{ "id": 431384, "title": "Species Description", "url": "/competitions/cafa-5-protein-function-prediction/discussion/431384", "author_name": "", "post_date": "2023-08-13T10:09:21.928395900Z", "votes": 2, "comment_count": 2, "views": 0, "content": "

I looked at the species distribution in taxonomyID and wanted to publish a notebook but saw something similar in https://www.kaggle.com/code/alexandervc/cafa5-towards-eda. I wanted to add some context, used GPT4 for descriptions, rewrote and fact checked them quickly on Wikipedia. I did this for all species with 1000+ entries. 85.1% of all sequences are attributed to these 18 species. Hope it helps!
\nEDIT: apparently there are many errors in there, I decided against deleting it and let it stay here as a word of caution against LLM use if you are not familiar enough with the topic to spot errors well.

\n

\"Species

", "messages": [ { "id": "2388352", "postDate": "08/13/2023 10:09:21", "content": "

I looked at the species distribution in taxonomyID and wanted to publish a notebook but saw something similar in https://www.kaggle.com/code/alexandervc/cafa5-towards-eda. I wanted to add some context, used GPT4 for descriptions, rewrote and fact checked them quickly on Wikipedia. I did this for all species with 1000+ entries. 85.1% of all sequences are attributed to these 18 species. Hope it helps!
\nEDIT: apparently there are many errors in there, I decided against deleting it and let it stay here as a word of caution against LLM use if you are not familiar enough with the topic to spot errors well.

\n

\"Species

", "rawMarkdown": "I looked at the species distribution in taxonomyID and wanted to publish a notebook but saw something similar in https://www.kaggle.com/code/alexandervc/cafa5-towards-eda. I wanted to add some context, used GPT4 for descriptions, rewrote and fact checked them quickly on Wikipedia. I did this for all species with 1000+ entries. 85.1% of all sequences are attributed to these 18 species. Hope it helps!\nEDIT: apparently there are many errors in there, I decided against deleting it and let it stay here as a word of caution against LLM use if you are not familiar enough with the topic to spot errors well.\n\n![Species Descriptions](https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F3747152%2Fd42b2f0ca709f7a55cde515b90b53fbb%2FScreenshot%20from%202023-08-13%2012-08-01.png?generation=1691921299784462&alt=media)", "votes": null }, { "id": "2388409", "postDate": "08/13/2023 11:28:21", "content": "

Note that there are quite a few errors there: oryza is a plant (rice) not a bacterium, plasmodium and trypanosoma are not fungi, and there are more.

\n

Your effort is appreciated. But this is probably a good time to caution against the use of chatgpt and its like for fact finding. Probably a good opportunity to also plug this paper on LLMs
\nhttps://dl.acm.org/doi/10.1145/3442188.3445922

", "rawMarkdown": "Note that there are quite a few errors there: oryza is a plant (rice) not a bacterium, plasmodium and trypanosoma are not fungi, and there are more. \n\n\nYour effort is appreciated. But this is probably a good time to caution against the use of chatgpt and its like for fact finding. Probably a good opportunity to also plug this paper on LLMs \nhttps://dl.acm.org/doi/10.1145/3442188.3445922", "votes": null }, { "id": "2388432", "postDate": "08/13/2023 11:41:43", "content": "

Thanks for the feedback, a bit embarassing having overlooked such huge errors, but I'll just edit instead of deleting it, as a word of caution.

", "rawMarkdown": "Thanks for the feedback, a bit embarassing having overlooked such huge errors, but I'll just edit instead of deleting it, as a word of caution.", "votes": null } ], "comments": [ { "id": 2388409, "author_name": "iddux1", "author_url": "", "post_date": "08/13/2023 11:28:21", "content": "

Note that there are quite a few errors there: oryza is a plant (rice) not a bacterium, plasmodium and trypanosoma are not fungi, and there are more.

\n

Your effort is appreciated. But this is probably a good time to caution against the use of chatgpt and its like for fact finding. Probably a good opportunity to also plug this paper on LLMs
\nhttps://dl.acm.org/doi/10.1145/3442188.3445922

", "votes": null, "replies": [ { "id": 2388432, "author_name": "raki21", "author_url": "", "post_date": "08/13/2023 11:41:43", "content": "

Thanks for the feedback, a bit embarassing having overlooked such huge errors, but I'll just edit instead of deleting it, as a word of caution.

", "votes": null, "replies": [] } ] } ], "raw_markdown_by_id": { "2388352": "I looked at the species distribution in taxonomyID and wanted to publish a notebook but saw something similar in https://www.kaggle.com/code/alexandervc/cafa5-towards-eda. I wanted to add some context, used GPT4 for descriptions, rewrote and fact checked them quickly on Wikipedia. I did this for all species with 1000+ entries. 85.1% of all sequences are attributed to these 18 species. Hope it helps!\nEDIT: apparently there are many errors in there, I decided against deleting it and let it stay here as a word of caution against LLM use if you are not familiar enough with the topic to spot errors well.\n\n![Species Descriptions](https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F3747152%2Fd42b2f0ca709f7a55cde515b90b53fbb%2FScreenshot%20from%202023-08-13%2012-08-01.png?generation=1691921299784462&alt=media)", "2388409": "Note that there are quite a few errors there: oryza is a plant (rice) not a bacterium, plasmodium and trypanosoma are not fungi, and there are more. \n\n\nYour effort is appreciated. But this is probably a good time to caution against the use of chatgpt and its like for fact finding. Probably a good opportunity to also plug this paper on LLMs \nhttps://dl.acm.org/doi/10.1145/3442188.3445922", "2388432": "Thanks for the feedback, a bit embarassing having overlooked such huge errors, but I'll just edit instead of deleting it, as a word of caution." }, "source": "meta" }