Preserving AI for Social Good — How AI can Preserve History

Original publication on Medium: Preserving AI for Social Good — How AI can Preserve History | by Sam Bobo | Feb, 2025 | Medium

History manifests itself across multiple facets of the human experience. People desire memorability once they pass away, society yearns to document the past for recollection and contextualization, and the human condition itself gets imbued on a timeline marked by historical milestones that build into the society today. Throughout the fabric of time, change transpires, juxtaposed by the loss resulting from it. Our world consists of nearly infinite instances of the human experience, from each individual and their unique story to the collective whole of a society, to the events of nature that unfurl. I’ve been watching a YouTube channel called Yes Theory that seeks to expose the human experience by “Seek[ing] Discomfort” and embracing the human condition through travel and pure immersion within society, from talking to locals, hearing their stories, and embracing their culture.

Too often our minds enter fight or flight mode with the unknown, and the same goes with Artificial Intelligence. There continues to be ongoing discourse on the malicious use of AI, replacement of jobs, and regulation around use in society. Business people seek to make capital gains on its investment and practitioners such as myself continue to push the limits. One area we forget to embrace are the societal benefits. I am not talking about productivity gains, rather, the social good aspect of AI. I’ve spoken previously about accessibility using AI, specifically in voice accessibility:

Reading Assistance (Visual Impairment): Voice AI can read out text from books, websites, or other sources for visually impaired individuals or those with reading difficulties.

But today I wish to spotlight another area of AI for social good — preservation. As I opened within this piece, time forces upon change, both predictable change and unpredictable. Most notably, as I continue to claim, AI’s strength is about understanding niche at scale, and with 8.2 billion people in the world on average, there are certainly niches to be explored.

Endangered Languaes: Take the Microsoft Language Bank, a project derived from a 2022 hackathon by colleges of mine within the AI Speech organization. They sought to use text-to-speech technology to record audio and written transcription of endangered languages in China, including Inuktitut, Kurmanji, Hakka, and Minnan Chinese. From their research, 40% of the world’s languages are endangered of being forgotten and with the use of AI, we can try to memorialize and preserve them through sophisticated modeling. (Another article outlining its importance)

Lost Human Memory: Another impactful text-to-speech product includes the preservation of loved ones. Many whom have experienced a loss hold on to precious memories of that individual and long to hear their voice again. People keep voicemails of passed loved ones simply to hear their voice and replay over and over again. Imagine a world where a voice model was built around that person either during their lifetime or after based on historical data. Could you also create a model that encapsulates that person, similar to use cases for historical figures used in Education or Gaming industries, so future generations can ask questions about their long lost relatives. I’ve had family members, my uncle specifically, who interviewed my great grand aunt at 106yrs of age (she is 108 now) to get her life story and ask probing questions about her history. Imagine that at scale.

Yes, these preservation techniques do come with unintended consequences such as deep fakes and ethical considerations such as whether or not these personal models should be created, yet I implore you to look past those and focus on the benefits they could provide to individuals and society as a whole.

Restoration: Shifting quickly to visual rendering. AI can play an important role in both analysis of visual imagery and recreation thereof. Start off with an early example such as Ubisoft aiding the restoration of Notre Dame through its detailed modeling in Assassins Creed Unity game (the amount this aided the restoration is currently disputed)? Through detailed digitization and 3D rendering, AI models can interpret and possibly recreate models so, as restoration is required, architects, artists, and the like can recreate historical structures with significant meaning back to its original.

Biodiversity: AI monitors and protects endangered species by analyzing data from sensors and cameras. Projects like Wildbook use AI to identify and track individual animals, aiding in conservation efforts

Internet Archive: Taking digitized books aside, a non-profit called the Internet Archive exists to, as it’s name suggests, archive the internet, capturing everything from human expression across social media, long lost webpages like Ask Jeeves, news articles, and more.

I have no doubt that historians are looking to AI for analysis of the past with researching models similar to DeepSeek or Deep Research. What can we learn? What patterns can we uncover when we unleash AI models to explore the past? Forbes has an interesting piece about AI in preservation, siting a Project Time Machine to reconstruct 2000 years of European History:

The international research teams involved in the open-source Time Machine project aim to not only digitize huge amounts of information currently stored in archives and museums, but to then use AI to analyze the data to reconstruct 2000 years of European history with a “Large Scale Historical Simulator”. This project, which has just secured 1 million euros in funding from the European Union, would serve as a time capsule to facilitate exploration of the cultural, economic, and historical development of European cities, vastly improving our current understanding.

Our animalistic, humanized, intrinsic need for preservation manifests itself in fascinating ways. Emphasizing, again, that AI can be a boon to society, for social good, simply by creating models of niche parts of society and getting to know every facet of the human condition, learning from individuals with decades of life experiences, stories, and insights of about the world. Some estimates predict that there is 402M TB of data generated every day. That level of scale is perfect for AI! In effect, there might be a virtuous cycle at play whereby the societal insights we gain can help understand and preserve the past while helping making progress on the future.

We should all go back to the concept of Yes Theory and embrace discomfort, both in the human sense but in the digital one as well. As AI gets democratized, we should continue to think how AI can be a social boon and preserve our societal legacy for future generations to learn, restore the past, and progress the future!