{ "id": 393584, "title": "What does ChatGPT think of the solution to ISLR ", "url": "/competitions/asl-signs/discussion/393584", "author_name": "", "post_date": "2023-03-09T23:33:14.428707Z", "votes": 1, "comment_count": 3, "views": 0, "content": "

Hello Kagglers 👋

\n

Since it's the trend of the year. I used ChatGPT to see what methods it suggests for ISLR. Here's the answer:\"\"

\n

What do you think? I personally I'm a big advocate of DL methods but I always try with other classical ML methods as it can give some surprises.

", "messages": [ { "id": "2175536", "postDate": "03/09/2023 23:33:14", "content": "

Hello Kagglers 👋

\n

Since it's the trend of the year. I used ChatGPT to see what methods it suggests for ISLR. Here's the answer:\"\"

\n

What do you think? I personally I'm a big advocate of DL methods but I always try with other classical ML methods as it can give some surprises.

", "rawMarkdown": "Hello Kagglers 👋\n\nSince it's the trend of the year. I used ChatGPT to see what methods it suggests for ISLR. Here's the answer:![](https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F3775861%2F019db4ae9bfff7f091768eaba4a9775a%2FCapture.PNG?generation=1678404554706335&alt=media)\n\nWhat do you think? I personally I'm a big advocate of DL methods but I always try with other classical ML methods as it can give some surprises.", "votes": null }, { "id": "2178077", "postDate": "03/12/2023 04:43:04", "content": "

👍 nice.
\nYou can check my notebook, if you like please upvote
\n https://www.kaggle.com/code/sagniksanyal/google-islr-visualization-pytorch-tflite

", "rawMarkdown": "👍 nice. \nYou can check my notebook, if you like please upvote\n https://www.kaggle.com/code/sagniksanyal/google-islr-visualization-pytorch-tflite", "votes": null }, { "id": "2179140", "postDate": "03/12/2023 23:06:11", "content": "

I personally prefer HMMs as they tend to get good results on time-series data with language structure and require less data to converge than deep learning models. For this competition, which is on isolated sign (i.e., no co-articulation or language grammar to leverage) and has many examples per sign, I do not know which methods will work the best. Note that ChatGPT only mentioned the issue of sequential data when discussing RNNs - it did not provide any help is how to handle sequential data with the other methods.

", "rawMarkdown": "I personally prefer HMMs as they tend to get good results on time-series data with language structure and require less data to converge than deep learning models. For this competition, which is on isolated sign (i.e., no co-articulation or language grammar to leverage) and has many examples per sign, I do not know which methods will work the best. Note that ChatGPT only mentioned the issue of sequential data when discussing RNNs - it did not provide any help is how to handle sequential data with the other methods.", "votes": null }, { "id": "2181588", "postDate": "03/14/2023 16:11:16", "content": "

According to OpenAI, ChatGPT is still trained on data from 2021, so it doesn't even know about recent advancements in AI/ML.

", "rawMarkdown": "According to OpenAI, ChatGPT is still trained on data from 2021, so it doesn't even know about recent advancements in AI/ML.", "votes": null } ], "comments": [ { "id": 2178077, "author_name": "sagniksanyal", "author_url": "", "post_date": "03/12/2023 04:43:04", "content": "

👍 nice.
\nYou can check my notebook, if you like please upvote
\n https://www.kaggle.com/code/sagniksanyal/google-islr-visualization-pytorch-tflite

", "votes": null, "replies": [] }, { "id": 2179140, "author_name": "thadstarner", "author_url": "", "post_date": "03/12/2023 23:06:11", "content": "

I personally prefer HMMs as they tend to get good results on time-series data with language structure and require less data to converge than deep learning models. For this competition, which is on isolated sign (i.e., no co-articulation or language grammar to leverage) and has many examples per sign, I do not know which methods will work the best. Note that ChatGPT only mentioned the issue of sequential data when discussing RNNs - it did not provide any help is how to handle sequential data with the other methods.

", "votes": null, "replies": [] }, { "id": 2181588, "author_name": "timsereno", "author_url": "", "post_date": "03/14/2023 16:11:16", "content": "

According to OpenAI, ChatGPT is still trained on data from 2021, so it doesn't even know about recent advancements in AI/ML.

", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "2175536": "Hello Kagglers 👋\n\nSince it's the trend of the year. I used ChatGPT to see what methods it suggests for ISLR. Here's the answer:![](https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F3775861%2F019db4ae9bfff7f091768eaba4a9775a%2FCapture.PNG?generation=1678404554706335&alt=media)\n\nWhat do you think? I personally I'm a big advocate of DL methods but I always try with other classical ML methods as it can give some surprises.", "2178077": "👍 nice. \nYou can check my notebook, if you like please upvote\n https://www.kaggle.com/code/sagniksanyal/google-islr-visualization-pytorch-tflite", "2179140": "I personally prefer HMMs as they tend to get good results on time-series data with language structure and require less data to converge than deep learning models. For this competition, which is on isolated sign (i.e., no co-articulation or language grammar to leverage) and has many examples per sign, I do not know which methods will work the best. Note that ChatGPT only mentioned the issue of sequential data when discussing RNNs - it did not provide any help is how to handle sequential data with the other methods.", "2181588": "According to OpenAI, ChatGPT is still trained on data from 2021, so it doesn't even know about recent advancements in AI/ML." }, "source": "meta" }