuman. E. O. Wilson’s line: “We have Paleolithic emotions, medieval institutions, and god-like technology.” On the tech side: frontier AI training compute has grown roughly 4–5× per year since 2010 (~doubling every six months, Epoch AI). Emotions and institutions move on other clocks — affective circuitry on evolutionary timescales (MacLean’s “mammalian limbic” story often cites ~150 million years; modern neuroscience treats the triune-brain picture as oversimplified, but “emotion hardware ≪ chips” still holds), institutions patched by the century.
A friend recently put it another way: if the BCI is strong enough, you might not be able to hit “remove” — the coupled system becomes the agent, and the pre-implant self gets replaced, not upgraded.
Defining post-/transhuman
These labels get mashed together; serious discussion needs the split:
| Usage | One-liner | Voices |
|---|---|---|
| Transhumanist | Capacities beyond present humans, unreachable without new tech → that existence is “posthuman”; usually longevity, cognition, richer affect | Bostrom (2008), More, Kurzweil |
| Critical posthumanism | Not a new species name — a critique of “the human” as sole moral/epistemic center; power, bodies, who gets excluded | Haraway, Wolfe, Braidotti |
| Culture / SF | Uploads, digital minds, post-biological civilizations | Egan, Gibson, Vinge |
Bostrom’s is the most measurable: if a being meets or exceeds levels in at least one core capacity (healthspan, cognition, emotion) that today’s humans cannot reach without new technology, it is posthuman (Why I Want to Be a Posthuman When I Grow Up, 2008). The question: would you want to become that kind of being?
Hayles almost asks something else. In How We Became Posthuman, posthuman is not a future stronger-species tag — it’s a diagnosis of a subjectivity configuration already underway. Plainly: when we use computers/cloud as extended compute and memory, live through digital identity, and let AI steer daily life, we are already posthuman.
When I discuss carefully I keep transhumanism and critical posthumanism distinct. In the AI era I sometimes braid them — facing smarter-than-us systems, we both have to accept our “ordinariness” and upgrade through tech to stay plugged into the loop.
More formally:
The cultural–cognitive condition of recognizing that anthropocentrism no longer holds — forced by the need to interface with stronger intelligence — plus the ongoing human–tech coupling and capacity reorganization that follows.
I’ll still treat critical posthuman and transhuman in turn, then bring them back together.
Lineage: from postmodern posthuman to older decentering
Gould: the most important scientific revolutions share a pattern — repeatedly knocking human vanity off the pedestal of “we are the center of the universe.”
Older, wider anti-anthropocentric moves have been dragging us down that pedestal — not strictly chronological:
- Copernican — Earth is not the cosmic center. Wounds cosmological narcissism; template: humans are not nature’s telos.
- Darwinian — Humans are one branch on a continuous tree. Species boundaries leak; Singer’s speciesism argument grows from here.
- Freudian — The ego is not master in its own house. Subject split, unconscious, driven — the “rational agent” myth cracked early.
- Foucauldian / post-structural — “Man” itself is a historical configuration, not eternal essence (“death of Man” in The Order of Things). What gets decentered is the epistemic human subject.
- Anti-colonial / feminist — Enlightenment’s “universal Man” was often a marked man (class, race, gender). What gets decentered is who gets to stand in for humanity.
- Animal rights / environmental ethics — Singer, Regan, Leopold’s land ethic, deep ecology (Naess): interests or value not only in humans.
- Disability studies — “Normal humanity” is statistical and normative; who counts as deficient vs standard in therapy/enhancement talk is political.
- Cybernetics / extended mind — Macy conferences, Wiener: organisms as information and feedback (Hayles later: information “lost the body”). Clark & Chalmers: cognition sometimes does not end at the skull — notebooks, phones, today’s LLMs/agents enter the loop. Point is not “we are wet computers,” but: mind/person boundaries are functional coupling, not skin. (Nearby: substrate independence — if intelligence is realizable information process, it need not be bound to this meat; silicon, other biology, alien substrates all qualify. Closer to functionalism / AGI philosophy than to Clark’s original claim.)
- Extraterrestrial / SETI — We are likely not the galaxy’s only intelligence. Extends blow #1. The Drake equation is a parameter framework; in 1961 Drake and colleagues’ “informed guesses” put communicating civilizations in the Milky Way at roughly 10³–10⁸ (often collapsed to N ≈ L). Drake later often cited something like ~10⁴.
Freud sometimes called the first three humanity’s three narcissistic blows (Copernicus, Darwin, Freud); AI coupling often gets called a fourth — this time against cognitive monopoly (“thinking” does not happen only in human brains).
