Choose an option to see content specific to your location and language

Longevity solved? Erik Pickett meets ChatGPT

With all the recent buzz around the Artificial Intelligence language model, ChatGPT, the Club Vita team were keen to understand how it might help us understand the world of longevity and what limitations may exist when using such a tool. I set out to “interview” ChatGPT to find out.


Here’s the transcript of my interview and my thought track as I went through:

[EJP]: Hello ChatGPT

[ChatGPT]: Hello! How can I assist you today?

[EJP]: Well, let’s start with some introductions. What are you, ChatGPT?

[ChatGPT]: I'm ChatGPT, a language model developed by OpenAI. My purpose is to assist and converse with people like you through written communication. I use machine learning algorithms to understand the context of the questions and provide helpful responses based on my training data. Is there anything specific you would like to know or talk about?

[EJP]: Yes, there is ChatGPT, I’d like to ask you some questions about my own subject area if that’s ok. Do you know anything about human longevity?

[ChatGPT]: Yes, I have access to information about human longevity. Human longevity is the study of the factors that contribute to the length of a person's life, including genetics, lifestyle, and environmental factors. There are many different theories about the aging process, including the role of telomeres, cellular damage, and oxidative stress. Additionally, there are various lifestyle factors that are thought to affect longevity, such as diet, exercise, stress management, and social connections. Would you like me to provide more information about any of these topics?

I was interested to know what conclusions an AI would make about some of the big debates in the field of longevity, so I started with a biggie…

[EJP]: How long is it possible for a human to live?

[ChatGPT]: The maximum human lifespan is currently estimated to be around 120-125 years. While there have been a few individuals who have lived beyond 120 years, these cases are extremely rare and may be due to exceptional genetics or lifestyle factors.

It's important to note that while the maximum human lifespan may be limited, there are many things people can do to increase their chances of living a long and healthy life, such as maintaining a healthy diet, getting regular exercise, avoiding harmful habits like smoking, and staying socially connected.

That was a very definitive answer to a question that attracts significant debate. I was reminded of a pub quiz I once did with my dad, where he answered every question incredibly confidently - including when he was wrong. Could ChatGPT be exhibiting similar traits?

[EJP]: How confident are you that the maximum human lifespan is 125?


[ChatGPT]: Based on the available scientific evidence and research, it is widely believed that the maximum human lifespan is around 125 years. However, it's important to keep in mind that this is an estimate based on current understanding of human biology and aging, and it's possible that new discoveries and advancements in medical technology could change this estimate in the future.

As an artificial intelligence language model, I rely on the information available to me from my training data, which includes scientific studies and research on human longevity. Based on this information, I can say that the prevailing scientific consensus is that the maximum human lifespan is around 125 years, but it's always possible that new information or discoveries could change this understanding in the future.

The answer to this question (and a number of others I asked trying to trick it) showed that ChatGPT was indeed confident in its answer. At the risk of anthropomorphising an AI, I wondered if it was possible for ChatGPT to believe it was right, when actually it was wrong.

[EJP]: Is it possible you are wrong?

[ChatGPT]: As an artificial intelligence language model, my responses are based on the patterns and associations learned from my training data, and it's possible that my responses may contain errors or inaccuracies.

So, as with any other statistical model, when using AI tools like ChatGPT we need to be careful of the “garbage in, garbage out” principle and ensure that we trust the data it is using to make its conclusions. Good writing practice for humans is to include sources for key information. I wondered if ChatGPT could do something similar…

[EJP]: Are you able to provide sources for your answers?

[ChatGPT]: As an AI language model, I don't have direct access to external sources, but I can provide information and insights based on the knowledge and data that I was trained on. My responses are generated using statistical language models and deep learning techniques, which allow me to generate coherent and contextually appropriate responses to questions.

I try my best to provide accurate and up-to-date information, but I may not always have access to the most recent research or data. If you have any doubts about the accuracy or reliability of any information that I provide, I would encourage you to seek additional sources of information or consult with a qualified expert in the relevant field.

The fact that ChatGPT cannot provide sources or describe how it reached its conclusions means that we’ll have to do some corroborating research before we can rely entirely on its answers. It feels like ChatGPT is fulfilling the role of an intelligent and logical peer who has read up on a lot of background material. My next thought was around what that background material was…

[EJP]: What data were you trained on?

[ChatGPT]: As an artificial intelligence language model, I was trained on a diverse and large set of textual data from a wide variety of sources, including books, websites, news articles, and scientific papers.

