Danish researchers at Danmarks Tekniske Universitet (DTU) have introduced Life2Vec, an artificial intelligence model akin to OpenAI’s transformer model. Departing from traditional language-based data processing, Life2Vec delves into sequences of key life events to predict an individual’s personality and, more remarkably, their anticipated lifespan. The model, dubbed an “AI death calculator,” claims an impressive 78% accuracy in forecasting the time of death based on an individual’s life trajectory.
Life2Vec predicting lifespan – A revolutionary leap in predictive analytics
Life2Vec, developed at DTU, represents a significant leap in predictive analytics, promising transformative applications within the healthcare industry. The tool works by crunching diverse datasets encompassing the entire population of Denmark, including medical history, education, job details, income, and marital status.
According to DTU Professor Sune Lehmann Jørgensen, Life2Vec aims to enhance our understanding of key events in human lives, with promising applications in medicine. The model’s ability to detect specific diseases opens the door to earlier and potentially life-saving interventions.
Life2Vec utilizes methodologies akin to Language Model Models (LLMs) like ChatGPT and Google LLC’s Gemini. Rather than delving into textual data, Life2Vec delves into the minutiae of life events. In its embryonic stages, the focus primarily centered on prognostications of mortality, leveraging the copious data stream emanating from insurance entities.
Employing rigorous methodology, researchers meticulously nourished the algorithm with an extensive dataset, encapsulating the life-related particulars of thousands of individuals who had procured health insurance policies spanning the temporal expanse from 2008 to 2016. The algorithm, in turn, showcased a remarkable 78% accuracy rate in its forecasts pertaining to the survival outcomes of these individuals as of the year 2020, affirming the robustness of its predictive capabilities.
Beyond mortality predictions – LLM’s mastery in vector representations and embedding spaces
The model transforms life events into vector representations within embedding spaces. These digital values assigned to each event enable Life2Vec to categorize and establish connections between events.
This forms the foundation for accurate predictions, considering the sequence and minutiae of each life event. Consider instances where specifics such as Francisco’s monthly earnings of 20,000 kroner in 2012 while working as a guard at Kronborg Castle in Helsingor or Hermione’s high school academic pursuits involving five distinct A-level subjects are encoded as embeddings, facilitating accurate prognostications.
While the initial focus has been on mortality predictions, the researchers emphasize that Life2Vec’s capabilities extend beyond this realm. The model’s potential to predict various aspects of human society represents just the beginning of a more extensive effort to harness AI for societal predictions. As AI continues to evolve, the applications of Life2Vec are poised to expand, revolutionizing our understanding and predictions of human behavior.
In conclusion, Life2Vec’s unveiling marks a pioneering step in predictive analytics, specifically in predicting human lifespan with a remarkable 78% accuracy. The implications for healthcare and societal predictions are vast, opening new avenues for early disease detection and intervention. As we stand on the cusp of this technological advancement, the question arises: How will Life2Vec reshape our approach to understanding and predicting the complexities of human life and society?
Source: https://www.cryptopolitan.com/life2vec-llm-predicting-78-accuracy/