Od rekonstrukcji do kreacji: generatywna AI w pracy z obrazem archiwalnym [Panel III]

Od rekonstrukcji do kreacji: generatywna AI w pracy z obrazem archiwalnym [Panel III]

From reconstruction to creation: generative AI in working with archival image [Panel III]

Title history
  • Od rekonstrukcji do kreacji: generatywna AI w pracy z obrazem archiwalnym [Panel III] — observed on 20261102-1025 UTC (YouTube)

by IPNtv Konferencje · View the channel on PSI

Humanities, Social Sciences & Thought AI — Models & Research UYQArtificial intelligence (AI)UYQFGenerative artificial intelligence
🎙 IPNtv Konferencje 👥 4K 📅 November 2, 2026 ⏱ 75 min 👁 2 📄 expert opinion 🧭 2026-11-02
Available in: English (current) Français

Keywords

AIarchivescolorizationreconstructiondigital history

Summary

This conference panel, part of a larger event, focuses on the application of generative AI to historical photographs and films. The first speaker, Dr. Marcin Wilkowski, a historian and programmer, outlines different levels of processing, emphasizing that digitization itself is an interpretative act. He discusses automatic and language-based colorization methods, noting the lack of historical training data and the risk of anachronistic biases, as exemplified by models trained on modern images. Wilkowski also examines animation techniques like parallax, deepfakes, and frame interpolation, and warns about the ‘presentification’ effect that strips archival materials of their historical context. He proposes algorithmic criticism as a defensive measure. The second speaker, Mariusz Zając, a professional photographer, shares his practical journey since 2020 using AI tools to colorize and restore pre-war Warsaw photographs and films. He demonstrates the evolution of tools like DeOldify and discusses their limitations, while also highlighting how his work gained public attention. The panel concludes with a discussion on the risks of AI-generated images infiltrating archives and the need for education and detection mechanisms.

173 words

Critical Evaluation

Value of the Information & Strength of the Argument

The value of the presentation lies in the complementary perspectives: Wilkowski provides a systematic theoretical framework for understanding the epistemic implications of AI on historical sources, while Zając offers concrete examples and hands-on experience. The argumentation is solid, relying on cited tools and concepts such as ImageNet, DeOldify, and the notion of ‘cognitive contamination.’ However, some claims are anecdotal and based on the speaker’s personal observations, which limits the generalizability. Still, the reasoning is coherent and well-illustrated with practical examples, making a compelling case for cautious and critical engagement with AI in archives.

Scientific Rigor, Source Quality, Title Accuracy

Scientific rigor is adequate: both speakers are credible professionals, and their statements are grounded in known tools and research. The discussion references academic concepts like ‘presentification’ and ‘algorithmic criticism,’ though no formal citations are provided with URLs. The title accurately describes the content, and the panel stays on topic throughout. The lack of accessible external sources from the description reduces the ability to verify claims directly, but the internal consistency and expertise lend credibility.

182 words

Title / Content Match

The title accurately reflects the content, which explores the use of generative AI on archival images, covering both theoretical implications and practical applications.

Quality & Reliability

7/10

The panel features established experts in history, programming, and professional photography. They provide a nuanced discussion grounded in both theoretical frameworks and practical experience. However, the content is not peer-reviewed and relies partly on personal insights and examples, limiting its formal reliability.

Key Moments

Markers derived by PSI from the transcript: the creator did not define chapters.

Contribution & Novelties

The video contributes a rare dialogue between theoretical critique and practical application of AI to archival images. Wilkowski’s framework of ’levels of processing’ and the concept of ‘algorithmic criticism’ offer novel pedagogical and methodological tools. Zając’s demographic-scale experience provides real-world benchmarks and reveals the rapid evolution of accessible AI tools.

Pour aller plus loin :

  • Image colorization — Provides foundational context on the technical processes and challenges of colorizing black-and-white images.
  • Deepfake — Explains the technology behind face-swapping and animation, relevant to Wilkowski’s discussion of animation techniques.
  • Digital humanities — Offers an overview of the field that encompasses the critical use of computational methods in historical research.

107 words

Radar Profile

The radar profile shows high scores in quantity and quality of information, with moderate technical level and reliability. This reflects a well-rounded expert discussion that is informative and technically accessible, while the reliability is slightly lower due to the lack of formal citations and reliance on practical experience.

Reliability 7/10