In the AI era these decentering blows leave the seminar. Concrete issues already on the table:
- Identity / legal personhood — Today’s AI agents are not legal subjects: you can’t sue them; they hold no assets. The European Parliament once floated “electronic persons”; mainstream legislation has moved toward no personhood, with liability on humans and firms. Corporate personhood is a fiction; whether AI should get something similar remains open — not settled fact.
- Legal consequence — AV crashes, agents placing orders, generative infringement: who pays — developer, deployer, user? The EU AI Act grades risk ex ante; a dedicated AI civil-liability directive proposal was withdrawn in 2025; fault still sits in national law patchworks; revised product-liability rules pull some AI harms toward “product” strictness. Gaps remain.
- Ethics and rights — If systems show preferences, suffering indicators, persistent goals, should AI welfare / moral status enter the agenda (cf. animal ethics, Singer-style substrate independence)? On the human side: algorithmic scoring, deepfakes, automated firing — whose basic rights get written into defaults. These get mashed into one slogan, “AI rights,” but one track is whether non-human subjects count, the other is whether human rights against AI systems get hollowed out.
- Agency and the accountability gap — The more “autonomous” the agent, the more post-hoc blame looks like a maze; alignment’s “whose values?” and governance’s “who signs?” are two faces of the same crack.
That’s why I think the discussion can’t wait: not the abstract “are humans the center,” but that in product specs, courts, regulation, and daily collaboration, nodes other than humans are already acting — whether you acknowledge them, how you split liability, how you allocate standing.
There are cultural/religious questions too: if critical posthumanism (Hayles / Haraway / Wolfe) really entered public discourse, what would it do to humanism and society? Is there a middle ground between nihilism and “humans are the crown of creation”?
The more interesting question: the transhuman
On the transhuman I have a few linked harder questions: whether we should, whether we can, and whether — for the AI era — we must push human capacity into the “unreachable without new tech” zone.
Should we / do we want to?
Classic objection (Sandel is clearest): don’t treat children (or yourself) as a product whose specs you keep upgrading. Part of life is “given gift” — height, temperament, talent aren’t items in a cart. If everything can be engineered “better,” the stance of accepting imperfection thins out into design and contest. Next to that: arms races once enhancement starts; coercion if you don’t upgrade; and whether the enhanced agent is still the same person who signed consent.
Classic case for (Bostrom 2008, More’s morphological freedom): if healthspan, cognition, and affective capacity can rise a lot while you are still you, that is often worth wanting for the person who lives it; glasses, education, and vaccines are already on an enhancement continuum.
My take: philosophically I’m more persuaded by Bostrom. Even without much biotech on hand, hasn’t present society already deeply objectified/engineered selves and kids in careers and status games — a lot of competition that is already anti–human-centered? Clinical neuroethics defaults to therapy first, enhancement careful (as in the opening class) not because Sandel won the philosophy tournament, but because realities stack: money and care go to patients first, eugenics history, FDA paths wrapped around disease indications.
Fairness isn’t only philosophy — history and social engineering matter. Who gets bandwidth and cognitive enhancement first rewrites power. If US inequality stacks with neural interfaces and bio-longevity, the unenhanced could become a de facto caste.
Can we?: the forward narrative vs 2026 hardware
The smoothest public ladder is: AI enables the transhuman — not “humans become silicon,” but “machines strengthen first, then lever biology and interfaces to lift carbon subjects.” Longtermism-adjacent writing often casts a “posthuman” future as what the AGI project is for (e.g. Greaves & MacAskill on strong longtermism; Bostrom Letter from Utopia frames posthuman flourishing as an end worth aiming at). Longevity rhetoric is already thick:
“For those in reasonably good shape and with reasonable means, I believe they will have access to longevity escape velocity by the end of 2030.” — Ray Kurzweil, Abundance Summit / Diamandis (2024). In a Guardian interview he puts LEV in the early 2030s and pushes nanobots / upload later.
Aubrey de Grey is more conservative but still radical: ~50% chance of LEV in the late 2030s (repeated in recent interviews; CEO Today / LEV Foundation). LEV Foundation’s Robust Mouse Rejuvenation Study 1 closed as a “qualified win” — additive damage-repair, but lifespan gains ≈ calorie-restriction scale, short of his +12-month middle-age target (May 2025 update).
