Written by 8:34 pm AI Prediction

### Can Artificial Intelligence Forecast Your Demise?

A new study used machine learning to “autocomplete” the life trajectories of six million Danish peo…

Lauren Leffer: Forecasting the future holds immense power. Throughout history, humans have sought to predict what lies ahead, achieving notable success in fields like meteorology and climate modeling.

Tulika Bose: However, from a scientific perspective, the realm of crystal balls and oracles remains distant. Certain aspects of the future elude reliable prediction, despite the allure of horoscopes.

Leffer: Imagine if artificial intelligence could bridge the gap, delving into the supernatural realm of divination. Could an AI model accurately project an individual’s demise?

Bose: That concept sounds alarming, doesn’t it?

Leffer: Indeed, the implications are unsettling, albeit not necessarily for the reasons one might anticipate. I’m Lauren Leffer, a tech reporting fellow at Scientific American.

Bose: And I’m Tulika Bose, Senior multimedia editor.

Leffer: Welcome to the Science, Quickly podcast by Scientific American.


Leffer: Tulika, the inevitability of death is a universal truth. Yet, if given the chance, would you want to know the precise moment of your passing?

Bose: Oh, no, absolutely not.

Leffer: You didn’t even let me finish my question.

Bose: Regardless, the answer remains the same.

Leffer: Agreed. Some things are better left unknown.

Bose: Indeed.

Leffer: Despite the disconcerting nature of an “AI death predictor,” researchers have developed a tool named life2vec, as detailed in a recent study.

Bose: So, it’s essentially an AI widget that predicts one’s lifespan?

Leffer: To an extent, yes. However, life2vec transcends mere mortality forecasts. It offers a broader spectrum of predictions. It’s essential to note a significant limitation—its accuracy is uncertain, particularly since it’s trained on data specific to Danish residents.

Bose: That’s somewhat reassuring. But could you elaborate on its other predictive capabilities?

Leffer: Certainly. Apart from mortality projections over a four-year span, the tool was tested for predicting individuals’ responses to personality assessments and the likelihood of relocating internationally. Impressively, life2vec demonstrated higher accuracy compared to alternative prediction methods, achieving around 78% accuracy for mortality and 73% for international relocation.

Bose: Fascinating.

Leffer: However, interpreting the concept of accuracy in this context poses challenges.

Bose: Wait, how can a death calculator also predict personality traits and relocation patterns? What’s the connection?

Leffer: Let’s delve deeper. Life2vec operates on a machine learning model, analyzing extensive datasets to identify patterns. By scrutinizing diverse information, such as employment history, demographics, and medical interactions, the model discerns correlations between various life events. For instance, it explores the relationship between location and income, the survival rates of individuals with specific illnesses, or the influence of professions on longevity.

Bose: So, career choices could influence personality traits?

Leffer: Precisely. This transition from basic data to complex predictions showcases the model’s predictive capabilities. Notably, life2vec functions akin to an advanced autocomplete system commonly used in chatbots and language processing.

Bose: It’s intriguing how these death predictions are essentially completing someone’s life trajectory.

Leffer: That’s one way to perceive it. By inputting a life event timeline and posing specific queries, the model anticipates the subsequent steps based on the provided information, sometimes including mortality predictions.

Bose: It almost resembles a scenario from a Black Mirror episode, with phones predicting one’s demise.

Leffer: An intriguing analogy indeed. Perhaps a storyline worth exploring further.

Bose: Earlier, you seemed hesitant about the tool’s accuracy. Could you elaborate on that?

Leffer: Certainly. Upon consulting statistical and life modeling experts, doubts were raised regarding the reported accuracy rates exceeding 70%. One expert, Christina Silcox, expressed skepticism, emphasizing the tool’s limitations by stating, “I would not quit my job and go to the Bahamas based on this thing.”

Bose: A humorous yet insightful perspective.

Leffer: Indeed, predicting something as complex as death poses significant challenges, especially when assessing the accuracy of a novel tool like life2vec.

Bose: So, determining the benchmark for accurate death prediction is crucial for evaluating life2vec’s efficacy.

Leffer: Absolutely. Considering the rarity of death events and the inherent complexity of mortality forecasting, establishing a comparative standard is essential.

Bose: That sheds light on the intricacies of mortality prediction and the evaluation of such advanced tools.

Leffer: Indeed, it’s a realm filled with uncertainties and complexities that necessitate careful scrutiny.

Bose: Moving forward, what practical applications could this tool have beyond mere mortality predictions?

Leffer: With further refinement and testing, life2vec holds promise in enhancing disease prognosis and understanding health outcomes. Moreover, from a sociological perspective, it could uncover hidden societal biases, revealing unexpected correlations between factors like age, nationality, and career progression.

Bose: The potential implications are vast, although concerns regarding misuse linger. Could this technology pave the way for pre-emptive actions akin to scenarios depicted in Minority Report?

Leffer: Valid apprehensions indeed.

Bose: It’s reassuring to hear about the regulatory measures in place to govern the tool’s usage, emphasizing data protection and ethical considerations.

Leffer: Certainly. The transparency and accountability maintained by the researchers, coupled with stringent privacy laws in Denmark, offer a level of assurance. However, the ethical dilemmas surrounding predictive technologies persist, especially in regions lacking robust data privacy regulations.

Bose: As a U.S. resident, such discussions evoke mixed sentiments, considering the ethical implications of predictive tools in various domains.

Leffer: The ethical landscape surrounding predictive technologies is indeed complex, with existing applications like predictive policing and sentencing algorithms raising pertinent concerns about bias and accuracy.

Bose: It’s a stark reminder of the delicate balance between technological advancements and ethical considerations, particularly in realms impacting societal structures.

Leffer: True. The transparency and intent behind tools like life2vec aim to foster dialogue and critical reflection, steering conversations towards a more ethical and informed future.

Bose: Striving for a utopian ideal amidst the complexities of predictive technologies is a noble pursuit, albeit challenging in practice.

Leffer: Agreed. Progressing towards a more just and equitable future demands vigilance and active engagement with evolving technologies.

Bose: Aiming for a balanced and ethical technological landscape seems like a worthwhile endeavor, steering us away from potential dystopian outcomes.

Leffer: Indeed, navigating the complexities of technological advancements requires a nuanced approach, fostering understanding and accountability.

Bose: Striking that balance between innovation and ethics is key to shaping a future that aligns with societal values and aspirations.

Leffer: Precisely. Embracing transparency and ethical frameworks can pave the way for a more conscientious integration of predictive technologies into our societal fabric.

Bose: Reflecting on these discussions, it’s evident that the trajectory of predictive technologies hinges on our collective responsibility and ethical stewardship.

Leffer: Absolutely. By engaging in informed dialogues and upholding ethical standards, we can steer technological advancements towards a more equitable and sustainable future.


Leffer: This episode of Science, Quickly was crafted by Jeff DelViscio, Tulika Bose, Kelso Harper, and Carin Leong, with editing by Elah Feder and Alexa Lim. Our theme music was composed by Dominic Smith.

Bose: Remember to subscribe to Science, Quickly on your preferred podcast platform. For comprehensive science news and features, visit ScientificAmerican.com. And if you enjoy the podcast, please consider leaving a rating or review!

Leffer: For more captivating insights from Scientific American’s Science, Quickly, I’m Lauren Leffer.

Bose: And I’m Tulika Bose. Until next time!

Visited 1 times, 1 visit(s) today
Last modified: January 23, 2024
Close Search Window