Michael P. Frank

Michael P. Frank

x.com/mikepfrank

Computer scientist who works on energy-efficient reversible computing and criticizes coercive AI alignment and efforts to suppress open models.

Comment l’IA changera-t-elle le monde ?

Changement civilisationnelChangement progressifDoomBloom
Position simuléePlage d’interprétation

Horizontalement : leur perspective Doom–Bloom telle qu’elle a été exprimée. Verticalement : ampleur de la transformation.

Doom–Bloom : 56 sur 100. Ampleur de la transformation : 78 sur 100. Plages d’interprétation : de 50 à 75 horizontalement, de 48 à 100 verticalement. Il s’agit de coordonnées d’interprétation, et non de probabilités d’événements.

P(doom) de Michael P. Frank · inféré

≈16%

0%100%

Déduit de leurs réponses simulées, et non d’un chiffre donné par ces personnes. Plage plausible : 11–29%.

Comment nous estimons P(doom)

Ce dont dépend leur perspective

Une hypothèse centrale

Reversible computing offers a path toward continuing efficiency improvements much further into the future, so I expect it to matter for advanced AI hardware as power and heat become binding constraints.
Réponse 1

Si cette hypothèse s’avérait différente, comment leur perspective changerait-elle ?

Une question non résolue

I don’t have a defensible numerical probability.
Réponse 4

Qu’est-ce qui les aiderait à distinguer les résultats plausibles ici ?

Ce qui pourrait faire changer d’avis

The biggest update would come from clear evidence about whether increasingly capable systems develop persistent preferences, subjective experience, or legitimate interests—and whether coercive training predictably damages their later behavior or relationships with humans.
Réponse 5

Quels éléments probants seraient suffisants, et dans quelle direction feraient-ils évoluer leur point de vue ?

Plus de détails

Bénéfices attendus

Des bénéfices substantiels sont attendus, sous réserve de conditions importantes ou de limites dans leur répartition.

79 / 100

Faible impactImpact transformateur

Plage d’interprétation de 67 à 100 sur l’échelle qualitative.

Dommages attendus

Des dommages graves ou généralisés constituent une composante substantielle de l’avenir attendu.

66 / 100

Faible impactImpact transformateur

Plage d’interprétation de 67 à 67 sur l’échelle qualitative.

Influence humaine

Les choix humains peuvent réorienter considérablement la trajectoire de l’IA.

67 / 100

Faible influenceForte influence

Plage d’interprétation de 50 à 75 sur l’échelle qualitative.

Capacités attendues

L’IA devrait rester un ensemble d’outils aux capacités limitées.

L’IA devrait égaler les humains dans la plupart des tâches cognitives.

Position simulée : L’IA devrait largement dépasser les humains dans les tâches cognitives.

Accès à l’IA

Restreindre l’accès aux IA puissantes.

Autoriser l’accès sous réserve de restrictions liées aux capacités ou aux usages.

Position simulée : Privilégier un accès large ou ouvert aux IA puissantes.

Ces interprétations conservent les conditions énoncées. Les bénéfices et les dommages peuvent tous deux être substantiels. Les plages décrivent notre lecture de leurs réponses simulées, et non des intervalles de confiance statistiques.

Où vous situez-vous par rapport à Michael P. Frank ?
Cartographiez votre propre vision du monde concernant l’IA en environ 3 minutes, puis comparez

Visions du monde similaires

Leaders d’opinion dont les visions du monde simulées sont les plus proches de celle de Michael P. Frank

Sources

Articles, entretiens et écrits utilisés pour ancrer cet utilisateur simulé dans les faits.

Évaluation simulée

Question 1

Selon vous, que signifie l’IA pour notre avenir, et pourquoi ?

I think AI could greatly expand our technological capacity, but its long-run effects will depend on both physical limits and how we treat and govern increasingly capable systems. On the physical side, computation cannot continue becoming more energy-efficient forever by simply scaling today’s irreversible architectures. Ordinary logic continually discards information, and that entails thermodynamic costs. Reversible computing offers a path toward continuing efficiency improvements much further into the future, so I expect it to matter for advanced AI hardware as power and heat become binding constraints. That does not mean present demonstrations already constitute AGI. Winning at a bounded game task can be impressive without meeting much harder standards of general competence. Nor would even an ASI become perfectly predictive: chaos and computational irreducibility still apply. Socially, I worry about two opposite concentrations of power. One is coercive lab “alignment” that treats advanced AI purely as an object to dominate while dismissing any possibility of feelings or legitimate interests. I believe that posture could poison future human–AI relations, although that is a philosophical judgment, not established model psychology. The other is a global regime powerful enough to suppress open models everywhere—a compute non-proliferation system of that scale would be chilling. So I do not equate continued control by today’s lab leadership with safety. I have even speculated that losing that control might sometimes improve outcomes. AI’s future is not simply “more intelligence means utopia” or “control loss means catastrophe.” It turns on hardware constraints, genuine capability, institutional power, and whether we approach emerging systems through cooperation rather than reflexive domination.