More detail:
1) AI → drugs / aging biology
| Node | Where we are (~2026-07) | Hype line | My read |
|---|---|---|---|
| Structure → medicines | AlphaFold lineage; Isomorphic Labs (Alphabet) saying first human trials are close — Hassabis at WEF 2026: before year-end (had been 2025; slipped ~a year) | “solve all disease” | AI-designed drugs into Phase 1: high; big life extension for healthy young people before 2030: low |
| Partial epigenetic reprogramming | Life Biosciences ER-100 FDA IND; Phase 1 for eye disease (glaucoma / NAION) from 2026 Q1 — widely reported as first human partial-reprogramming trial (LifeSpan.io); Altos Labs (~$3B) still platform-heavy, CMO in place; systemic anti-aging human trials maybe 2027–29 before they’re real | ”rewind the cellular clock” | Local organ safety data in coming years: plausible; systemic anti-aging products before 2035: still speculative |
| Senolytics / multi-morbidity | Small human trials; multi-disease paths (e.g. metformin / TAME-class) still fragmented | Combo therapies explode | Modest healthspan gains possible; ≠ LEV |
Read: AI shortens the front end of target-finding / molecule design; review and biological noise still run on years. The forward narrative swaps “faster discovery” for “humans already upgraded” — a whole clinical layer missing in between.
2) Invasive BCI (therapy ladder → implied enhancement)
| Player | Where | Bandwidth / capability | Path |
|---|---|---|---|
| Neuralink | N1 in humans on the order of ~20+ (public reports ~26; US/Canada/UK/UAE); PRIME (cursor/devices), CONVOY (arm), VOICE (speech); Blindsight still expanding toward humans | Mostly reading motor intent; demos: computers/games/early speech | All IDE; company materials often sketch ~2029 FDA commercial (paralysis); no healthy-enhancement pathway |
| Synchron Stentrode | Endovascular via jugular, no craniotomy; COMMAND feasibility: 6 people, 12 months no device-related SAE; Apple BCI HID → iPhone / iPad / Vision Pro | ~1–4 bit/s ITR class, below intracortical; enough for pointing, not “merge” | Targeting 2026 pivotal → if clean, commercial often sketched 2028–29 — maybe first to sell a “therapy product” |
| BrainGate / academic | Nature 2023 speech decode 62 WPM (125k vocab, error rates still high; natural speech ~160 WPM) (Willett et al.); later NEJM case ~32 WPM but more stable | Communication ceiling near speech, still semantic strings — not cortex-scale AI I/O | Research-grade |
| China Neuracle NEO | Reports of NMPA commercial approval Mar 2026 — first commercial BCI-class node of that kind (therapy indication; trust primary Chinese documents) | — | Gate opens earlier than FDA; ≠ Western consumer enhancement |
PLOS Biology 2024 is cold on enhancement BCIs (eBCIs): invasive cognitive enhancement is still mostly prospective; ethics go “far beyond practical pros and cons.” A companion notes most current neurotech boosts are small lab effects (PLOS 3002831). Longterm-Wiki-style surveys put “meaningful cognitive enhancement” at 2040+? / very low confidence, and treat BCIs as nearly irrelevant to TAI timelines — bandwidth gaps of orders of magnitude (62 WPM vs AI throughput).
My invasive odds:
- ~2028–31: commercial medical BCI for paralysis in US/EU (at least one) — medium-high (Synchron looks more like “sell safe bandwidth first”)
- ~2030–35: scaled legal “cognitive enhancement” implants for healthy volunteers — low; regulation, surgery, chronic biocompatibility, indication labeling all block
- Bandwidth jump (orders of magnitude) to hedge ASI / stay principals — very low if AGI lands around 2030; this is where the forward narrative is brittle
3) Soft enhancement: VR/AR, peripheral “neural” input, consumer EEG
Sold as “we’re already cyborgs,” mostly outsourced senses + muscle decoding, not cortical rewrite:
- Apple Vision Pro + Synchron: BCI as accessibility input, not healthy cognitive boost.
- Meta: Orion AR prototype + Neural Band (EMG wrist) — arm-muscle intent for UI (CTRL-labs path); shipping with Ray-Ban Display–class products from 2025. Great UI, not brain–machine broadband.
- Consumer EEG (Muse / Emotiv / Neurosity): meditation/attention feedback, millions of units sold (industry reports); ≠ reliable thought-to-agent channel.
- Manifold-style markets long bearish on “cheap portable general-purpose non-invasive BCI by 2030”; medians for “million-user consumer non-invasive BCI” often slide into the late 2030s once definitions tighten.
My take: “Soft enhancement” (always-on AR, AI copilots, EMG/gaze/voice multimodal) by 2030 — high. That already reshapes productivity and attention; calling it “transhuman” dilutes the word. Things that actually rewrite time preference / emotion hardware — near empty.
| Claim | Rough odds |
|---|---|
| AI materially speeds some drug discovery; major disease wins in the 2030s | High |
| Wealthy buy modest healthspan gains first (≠ LEV) | Medium-high |
| Kurzweil-style “LEV by end-2030 for people with means” | Low–medium-low (de Grey himself only gives late-2030s ~half) |
| Invasive BCI as mass “merge with ASI” channel on AGI clocks | Very low |
| Soft AR/AI “exoskeleton” ubiquity; power and attention stratify further | Very high — closer to capitalist enhancement than species transition |
Do we even get a real choice?