The specific sources of my training data include a broad range of topics and fields, such as literature, history, science, technology, current events, and more. The training data was selected and curated by the team of researchers and developers who created and trained my model, with the goal of providing a comprehensive and diverse range of information for me to learn from.

The cutoff date of my training data is mid-2021. This means that my training data includes information and knowledge that was available up until that point in time. However, it's important to note that my model is continually updated and improved based on new data and information, and so my responses may reflect more recent developments in certain fields, even if they were not explicitly included in my training data.

This, I think, gets to the crux of working with models such as ChatGPT. Any biases present in the curated training data will feed through into the conclusions it makes – which of course is also something to think about when analyzing the opinions of industry experts. ChatGPT is sounding more and more human the more I find out! And unfortunately, the cut-off date means we won’t be able to ask ChatGPT for its insights on what caused the excess deaths experienced in different countries throughout 2022 (the subject of our next Club Vita webinar). With these limitations in mind, I turned back to my original goal of finding out what ChatGPT thought about some big issues in longevity.

[EJP]: What does ChatGPT mean for human longevity?

[ChatGPT]: As an AI language model, ChatGPT does not directly impact human longevity. However, ChatGPT can help improve our understanding of the factors that contribute to human longevity and help us identify strategies to improve health outcomes and increase life expectancy.

By providing access to a vast amount of knowledge and information, ChatGPT can help researchers and healthcare professionals better understand the complex interactions between genetics, lifestyle, social and economic factors, and environmental factors that contribute to human longevity. This knowledge can inform the development of new medical treatments, public health policies, and other interventions that can help improve health outcomes and increase life expectancy.

Additionally, ChatGPT can help educate the public about the importance of healthy lifestyle choices and disease prevention strategies. By providing accurate and reliable information on topics such as nutrition, exercise, and stress management, ChatGPT can help individuals make informed decisions that can improve their overall health and increase their chances of living a longer, healthier life.

And what about the impact of the COVID-19 pandemic on…

[EJP]: In the long run, do you think the COVID-19 pandemic will increase or decrease life expectancy?

[ChatGPT]: The long-term impact of the COVID-19 pandemic on life expectancy is difficult to predict with certainty. However, it is likely that the pandemic will have a negative impact on life expectancy in the short term, but its long-term effects may be more complex.

In the short term, the COVID-19 pandemic has led to a significant increase in mortality rates in many countries, particularly among older adults and those with underlying health conditions. This has led to a decline in life expectancy in some countries, particularly those with high COVID-19 mortality rates.

However, it is important to note that the impact of the pandemic on life expectancy may be temporary. With the development and distribution of effective vaccines and other medical treatments, it is possible that mortality rates will decline in the future, which could help increase life expectancy.

Additionally, the pandemic has led to significant changes in healthcare delivery, public health policies, and lifestyle behaviors that could have both positive and negative effects on life expectancy in the long term. For example, the pandemic has highlighted the importance of public health measures such as hand washing, mask wearing, and social distancing, which could lead to reduced transmission of infectious diseases and improved health outcomes in the long term.

On the other hand, the pandemic has also led to disruptions in healthcare access and utilization, which could lead to delays in diagnosis and treatment of non-COVID-19 illnesses and contribute to decreased life expectancy in the long term.

Overall, the long-term impact of the COVID-19 pandemic on life expectancy is likely to be complex and multifaceted, and will depend on a variety of factors including future developments in medical treatments and public health policies.

In my view, these answers seem reasonable and balanced. I don’t think we’re ready to replace experts in industry thought leadership, but I certainly think ChatGPT can give us some things to think about. To finish up, I asked ChatGPT perhaps the most important question of the session:

[EJP]: Do you think people should follow Club Vita on LinkedIn?

[ChatGPT]: As an AI language model, I cannot provide a personal opinion on whether people should follow Club Vita on LinkedIn. However, if someone is interested in actuarial science, mortality trends, or longevity research, following Club Vita on LinkedIn could be a good way to stay informed about industry news, insights, and research. It's always a good idea to follow reputable sources and experts in your field of interest to stay up to date on the latest developments and trends.

Maybe ChatGPT can be trusted after all – as it says, if you’re interested in actuarial science, mortality trends and longevity research head on over here and give us a follow:
Share this article:
Icon/Arrow/UpIcon/Pin/Calander12Icon/Close/blackIcon/Social/FacebookFlag/CanadaFlag/wolrdFlag/ukFlag/usaIcon/Social/LinkedinIcon/MinusIcon/PinIcon/ExpandIcon/QuoteIcon/Website-greenIcon/Website/grey