It’s even more surprising that it’s good enough at it that programmers are already incorporating it into their workflow.
This has the potential to make coding far more accessible as people can now generate some using plain English.
Because of this in the simplest of terms, all you need to do in order to code with ChatGPT is ask it to code for you. That said, it can be prone to errors and there is a set of techniques and approaches you can use to try and minimise these. We will be covering some of these here.
There are also limits to what ChatGPT can do. If you try to get it to write a fully functional program itself, it won’t be able to, no matter how detailed your prompts. This is why think you should view it as an assistant, rather than the coder in itself.
It should also be noted that while a complete beginner to coding will be able to get usable code from ChatGPT, you will still need some knowledge in order to actually implement it into any program. With all that in mind let’s get into our tips.
Here are a few specific things you can do to get the best code out of ChatGPT. These are a set of starting points and things to keep in mind to get you going.
If you spend enough time getting ChatGPT to code for you, you will eventually fine-tune your own system and exact ways of wording prompts in order to get the kind of code you are seeking.
ChatGPT will respond with complete confidence no matter what prompt you give it. This means that it won’t tell you when you haven’t given it enough information to know what kind of code to generate.
For example, you need to make sure you state clearly the programming language you want it to respond in.
You can actually think of this in a similar way to if you were hiring a coder. You need to make sure you have provided enough information so that they can complete the desired task. The more detail, the less likely ChatGPT is to create error-ridden code.
One possible way to provide this information is by giving an example of the kind of process you want it to make.
Prompt engineering is an emerging field in the world of AI. It’s based on a set of techniques and approaches used to refine and alter prompts so as to maximise the efficiency of a chatbot. One aspect of prompt engineering is the use of ‘magic phrases’.
These are a few phrases that can be added to any prompt and will alter the process the Chatbot is going through, making it more it more visible to you in the process.
For example, you can ask ChatGPT to ask any questions that it thinks would improve the program if answered. This will not only ensure you’ve given it enough information but also can sometimes give you useful suggestions that you might not have thought about.
Another useful one is to ask ChatGPT to ‘‘Explain your process step by step’ at the end of a prompt. This makes more visible to logic the program is going through more, making it easier to identify steps where it has made a mistake.
Surprisingly it also causes fewer mistakes to be made, as the chatbot is forced to justify its own actions to itself, it can actually spot mistakes it’s making and correct them.
ChatGPT may be prone to making coding errors but it is also very good at correcting them when they are pointed out. Sometimes that can be as simple as telling it the error message that was received and asking if it can think why that happened.
At other times, it still won’t be able to correct the code from here, but if you can proofread it and see the error, you can save some time by pointing this out in a specific detailed prompt and letting ChatGPT do the hard work of coding the fix.
ChatGPT cannot maintain or edit existing code, at least not yet. As a result, if you want to keep your code alive you will need to understand how it functions so you can make changes or improvements as you need to.
As a result, it pays to understand the exact code ChatGPT is creating, and a good way to do this is to view the code its generating as a starting point for your own code.
Viewing ChatGPT as a tool within your coding toolkit, rather than a coder itself is the best way to make use of this new and pwoerful tool.
So for example you could ask ChatGPT to break down what you would need to code in order to create the program you want to make, then get it to code smaller aspects of it, while you collate it together, proofreading and testing its code along the way.
