Bonus points: It was for an environmental conservation position
I had a climatology professor who was in his 50s or 60s. He was obsessed with LLMs and thought they were the greatest boon to research because he no longer had to create a spreadsheet of environmental conditions over time. Granted we're talking hourly readings of half a dozen variables for 10-20 years across multiple weather stations, but it's literally just exporting that data as a .CSV and importing it into Excel.
He tried to demonstrate this to us, spending an entire class period showing how ChatGPT will read the file and generate the same spreadsheet instantly. It kept fucking up and hallucinating huge chunks of it. Sometimes entire years would be missing, sometimes the data and graphs were nonsensical. There was no way I could trust any of it without manually verifying every line, which is over 100k lines and all of the different categories.
That week our assignment was to ask ChatGPT to code an import tool and have it process a different, much simpler dataset. The answers it should have returned were listed below the input boxes. I couldn't get over how lazy that class was for what should have been one of my most intense interdisciplinary ones. Using it in any other class was considered cheating.
Oh, and we were specifically looking at climate change's impact on this region making it significantly hotter over that time period. The whole point was to demonstrate climate change exists. He had 50+ students use ChatGPT and make it parse 100k+ lines in 50+ spreadsheets to prove that climate change is happening, without ever mentioning the inputs going into LLMs and data centres.



Yeah, I feel the exact same way. I want there to be less bullshit too much to consider adding to it.
The planet's dyin', Cloud.
were right all along and the only genuine career path in this space is 