As research into artificial intelligence (AI) continues its march forward, computers are becoming more and more human-like all the time.
Making headlines of late has been the new ChatGPT, developed by OpenAI - an artificial intelligence research and deployment company that says its mission is "to ensure that artificial general intelligence benefits all of humanity."
But now, ChatGPT is moving into the text-based world of AI, allowing users to carry on human-like conversations but with a (mostly) know-it-all computer that is ever-learning.
Move over, college professors?
Of course, we're all weather geeks here at FOX Weather, so I had to test its meteorological chops. I began a conversation by asking some basics about weather and meteorology:
Q: "Hello ChatGPT, I'm looking to learn more about the weather. What can you tell me about meteorology?"
ChatGPT responded by defining meteorology in colloquial terms -- think less Merriam-Webster and more like texting a college meteorology professor.
"Meteorologists use a variety of instruments and techniques to measure and monitor the weather, including satellites, radar, and computer models," was part of ChatGPT's response. "They also study the interactions between the atmosphere and other Earth systems, such as the oceans and land masses, in order to better understand and predict the Earth's weather and climate."
Q: "Can you tell me how wind is affected by high and low pressure centers?"
Again, I received a nice conversational reply about the nuances of wind. Here's the opening paragraph:
"Yes, wind is strongly influenced by differences in air pressure. When there is a difference in air pressure between two locations, the air will tend to flow from the area of high pressure to the area of low pressure. This creates wind."
Not quite perfect… yet…
I managed to trip up its response just a little in a question about temperatures:
Q: "How about temperatures? What weather conditions will make temperatures warmer?"
I was angling for meteorological conditions, such as the influence of warm fronts or storm systems.
Its reply: "There are several factors that can influence temperature and cause it to become warmer." But the factors given included time of year, latitude, altitude, and land use -- more overall influences on temperature but not day-to-day influences––though it also mentioned cloud cover.
But maybe my question was too vague, so like any good reporter, I hit my subject with a follow-up question.
Q: What meteorological conditions will cause an increase in temperature?
That seemed to do the trick, as now it's talking about sky cover, humidity, and winds -- and added a note about climate change's influence. Also, note that the computer remembered our earlier conversation in its second reference to land use:
But it nailed the answer when I asked about describing a cold front:
Q: "And what about making temperatures colder? For example, what does a cold front do?"
"A cold front is a boundary between two air masses, with the colder air mass replacing the warmer air mass. When a cold front moves into an area, it can bring colder temperatures, along with other changes in weather, such as precipitation and wind," ChatGPT's opening paragraph read. It gave a full three-paragraph answer that explained it well:
Glimpse into how artificial intelligence works
All right then, let's really put it to the test. Here in the Seattle area where I am based, we have a local microclimate effect known as the Puget Sound Convergence Zone -- well-known to local meteorologists but not so much to those outside the area. Does the computer know the deal? I asked this question nearly the same way twice:
Q: "I'd like to learn more about the Puget Sound Convergence Zone."
"The Puget Sound Convergence Zone is a weather pattern that occurs in the Puget Sound region of Washington state, in the United States. It is a (sic) area where two different air masses meet and converge, often leading to cloudy, rainy, and sometimes stormy weather."
And here is the second reply:
The second paragraphs in both answers are where you can glimpse how artificial intelligence would answer a question when it might not have all the answers:
"As the warm, moist air moves over the colder land, it is forced to rise, which can cause clouds and precipitation to form." And "the convergence zone typically forms in the afternoon when the land heats up and causes air to rise. The rising air creates an area of low pressure, which draws in air from surrounding sides..."
Converging air masses, in a general sense, do cause clouds and precipitation, as does the afternoon heating-triggered convergence, such as what is typically found along Florida's peninsula.
Still, in this particular case around Seattle, their convergence zone is caused by colliding winds that are split around the Olympic Mountains range, not the warm air moving over colder land or afternoon heating and rising air as ChatGPT suggested.
And Seattle's convergence zones occur throughout the year at any time of day and generally remain stationary or move north to south. But you could tell the computer gleaned from what it knew about the Puget Sound local area, and what it knew about converging weather air masses in general to facet a response.
Artificial Intelligence is already being used in meteorology
While ChatGPT continues to hone its conversational learning skills, the National Center for Atmospheric Research in Colorado has already developed a forecasting tool for severe weather that uses artificial intelligence trained from 500 past forecasts of a high-resolution model then matched to actual conditions.
The machine analyzes 40 current weather measurements and uses what it learned from experiences to give detailed probabilities of tornadoes, large hail and damaging wind.
ChatGPT is still learning too, and there are some limitations to the bot – it's not quite ready for Terminator 2 status just yet. ChatGPT is not connected to the internet, so don't ask for current specific weather conditions or forecast-related answers (if you need that information, there's always the free FOX Weather app!), but it does know general climatology.
I asked if it could tell me where else I could live that would have weather similar to Seattle. It suggested San Francisco, Portland, Oregon, Halifax, Nova Scotia and Vancouver or Victoria, B.C., as cities in coastal regions with temperate marine climates.
"These cities are located in coastal regions and have weather patterns that are similar to Seattle's, with mild, wet winters and cool, dry summers," ChatGPT wrote. "They also tend to experience more cloudy days than other parts of the country." Sign me up!
But ChatGPT can be more than just basic Weather 101 questions – it is smart enough to spout expertise on a variety of complex topics, scientific or otherwise, all while giving that human vibe – even adding humor!
My final question was to explain the terrifically complicated mathematical formula of the hydrostatic equation – a tenet of meteorology that describes how physical forces act on fluids and… well, it's complex; let me have ChatGPT explain…
Oh, and do it in the voice of a pirate…