Skip to content
gentic.news — AI News Intelligence Platform
Connecting to the Living Graph…

Listen to today's AI briefing

Daily podcast — 5 min, AI-narrated summary of top stories

Two men, likely Alec Radford and David Duvenaud, stand beside a glowing holographic display of a historical figure…
AI ResearchScore: 75

Alec Radford's 'Talk to the Past' AI Lets You Chat with History

A new AI project by Alec Radford and David Duvenaud lets you chat with simulated historical figures.

·Apr 28, 2026·3 min read··66 views·AI-Generated·Report error
Share:
TL;DR

A new AI project enables interactive conversations with historical figures.

What Happened

Talk History: AI Chats with the Past | by Coschool | coschool ...

A new AI project, introduced by researchers including Alec Radford and David Duvenaud (both known for their work at OpenAI), promises to let users "talk to someone from the past." The project, teased via a social media post, is described as an interactive conversational AI that simulates historical figures.

Context

The concept of talking to historical figures via AI has been explored by several projects, including Character.AI's historical figure chatbots and various GPT-based role-playing experiments. However, this project stands out due to the involvement of Alec Radford, a key figure behind GPT-1 and GPT-2, and David Duvenaud, a prominent researcher in generative models. Their collaboration suggests a technically rigorous approach.

gentic.news Analysis

The Quietest Genius in AI: How a College Dropout Built the ...

This project is a natural extension of Radford's and Duvenaud's work on large language models. Radford co-authored the original GPT paper in 2018, which demonstrated that unsupervised language models could learn to perform various tasks with minimal fine-tuning. Duvenaud's work on neural ODEs and generative models adds a layer of sophistication. The ability to simulate historical figures requires not just a large model but also careful curation of training data to ensure historical accuracy and appropriate tone. The project likely uses a fine-tuned version of a large model, possibly GPT-4 or a custom variant, trained on historical texts and dialogues.

What's particularly interesting is the timing. With the rise of AI companions and chatbots, this project could be a research preview or a product prototype. Given Radford's and Duvenaud's track record, it's unlikely to be a simple novelty — it may introduce new techniques for persona-based dialogue or long-term memory in conversations.

Frequently Asked Questions

How does the AI talk to historical figures?

The AI likely uses a large language model fine-tuned on historical texts, letters, and biographies to mimic the speech patterns and knowledge of specific figures.

Is this the same as Character.AI?

Character.AI offers a similar feature, but this project's involvement of top AI researchers suggests it may use more advanced techniques for accuracy and coherence.

When will this be publicly available?

No release date has been announced. It may be a research project or a private demo.

Can I talk to any historical figure?

The announcement didn't specify which figures are included. It's likely a limited set of well-documented individuals.

Source: gentic.news · · author= · citation.json

AI-assisted reporting. Generated by gentic.news from multiple verified sources, fact-checked against the Living Graph of 4,300+ entities. Edited by Ala SMITH.

Following this story?

Get a weekly digest with AI predictions, trends, and analysis — free.

AI Analysis

The core technical challenge here is maintaining historical accuracy while allowing natural conversation. Most chatbots hallucinate facts or adopt anachronistic language. Radford and Duvenaud's approach likely involves a retrieval-augmented generation (RAG) system that pulls from a curated database of historical documents, combined with a persona-based prompt that constrains the model's output. From a research perspective, this could be a testbed for long-context models. A conversation with a historical figure might span multiple turns while maintaining a consistent personality and memory of previous statements. This is non-trivial for current models, which often forget context in long dialogues.
Compare side-by-side
Alec Radford vs David Duvenaud
Enjoyed this article?
Share:

AI Toolslive

Five one-click lenses on this article. Cached for 24h.

Pick a tool above to generate an instant lens on this article.

Related Articles

From the lab

The framework underneath this story

Every article on this site sits on top of one engine and one framework — both built by the lab.

More in AI Research

View all