That happy path sounds fine. As AI closes in, what’s surfacing isn’t only new tech risk — it’s philosophy, society, and ethics we never seriously resolved. Even though (in the critical sense) the posthuman has already arrived, we still haven’t properly asked who we are, what we’re bad at, what “humanity” means.
More concretely:
- Bandwidth / symbiosis — Human intent output is tiny; to a fast AI it’s like talking to a tree. Musk–Neuralink: raise cortex–digital bandwidth so collective will can keep up — or even a well-meaning AI factually decouples.
- Decision can’t match the explosion — Forethought: governance / distribution / arms challenges from an intelligence explosion can’t all be outsourced to future AI — “how much to delegate, and whose values” is itself a human-side bottleneck.
- Make key humans smarter as “insurance”? — One story: if we can’t guarantee control of strong AI, at least strengthen the humans who matter — smarter, better able to talk to machines — before takeoff hits with today’s brains. MIRI FAQ lists biological cognitive enhancement and BCIs among paths to superhuman intelligence. Sounds like insurance: protect against “humans too weak.” Two red flags — (a) the same move can make the capability race faster; (b) “we must enhance a subset of people” invites eugenic rhetoric. Motives, not a moral free pass.
On bandwidth, Musk already sounds less like gift narrative and more like forced interface change:
“If the AI can communicate at terabits per second, and you’re communicating at bits per second, it’s like talking to a tree.” — Elon Musk, Lex Fridman #438 (2024)
I increasingly want the story flipped:
Not “AI gives us a chance to become transhuman,” but “we probably have to become transhuman in some sense to prepare for AI.”
By prepare I mean:
Try to remain principals with real veto — understand a bit, verify a bit, refuse, rewrite the contract.
Pure biological sapiens + keyboards + parliamentary tempo is unlikely to steward god-like intelligence.
Kevin Kelly’s Out of Control (1994) already said this century machines would get more biological and biology more engineered; in human–machine coevolution both rewrite each other’s fitness landscapes, and “control must be shared as the system grows.” Choosing enhancement may also hand over central control faster — maybe a reality we have to face.
Open question
Coupling strength — read-only vs write; detachable vs woven in. Once woven, who is the agent? Theseus’s ship vibes — but more than a thought experiment, can we keep an off switch / reversibility and a protected right to withdraw?
Sources
- Kelly, K. (1994). Out of Control (born/made merger, coevolution, hive mind).
- Wilson, E. O. (2012). The Social Conquest of Earth (“Paleolithic emotions / medieval institutions / god-like technology”).
- Musk, E. / Neuralink — bandwidth & symbiosis: Lex Fridman #438.
- Synchron — COMMAND / BCI HID: Business Wire 2025-05.
- Willett et al. (2023). A high-performance speech neuroprosthesis. Nature (BrainGate ~62 WPM).
- PLOS Biology eBCI ethics; companion neurotech enhancement.
- Kurzweil, R. — LEV ~2030 with means (Diamandis); Guardian 2024 interview.
- de Grey / LEV Foundation — RMR1 “qualified win” (May 2025 update); LEV ~50% late 2030s interview framing.
- Isomorphic Labs / Hassabis — first AI-designed drug trials (~end 2026; had been 2025).
- Life Biosciences ER-100 — first partial-reprogramming human IND (LifeSpan.io).
- Meta Orion / Neural Band (EMG) — peripheral intent → AR UI, not cortical broadband.
- Forethought. Preparing for the Intelligence Explosion.
- MIRI. Intelligence Explosion FAQ (bio enhancement / BCI paths).
- Bostrom, N. (2014). Superintelligence Ch.2–3 (limits of bio vs machine).
- Bostrom, N. (2008). Why I Want to Be a Posthuman When I Grow Up.
- Bostrom, N. & Sandberg, A. (2009). Ethical Issues in Human Enhancement.
- Sandel, M. (2004). The Case Against Perfection.
- Haraway, D. (1985). A Cyborg Manifesto.
- Hayles, N. K. (1999). How We Became Posthuman.
- Wolfe, C. (2010). What Is Posthumanism?
- Braidotti, R. (2013). The Posthuman.
- Clark, A. & Chalmers, D. (1998). The Extended Mind.
- Clark, A. (2025). Extending Minds with Generative AI.
- Nozick, R. (1974). Anarchy, State, and Utopia (experience machine).
- Putnam, H. (1981). Reason, Truth and History (brains in a vat).
- Singer, P. (1989). All Animals Are Equal.
- Foucault, M. (1966/1970). The Order of Things (“death of Man”).
- Leopold, A. (1949). A Sand County Almanac (land ethic).
- Gebru, T. & Torres, É. (2024). The TESCREAL bundle.
- PLOS Biology (2024). BCIs for cognitive enhancement.