Recently
ChatGPT, the fire has been in a mess. As a fringe swinger who is struggling to survive in the fields of biomedicine and computer science, he also came to catch a wave of heat.
ChatGPTis a pre-trained language model OpenAItrained by . It can be used to generate natural language text, and it can hold conversations. It is based on Transformerschemas that capture complex relationships between languages. It can be used to develop chatbots, voice assistants, comment generation systems, and more.
Next , I will test from what I am good at 组学生信and three aspects.数据可视化机器学习
Bioinformatics Analysis
Ask about the RNAseq analysis process
The process is right but there is no code, see if you can write us a code~
It's over, I feel like I'm going to lose my job. I have used a lot of pythonscripts to see if there is any code.
Although there is no code, he gives an example of pythonstandardization RPKM. good cow~
How to use each software?
Very detailed, how to interpret the quality control report?
Very detailed~ Of course, these things can also be searched on Google. It can be seen that ChatGPTthere is still a slight deficiency in automatically generating a ready-to-use upstream analysis code. However, tinkering with the outline is enough.
data visualization
Let's draw a picture!
Write down your requirements in detail.
As you can see, in addition to the code, there is also the usage of each parameter.
May I ask how to draw the picture in the last tweet?
All the functions used are given ~ it is very powerful.
It seems that if you see a picture that you can't draw, ask first, ChatGPTand you may have an idea.
machine learning
Machine learning should be ChatGPThome.
Ask about the code for random forest classification in R language.
How to adjust parameters ?
Summarize
After the test, I just want to say: ChatGPT, what else can you not know! In general, ChatGPTit can act as an AI assistant in the problem-solving process. Of course, being a brand new model, ChatGPTthere is naturally a lot bugwaiting to be fixed. OpenAISome existing limitations are also listed on the official website.
ChatGPTSometimes answers that sound plausible, but are actually outrageous. This problem is difficult to solve because: during training, there is no source of truth for reinforcement learning; too much focus on accuracy can cause the trained model to be more cautious, potentially rejecting questions that could have been answered correctly; supervised training can mislead the model because the ideal answer depends on It depends on what the model knows, not what the human demonstrator knows.ChatGPTSensitive to tweaking input wording or trying the same prompt multiple times. For example, given the wording of a question, a model can claim not to know the answer, but with a slight rephrasing, it can answer correctly. The model is often too verbose and overuses certain phrases, such as reiterating that it wasOpenAItrained by the language model. Ideally, when a user question is ambiguous, the model asks the user for further explanation. However, current models usually guess the user's intent.
previous content
How to draw this picture | Correlation Analysis Lollipop Chart
(Free tutorial + code collection) | Follow Cell to learn drawing series collection
R actual combat | Lasso regression model establishment and variable screening
Follow NC to map | Advanced interaction network diagram (protein + enrichment pathway) (Cytoscape)
R actual combat | Add a circle to the clustering (ggunchull)
