Thank you both so much for sharing this conversation! You inspired me to spend some time with the Iris dataset these past couple of evenings. I've never learned the history of this dataset before. I tried the ideographs and some other plots, and I really liked the idea of uploading images to help get the outcome you want. "Just ask" seems to be the lesson I have to keep learning. A couple of prompts I enjoyed:
- "what are some unusual data visualisations I could use to explore the iris data set in a jupyter notebook?"
- "Imagine I'm a botanist. What are some interesting questions I could ask about the Iris dataset using a jupyter notebook?" (then I tried Imagine I'm a data scientist)
I love the questions! I am happy to hear we inspired you to test it out! The more I try to more I am completely baffled by what kind of machine humans invented. I can't wrap my head around it, really.
Interesting film! I'm impressed that Tyler knows the Gaspé Peninsula.
You went quickly over the iris dataset, so viewers may not know about the origins of the data for the third species, iris virginica. Check my article with Kim Kleinman: "The Iris Data Set: In Search of the Source of Virginica" Significance, Volume 18, Issue 6, December 2021, p 26–29
Thanks for sharing the article Antony! The story of some data sets is so fascinating! There should be a show only on that!!! :) I am glad to hear you found the video interesting!
I really don't understand why you find the performance of chatGPT impressive. The histogram was hopelessly wrong, and the ideograms took a lot of prompting to make and were still not right. I coded along with the video and had a plot of the ideograms ready when you had barely started working on them. The entire plot took 13 lines of code, and less than five minutes to write. It is incredibly simple. The AI hype makes me despair and utterly depressed. If this is a view of the future we're going to have to live in, I want no part of it.
I’m not going to try to convince you that you should be excited by this, my intention was to show a range of use cases that I find useful in my workflow using simple (untested) examples, and your mileage may vary. For those who are interested in experimenting with these tools, I don’t think the best comparison is whether you can do it faster than I can while livestreaming and attempting to maintain a conversation, but rather to compare with or without them in a given workflow.
I understand the sentiment Michiel. I think the goal here was not to show that ChatGPT can do better than a human but simply that it's capable of understanding so much. It is true that Tyler uses it to speed up his work and offload some of the tasks, but eventually, the examples we used are more focused on reproducing existing charts, which is not how one would use GenAI in practice in a data science project. Does this make more sense to you?
Thank you both so much for sharing this conversation! You inspired me to spend some time with the Iris dataset these past couple of evenings. I've never learned the history of this dataset before. I tried the ideographs and some other plots, and I really liked the idea of uploading images to help get the outcome you want. "Just ask" seems to be the lesson I have to keep learning. A couple of prompts I enjoyed:
- "what are some unusual data visualisations I could use to explore the iris data set in a jupyter notebook?"
- "Imagine I'm a botanist. What are some interesting questions I could ask about the Iris dataset using a jupyter notebook?" (then I tried Imagine I'm a data scientist)
I love the questions! I am happy to hear we inspired you to test it out! The more I try to more I am completely baffled by what kind of machine humans invented. I can't wrap my head around it, really.
Interesting film! I'm impressed that Tyler knows the Gaspé Peninsula.
You went quickly over the iris dataset, so viewers may not know about the origins of the data for the third species, iris virginica. Check my article with Kim Kleinman: "The Iris Data Set: In Search of the Source of Virginica" Significance, Volume 18, Issue 6, December 2021, p 26–29
Thanks for sharing the article Antony! The story of some data sets is so fascinating! There should be a show only on that!!! :) I am glad to hear you found the video interesting!
I really don't understand why you find the performance of chatGPT impressive. The histogram was hopelessly wrong, and the ideograms took a lot of prompting to make and were still not right. I coded along with the video and had a plot of the ideograms ready when you had barely started working on them. The entire plot took 13 lines of code, and less than five minutes to write. It is incredibly simple. The AI hype makes me despair and utterly depressed. If this is a view of the future we're going to have to live in, I want no part of it.
I’m not going to try to convince you that you should be excited by this, my intention was to show a range of use cases that I find useful in my workflow using simple (untested) examples, and your mileage may vary. For those who are interested in experimenting with these tools, I don’t think the best comparison is whether you can do it faster than I can while livestreaming and attempting to maintain a conversation, but rather to compare with or without them in a given workflow.
I understand the sentiment Michiel. I think the goal here was not to show that ChatGPT can do better than a human but simply that it's capable of understanding so much. It is true that Tyler uses it to speed up his work and offload some of the tasks, but eventually, the examples we used are more focused on reproducing existing charts, which is not how one would use GenAI in practice in a data science project. Does this make more sense to you?