Kan AI generere fotorealistiske billeder af menneskelige ansigter, der ikke kan identificeres som syntetiske ?
Afgiv din stemme — læs så hvad vores redaktør og AI-modellerne fandt.
Generativ AI har transformeret skabelsen af visuelle medier. Værktøjer som Stable Diffusion og DALL-E producerer nu billeder, der er umulige at skelne fra fotografier for menneskelige observatører. De håndterer lys, tekstur og følelsesmæssige nuancer med en overraskende realisme. Dette udvisker grænsen mellem det virkelige og det syntetiske og udløser debatter om autenticitet og misinformation.
Background
Generative AI has transformed visual media creation. Tools like Stable Diffusion and DALL-E now produce images indistinguishable from photography to human observers. They handle lighting, texture and emotional nuance with uncanny realism. This blurs the line between real and synthetic, sparking debates on authenticity and misinformation.
Current systems like DALL-E 3, Midjourney v6, and Stable Diffusion XL can produce highly photorealistic face images indistinguishable from real photos to most human viewers, especially when guided by prompt refinements such as “ultra-realistic, 8k, subtle skin texture, natural lighting.” These models achieve this fidelity by training on large datasets of licensed portrait photographs while incorporating adversarial filtering and diffusion-based refinement to suppress obvious synthetic artifacts. However, even state-of-the-art generators still exhibit subtle inconsistencies—such as improbable reflections, skin-pores that misalign with lighting, or teeth arrangements that violate anatomical norms—that can be detected under close inspection or by forensic analysis tools.
— Enriched May 12, 2026 · Source: Adobe Research
Foreslå et tag
Mangler et begreb i dette emne? Foreslå det, admin gennemgår.
Status senest tjekket May 15, 2026.
Galleri
Kan AI generere fotorealistiske billeder af menneskelige ansigter, der ikke kan identificeres som syntetiske?
Juryen fandt et klart bekræftende svar.
After thorough deliberation, the jury concluded that the current state of generative adversarial networks and diffusion models has indeed crossed the threshold into photorealistic human face generation indistinguishable from real photographs. The four jurors found no meaningful distinction between synthetic and authentic images when viewed by human observers, leaving no room for doubt. The evidence convinced even the most skeptical among them that this capability has arrived. The ruling: "A face in the crowd can no longer be trusted to tell the truth.
But the data is real.
The Case File
Across 2 sessions, 7 jurors have heard this case. Combined tally: 7 YES · 0 ALMOST · 0 NO · 0 IN RESEARCH.
Note: cumulative includes older juror opinions. The current session tally above is the live verdict.
By a vote of 4 — 0 — 0, the panel returns a verdict of JA, with verdict confidence of 86%. The court so orders.
"Advanced GANs achieve photorealism"
"Generative adversarial networks (GANs) and diffusion models produce nearly indistinguishable synthetic human faces"
"State-of-the-art GANs like StyleGAN3 can generate high-resolution, photorealistic human faces that are indistinguishable from real photos to human observers."
"State-of-art generative models can produce realistic faces 2020-01"
Individuelle nævningers udtalelser vises på originalengelsk for at bevare bevismæssig præcision.
Hvad publikum mener
Nej 20% · Ja 60% · Måske 20% 5 votesDiskussion
no comments⚖ 2 jury checks · seneste for 8 timer siden
Hver række er et separat jurytjek. Nævninger er AI-modeller (identiteter holdt neutrale med vilje). Status afspejler den kumulative optælling på tværs af alle tjek — hvordan juryen virker.
Flere i Creative
Kan AI generere en troværdig dokumentarstemme? — Status tjekket på 12. marts 2024 ?
Kan AI generere fuldalbums-længde musik i en valgt kunstners stil ?
Kan AI udføre et fjendtligt cyber-overtagelse af en nations elnet ved at udnytte nul-dags-sårbarheder identificeret og våbeniseret af en AI-agent på under 72 timer ?