Post 7: ChatGPT 🤯
Published:
This post is a reminder for me in the future about the day I was genuinely surprised by an AI for the first time and how this has aged over time.
Today, you can ask an artificial intelligence like ChatGPT to write in numpy/python the Needleman-Wunsch algorithm, one of the most famous strategies for aligning protein sequences. However, if you ask it to implement AlphaFold2, it backs off, acknowledges its difficulty, and sends you to read the original article. Here you can see several of the responses on this topic made by Sergey Ovchinnikov To learn more about ChatGPT: https://youtu.be/c_PqM5lPvjY and https://openai.com/blog/chatgpt/
How could it not surprise you that ChatGPT explains the absurdity of a “paper” called “Can a good philosophical contribution be made simply by asking a question?” is just the mere question.
Do you want to do something in bioinformatics, but don’t understand some concepts? Do you not know if there is a program to do that, let alone how to run it? Well, now ChatGPT can explain and teach you the code needed to run what you had in mind. And so with everything: pytorch, scikit-learn, pymol, single-cell, etc. in Spanish, English, French, Morse code, etc.
Here’s an example with metagenome assembly, annotation, and binning.
