GPT-3 can do science!

Given the following data, what can you infer about Orange County? Generate a testable hypothesis, experiment, and follow-up questions:

Median home prices:
1990: $192,000
1995: $182,000
2000: $154,000

1990: 320,000
1995: 306,000
2000: 284,000

INFERENCE: The median home prices have decreased from 1990 to 2000.

HYPOTHESIS: Median home prices have decreased in Orange County from 1990 to 2000.

EXPERIMENT: Compare the median home prices in 1990, 1995, and 2000 in Orange County; if they are different, this is evidence that our hypothesis is correct.

FOLLOW-UP QUESTIONS: Why did median home prices decrease? What are the implications for homeowners, such as how much mortgage they can afford to carry? How does this impact the economy in Orange County and neighboring counties?


Oh Yeah! GPT-3 helped me with my Econ and Finance HW last semester!

1 Like

Observation is the first step in the scientific process.


I am also somehow planning to provide gene expression data to give some insights from it.

1 Like

No data to work with = no science.
As GPT-3 can’t access data by itself outside of uploaded documents and its own training data I find it to be highly unlikely for it to do actual science.
Still pretty impressive for a language transformer to observe, create a hypothesis and draw conclusions based on data.

1 Like

It seems like the more interesting question is why it decreased. My hypothesis would be, there is a decline in demand due to a decrease in population. (people are leaving town)

Also, is there enough data to establish there is an actual decline in home prices, instead of a fluctuating housing market?

I know these things are beside the point… its still pretty cool… I would like to experiment with this concept.