In the recent few months, AI has become quite powerful. It has been quite powerful for quite a while, but now normal people get to use it, and with many people using it, we are seeing many examples of how much people can do with AI and how much AI is capable of, and it has only just really started, so we have a lot to see.
Chat gpt is a new chatbot from open AI. This is not a normal chatbot; rather, think of it as a smart assistant that will answer your questions and provide assistance on many things, from science to technology to programming.
It can also help you with problems, like when you are having some trouble while programming; this can help you understand the problem and even suggest some possible solutions; or when you are having some difficulty understanding some scientific concepts; it can also help you understand difficult concepts.
How to use Chat GPT to learn and understand science (Even for kid)
Let’s ask the question: “Explain evolution as if I am a kid.”
Here is what Chat GPT answered:
How to solve the accuracy issue
But Chat GPT has issues with accuracy when it comes to recent events; here is how to solve that problem.
We know that AI doesn’t know about recent events, and we can solve that by giving that information to Chat GT. It will not learn the information you are giving it, but when you tell that information to Chat GPT, then it will remember that information for your Convo thread, so when you ask more questions regarding that, it will likely give you more accurate information.
Like, let’s give a recent scientific paper and ask it to explain that paper for a beginner or non-science person.
I asked to give an explanation for a science paper about Alcubierre’s drive, and I also told the Chat GPT that I am not a science person.
Exact command: Explain this science paper. I am a non-science person, and I can’t understand what this paper says. (Copy and paste the content of the paper here.)
Here is what it give me:
As you can understand from this, AI tools can help in educating science for all types of people, no matter their current levels of science knowledge.
You can also ask about problems with this paper or this method, or you can have a full-blown conversation with Chat GPT, and in the end you will understand more about the paper than you ever have.
This is not the only way to fix the accuracy issue, but rather one of the solutions. Let me give you more examples.
We know that Chat GPT doesn’t have much most current information about stable diffusion, so if we ask questions related to the most current stable diffusion, then it will give us some older information, so let’s find a way to get around that.
We can get around it by giving it the information that is missing, or by giving an example.
Like, let’s ask the question: “What is stable diffusion 2.1?”
It gave me an answer that is far from the truth, so I replied, “No, you are wrong; this is what this is” (information about stable diffusion 2.1 from the Stability AI stable diffusion 2.1 release page).
Now it gave me more accurate information.
So now let’s ask it to give us a prompt for an image about evolution.
At first, it will give you some normal description of the image as a prompt, but you and I know this will not give us a good image, so let’s give it an example of a true stable diffusion prompt.
If it still gives a description of an image rather than a prompt, then reply to it as “format this prompt like this” and then paste an example prompt so chat GPT can understand what type of formatting it needs to do to create a working prompt.
The key is to talk to it like a human, meaning that you have to explain, give it an example, and make it clear for it so it can understand what you are trying to say.
Also mention what Chat GPT is doing wrong and what it is doing right; pinpoint the direction you want.
What if it refuses to answer
If it refuses to answer, then you must be asking some question that it’s preprogrammed not to answer, so change the intended meaning to avoid a no response from the chat bot.
I cannot explain more because it may be used in an inappropriate manner, but understand that if you are asking a question that looks malicious, then, as I mentioned, talking to it like a human may work.
But if you ask an appropriate question but it still refuses, then try to give it context or more information, or try to make your intent clear.