ChatGPT protects your health
Graphs of pulse dynamics and cardio training
My assessment of the result: ChatGPT generally got it right that over the last two years my training has become more consistent and I have made progress. This can be seen by the decrease in average heart rate – the heart has become more efficient. At the same time, decreasing your cardio also indicates improvement: the fitter you become, the more effort you need to make progress. However, it seems to me that the conclusions could have been made more explicit – perhaps I did not formulate the request quite correctly.
Exercise helps you maintain weight
How to check: Let’s conduct a correlation analysis between the number of steps taken, activity (walking, running) and weight changes over the years. We will ask you to list the correlations of the parameters and draw conclusions whether training and walking help maintain weight based on my data.
Result from ChatGPT
Correlation matrix
With a minus sign there are negative correlations, without – direct correlations.
My assessment of the result: the more steps, the more time you spend walking – everything is logical. The correlation is 1.0. There is a negative correlation between steps and weight, it is 0.39 (weak relationship). The situation is similar with cardio training and average heart rate. These weak connections show that there is a relationship, but it cannot be explained by only two parameters – there are probably other factors, for example, calorie expenditure per day.
To lose 5 kg, I need to double my activity
How to check: let’s ask ChatGPT, based on the dynamics of my weight and activity, to calculate how much I need to move in order to lose 5 kg through activity alone and how long it will take.
Result from ChatGPT
My assessment of the result: This is a more difficult task, and in my opinion, ChatGPT failed. He suggested that in three months I could lose 5 kg if I took 2,000 more steps every day. Sounds doubtful. To get an accurate result, the neural network needs more data and refinements.
In how many years will I most likely die if I keep my activity level the same?
How to check: ask ChatGPT to analyze my current activity, as well as research and statistics, and based on these data, suggest the exact number of years in which I will die, without taking into account the factors of random death.
Result from ChatGPT
My assessment of the result: The result is close to zero, since ChatGPT apparently has built-in limitations that prevent it from predicting the date of death.
If I were a Pokemon, it would definitely be a powerful and fire type.
Spoiler: my favorite pokemon is Charizard.
How to check: Let’s ask ChatGPT which Pokemon, based on the data, I think I am, clarifying that most likely I’m powerful and fiery.
Result from ChatGPT
My assessment of the first result: Wow, he guessed Charizard! But I realized that I myself had prompted him to the answer when I indicated “fiery and powerful” in the hypothesis. I decided to ask again: “If I were a Pokemon, what kind?”
Second result from ChatGPT
My assessment of the second result: Overall, I'm pleased with the result, although I don't know such a Pokemon. It looks like this:
Forming a request for ChatGPT
All hypotheses given in the article can be tested using the promt, which is located below. It is important to test hypotheses one at a time, otherwise the neural network will freeze. Select one hypothesis to test, and then test the others as tokens are updated.
Request for ChatGPT
Hello ChatGPT. Attached is a file with data from Google Fit. Analyze it, please. Be an experienced data analyst and try to analyze these queries well.
I want to test the following hypotheses:
1. I am quite active and conform to the norms.
Analysis: look at the average number of steps per period, the average number of hours of sleep per period and compare with the norms in Russia (7-10 thousand steps). Count how many times the goal was achieved successfully; unsuccessful; above the norm and draw a final conclusion based on my activity for this period.
2. I am progressing in training and becoming more and more fit.
Analysis: study the dynamics of changes in average and maximum heart rate, as well as minutes of cardio training relative to time (quarterly), evaluate how much more fit I am becoming. Display the average and maximum heart rate over time, broken down by quarter, and draw a conclusion based on the data whether the body is more fit or not.
3. Exercising helps maintain weight.
Analysis: Correlation analysis between steps taken, activity (walking, running) and weight changes over the years. List the correlations of the parameters, draw conclusions about what helps support the weight better, draw a conclusion about the truth or falsity of the hypothesis.
4. To lose 5 kg, I need to double my activity
Analysis: based on the dynamics of my weight and activity, calculate how much I need to move in order to lose 5 kg through activity alone. Calculate in detail how many steps I need to walk in general, how many extra steps I need to walk per day on average, and how long it might take to lose 5 kg.
5. How long will it take me to die if I keep my activity level the same?
Analysis: analyze my current activity and other data available to you, as well as research and statistics, based on these data, predict the exact number of years in which I will die, without taking into account the factors of accidental death.
6. If I were a Pokemon, what kind would I be?
Analysis: Determine what Pokemon I would be based on the data and carefully justify your answer.
Conclusions
ChatGPT + Google Fit is great for testing simple hypotheses about your health. Complex hypotheses may require more detailed elaboration of data and queries, but you can always have fun and find out what kind of Pokemon you are.
PS I can go deeper into the archive from Google Fit in a new article: detailed data from the bracelet is stored there: activity, sleep, etc. It may be possible to test more interesting and complex hypotheses. Write in the comments if you are interested.
