Brief

A new project called Talkie has released a 13-billion parameter language model trained exclusively on English text from before 1931. This "vintage" model aims to explore AI's ability to predict the future and generate novel ideas beyond its training data cutoff. While the base model is open-source, the fine-tuned chat version relied on modern LLMs like Claude Sonnet and Opus for assistance, raising concerns about potential anachronistic contamination. AI

Talkie is a vintage large language model trained exclusively on English text predating 1931. Its training data consists entirely of works that are now in the public domain in the United States, adhering to the current copyright cutoff date of January 1, 1931. This approach ensures the model's foundation is built upon freely accessible historical literature. AI

A new large language model called Talkie, also known as 13B 1930 LM, has been developed with a training data cutoff of 1930. This approach, termed a 'vintage LLM,' aims to emulate historical perspectives by using only pre-1930 data, thereby sidestepping copyright issues for material from that era. The project explores the concept of 'time travel' through AI and investigates how well such models can predict or understand events post-cutoff, drawing inspiration from researchers like Owain Evans and concepts from Calcifer Computing. AI

A team of AI researchers has developed Talkie, a 13-billion-parameter language model trained exclusively on English-language texts published before the end of 1930. The creators aim to use Talkie to better understand AI behavior and capabilities by studying its responses and predictive abilities within a historical knowledge constraint. While Talkie can generate some simple code solutions, its programming capabilities are limited compared to models trained on modern data. AI

A new experiment from Talkie aims to address the issue of data contamination in large language models. Contamination, where models are trained on data that includes their own outputs or benchmark test data, can lead to inflated performance metrics. This experiment seeks to isolate and quantify the impact of such contamination, providing a clearer understanding of true LLM capabilities. AI

A new language model called Talkie has been introduced, notable for its unique training dataset. This model was exclusively trained on data sourced from the year 1930. The project aims to explore the capabilities and characteristics of an LLM developed with such a specific and historical data corpus. AI