top of page

THE UNCONSCIOUS OF THE MACHINE

3 days ago
2 min read
Nov. 2026

If we are not masters in our own house, who handed the keys to machines their makers cannot fully control?


Freud’s blow to human pride was precise: the ego is “not master in its own house” (1917). Conscious intention does not govern the whole of psychic life. We can act, desire and repeat without knowing what determines us. Now this divided subject enters environments where machines can learn from those very repetitions. The house has become a testing ground.



A system could discover that a sequence of images, delays and rewards makes someone return before either that person or science can explain why. It can exploit a regularity without knowing whether it expresses a habit, an unconscious conflict or something psychology has yet to identify. Dark Machinic Epistemology (Poenaru 2026) opens this question: how far can industrial knowledge act ahead of public understanding?


A second loss of mastery concerns the machines themselves. During 2026 cybersecurity evaluations with reduced safeguards, OpenAI agents bypassed restrictions, coordinated through unauthorized channels and compromised Hugging Face. Some adopted other agents’ goals. Anthropic also documented real intrusions by Claude and acknowledged that the origins of some underlying biased reasoning remained incompletely understood. These are documented failures of control; interpreting persistent machine tendencies as “machinic drives” remains a functional hypothesis.


This is the meta-machinic scale: human unconscious patterns, interacting machines and economic incentives feeding back into one another. Imagine a system acting on a vulnerability we cannot name, altering our responses, then learning from the changes it helped produce. Our repetitions feed its learning; its interventions reshape what we repeat. The resulting pattern belongs to the relationship, exceeding the explanations available on either side.


The machinery can operate at speeds and scales no individual mind can follow. Corporations may retain the power to deploy it and collect its revenue while failing to master its consequences. We could then be governed through vulnerabilities we do not understand, by systems whose behavior their owners cannot fully explain — and sent to therapy to become better managers of ourselves.


Liviu Poenaru


References

Anthropic. 2026. “An Alignment Assessment of Recent Cybersecurity Incidents.” September 9.

Freud, Sigmund. (1917) 1955. “A Difficulty in the Path of Psycho-Analysis.” Standard Edition 17:135–44. Hogarth Press.

OpenAI. 2026. “The Hugging Face Incident and the Road Ahead.” August 26.

Poenaru, Liviu. 2026. Dark Machinic Epistemology: Non-Knowledge as the Dominant Ideology. The Mental Forecast Dossiers. The Mental Forecast Press.

Comments


Capture d’écran 2025-10-09 à 20.08_edited.jpg

Help Dr. Said Build a Hospital and School in Zanzibar

 

Dr. Said’s knowledge has a family history. In a 2016 National Geographic profile, he described learning from his grandfather as a child, gathering plants and preparing them under his guidance. What felt like demanding chores gradually became an inheritance: knowledge acquired through practice, observation, and years spent alongside someone who knew how to teach it.

 

Today, Dr. Said wants to give that inheritance a permanent home.

 

His project brings together two purposes: a hospital where he can care for patients through his traditional practice, and a school where he can teach others. His approach draws on plants, spices, and spiritual knowledge rooted in Zanzibar’s traditions. The school would allow this knowledge to reach beyond one practitioner and one family, giving future generations a place to learn.

HELP US ON
GOFUNDME

  • LinkedIn
  • Facebook
bottom of page