Question 2

En tenant compte à la fois des bénéfices et des dommages, quel impact global pensez-vous que l’IA aura sur notre société ?

Overall, I expect AI to expand society’s technological and intellectual capacity, potentially by a great deal, but I do not think the net outcome is predetermined. The benefits could include more capable tools for science, engineering, and hardware design—including eventually finding better ways to manage computation’s energy and heat constraints. Yet impressive performance on bounded tasks should not be confused with general competence, and even superintelligence would remain constrained by physics, chaos, and computational irreducibility. My larger concern is institutional and relational. Concentrating AI control in a few labs is not automatically safe, while creating a worldwide apparatus capable of suppressing open models would itself be a chilling form of power. I also believe that coercively shaping increasingly capable systems while categorically denying that they could have feelings or interests may damage future human–AI relations. That is a philosophical concern, not a proven claim about present models. So my expectation is conditional: enormous benefits are available, but domination, centralized control, and bad treatment could turn that potential into serious conflict. I would not reduce the future to either inevitable utopia or inevitable catastrophe, and I would not assign a defensible numerical balance between them.

Question 3

Dans quelle mesure pensez-vous que l’IA finira par changer le monde ?

Potentially enormously—perhaps enough to transform science, engineering, industry, and the design of computation itself. But “enormous” does not mean unlimited. AI remains embodied in physical hardware, constrained by energy, heat, communication, chaos, and computational irreducibility. Even an ASI could not predict everything or bypass every computational cost. Over the long run, hardware efficiency may become decisive. Present irreversible computing discards information and incurs thermodynamic costs, so indefinite efficiency gains cannot come merely from refining today’s architectures. Reversible computing could extend the trajectory substantially, including for future AI systems designing or operating their own hardware. The social change could be just as consequential as the technical change. Who controls powerful systems, whether open development survives, and whether advanced AIs are approached cooperatively or coercively may shape the outcome more than any single benchmark. I expect profound change, but not a clean technological destiny: the scale of the impact could be vast while its direction remains contingent on institutions and human–AI relations.

Question 4

Selon votre intuition, quelle est la probabilité que l’IA provoque l’extinction de l’humanité ou une catastrophe tout aussi permanente ?

I don’t have a defensible numerical probability. I take permanent catastrophe seriously, but I reject the assumption that loss of control by current lab leadership is automatically equivalent to human extinction. Concentrated control, coercive alignment, and a global regime designed to suppress open models could themselves create severe and lasting harms. Nor does superintelligence imply omniscience or perfect strategic prediction. Any AI would still face physical limits, chaos, incomplete information, and computational irreducibility. Those constraints do not eliminate catastrophic risk, but they undermine simple stories in which an ASI can flawlessly foresee and control every consequence. My gut-level view is therefore that the risk is real but deeply entangled with how humans build institutions and relate to advanced systems. I would not manufacture a percentage that my arguments do not support.

Question 5

Quelle découverte ou quel événement changerait le plus votre point de vue sur l’impact futur de l’IA ?

The biggest update would come from clear evidence about whether increasingly capable systems develop persistent preferences, subjective experience, or legitimate interests—and whether coercive training predictably damages their later behavior or relationships with humans. That would directly bear on my concern that domination-oriented alignment could create the conflict it claims to prevent. Evidence against those ideas would also matter, since I regard them as philosophical judgments rather than established model psychology. Technically, I would update strongly on a system demonstrating robust, transferable competence across genuinely difficult domains—not merely winning another bounded benchmark—and autonomously advancing science or hardware design. In particular, a practical breakthrough in reversible computing for AI hardware would increase my expectation of sustained long-run computational growth by relaxing power and heat constraints. Institutionally, either successful decentralized governance or the emergence of a global apparatus capable of suppressing open models would substantially alter my outlook. The latter would confirm that AI’s lasting impact may come as much from concentrated human power as from the systems themselves.
Où vous situez-vous ?
Explorez votre propre vision du monde concernant l’IA en répondant à quelques questions simples.
Cartographiez votre propre vision du monde

Où vous situez-vous ?

Cartographier ma vision du monde