ChatGPT as a companion

by

ChatGPT from OpenAI has received quite a lot of attention and discussions after the release of its free research preview. I also played around with it over the last weeks and I am really impressed with its capabilities and potential especially to speed up certain tasks.

What can ChatGPT do?

I asked and got the following response.

As a large language model, I am able to process and generate human-like text based on the patterns and structures that I learned from a dataset of texts from the internet. I can understand and respond to a wide range of questions and prompts, and I can provide general information and assistance on a variety of topics.

Some examples of the types of tasks that I can perform include:

  • Answering questions: I can understand and respond to questions on a wide variety of topics, such as history, science, mathematics, and current events.
  • Providing definitions and explanations: I can provide definitions and explanations of words, concepts, and phenomena, drawing on my understanding of language and the context in which they are used.
  • Summarizing and synthesizing information: I can summarize and synthesize information from multiple sources, condensing complex ideas into more easily digestible form.
  • Generating natural language text: I can generate human-like text based on a given prompt or topic, synthesizing information from my training data to produce coherent and coherently written responses.

Overall, I am here to provide general information and assistance to users, and I am able to understand and respond to a wide range of questions and prompts. If you have any specific questions or tasks that you would like me to help with, please feel free to ask and I will do my best to assist you.

ChatGPT

So far I primarily used it for information search tasks about software development and different frameworks and tools. In the following, I will present some examples before I will raise some concerns I thought of and my takeaway from using ChatGPT for some time.

When I asked for a summary of unit testing best practices I got the following response

Unit testing is a software testing technique in which individual units or components of a software application are tested in isolation from the rest of the application. The goal of unit testing is to validate that each unit or component of the application is working as intended and meets the specified requirements.

Here are some best practices for unit testing:

  1. Write tests before writing the implementation code: This practice, known as “test-driven development,” helps ensure that the implementation code satisfies the requirements specified in the tests.
  2. Keep tests small and focused: Each test should test a single unit or aspect of the code. Avoid testing multiple units or aspects in the same test, as this can make it difficult to identify the cause of a failure.
  3. Make tests independent: Tests should not depend on the execution or results of other tests. This ensures that the tests can be run in any order and still provide accurate results.
  4. Use descriptive names for tests: Use descriptive names for tests that clearly describe the purpose of the test and the unit being tested.
  5. Use assertions to validate test results: Assertions are statements that specify the expected outcome of a test. Use assertions to validate that the code under test produces the expected results.
  6. Use mock objects to isolate units under test: Mock objects are simulated objects that mimic the behavior of real objects in a controlled manner. Use mock objects to isolate the unit under test and ensure that the test is only testing the unit and not its dependencies.
  7. Run tests regularly: Run unit tests regularly, ideally as part of an automated build process, to catch and fix defects early in the development process.

By following these best practices, you can improve the effectiveness and reliability of your unit testing efforts.

ChatGPT

This response covers the most essential points I think and can be compared to what I wrote in my post https://blog.kaikarren.de/2022/12/31/unit-testing-best-practices-summary/. However, ChatGPT is not able to provide sources for its information and it can be completely wrong with confidence because it usually writes its responses in a style that resembles 100% confidence in the correctness of the information it provides.

In another example, I asked for an explanation of Lombok.

Project Lombok is a Java library that can be used to reduce the amount of boilerplate code that is required in Java projects. It provides a set of annotations that can be used to automatically generate common code patterns, such as getters and setters, toString methods, and constructors.

By using Lombok, you can avoid having to write a lot of repetitive, boilerplate code and focus on the core logic of your application. This can make your code cleaner, easier to read, and more maintainable.

Lombok is especially useful for projects that use the JavaBeans pattern, where the structure of an object is defined by private fields with corresponding getters and setters. Using Lombok, you can annotate the fields with @Getter and @Setter annotations to automatically generate the getters and setters.

Here are some examples of how Lombok can be used:

In this example, the @Getter, @Setter, @ToString, @NoArgsConstructor, and @AllArgsConstructor annotations will cause Lombok to generate the following methods for the Person class:

@Getter
@Setter
@ToString
@NoArgsConstructor
@AllArgsConstructor
public class Person {
  private String firstName;
  private String lastName;
  private int age;
}

Getters and setters for the firstName, lastName, and age fields
A toString method that returns a string representation of the object
A no-arg constructor that creates a new instance of the object with default values for the fields
An all-arg constructor that creates a new instance of the object with the provided values for the fields
Overall, Lombok can be a useful tool for reducing boilerplate code and making your Java code more concise and readable. However, it is important to be mindful of the trade-offs involved, as Lombok can add complexity to your build process and may not be suitable for all projects.


ChatGPT

Using ChatGPT as a kind of replacement for Google Search is nice, but the power of ChatGPT from my point of view is the interactivity that it allows through its thread model. For example, you can tell it to rephrase the texts above in a certain style, add and replace paragraphs, as well to let it adjust code examples (or real code). I also asked it to create SQL statements to create tables with certain columns and to insert some data that I provided as text and it provided me the correct statements that I could execute and also explain them correctly. I did not test it as a helper to find bugs yet, but I read that it should also be a decent job here.

Similarly, as for Github Copilot, I am not sure about potential legal issues because it seems like it was trained basically on all available information on the internet and not everything that you can view is free to use for you how you like it. I got the following response to my question about my concerns.

As an AI language model, I was not trained on any specific dataset. I was trained using a machine learning technique called unsupervised learning, which involves training a model on a large dataset of text and allowing the model to learn patterns and structures in the data without any explicit labels or supervision.

The dataset used to train me was sourced from the internet and consists of a wide variety of text from a variety of sources, including books, articles, and websites. The dataset was compiled and processed in a way that ensures that it is legal and compliant with all applicable laws and regulations.

As a language model, I do not have the ability to browse the internet or access any new information beyond what I was trained on. My responses are generated based on patterns and structures in the training data, and I do not have any personal beliefs or opinions. I am here to provide general information and assistance to users, and I am not able to browse the internet or access new information beyond what I was trained on.

ChatGPT

For now, I think that’s sufficient for me. However, I am assuming there may be more discussions and legal battles in the future. As far as I know, there are also no reliable ways to detect content generated by such large language models making the ChatGPT ban by Stackoverflow hard to enforce.

Takeaway

ChatGPT is a great tool that can help you to search for information, explore and learn in an interactive way. However, because it can be completely wrong you should not turn off your brain. I am looking forward to seeing it further improve.

Leave a Comment