Ever since the launch of ChatGPT, our team at CoinCodeCap has fallen in love with it and has been using this ChatGPT chatbot for all forms of use cases like blogging, coding, marketing messages, etc that have greatly enhanced our productivity.ย
Shortcoming of ChatGPT
But there are limitations to ChatGPT: It isn’t always up to date with the general internet.
- The GPT 3.5 version was last updated in – January 2022
- The GPT 4o (“o” for “omni”) was last updated on May 13, 2024. The 4o version is only available after subscription, although this version can surf the internet for recent results.
So, if someone were to go on GPT and ask about the trends going on in the 2024 general elections in India, they wouldn’t know a thing.
Say you want to buy a new phone and you ask GPT for the best one-
Well, you better notโฆ
There’s another shortcoming to ChatGPT:-
Say you had an exam tomorrow and you had 2 large PDFs to go through and you don’t have enough time to go through them, youโd typically put them on chatGpt and ask for a summary, but its input limit is only around 16Kbs and if the pdf is more than that then you are on your own. Even if it was smaller than 16kbs it would summarise it in its own way that may or may not be helpful for you.ย
You could just put that topic into chatGPT and ask for an overview of that but then again thatโll be based on the general internet and might not be very relevant to what you needed or what was written in the PDF.
Therefore, except for the lack of being up-to-date with the world, it seemed to be fairly general when asked or talked about specific things. You need to use multiple prompts and varied questions to get the precise information you need.
And that’s where ChatGPT chatBOTs come in!
Also Read: 20+ ChatGPT Prompts to learn about Bots | Make NOW!
Custom Chatbots
So, Open AI rolled out a new exciting option called GPTs. With GPTs, you could create your own custom ChatGPT, Custom GPTs can be tailored to specific tasks or domains, allowing them to excel in certain areas where a general-purpose model such as the ChatGPT itself might not perform as well. You could train it with your own data and with your own set of instructions. These are basically mini ChatGPTs for a very specific purpose.
In what way are they better:-
- Domain-Specific Knowledge: Custom GPTs can be trained on specialized datasets that we can provide in the form of PDFs, documents, imagery, etc. These can be related to specific fields like medicine, law, finance, etc. This allows them to have a deeper understanding of domain-specific terminology, concepts, and nuances, ย ย we can also link some websites or articles that weโd like the bot to take information from and then get back to us.
- Task-Specific Generation: Custom models can be trained for specific tasks such as summarization, translation, code generation, etc., resulting in more accurate and relevant outputs.
- Specialized Applications: They can be customized for specific applications such as virtual assistants, chatbots, content generation for marketing, legal analysis, etc., delivering tailored solutions for these use cases.
- Privacy and Security: In scenarios where data privacy and security are critical, custom models can be trained on sensitive data without sharing it with third parties or cloud services, ensuring data confidentiality.
Also Read: 100+ ChatGPT Prompts for learning About AI | Check NOW!
How to build your own custom Chatbot
Making a custom GPT requires a premium ChatGPT subscription from openAI.
- To create your very own custom GPT, head to the ChatGPT website. Over on the left-hand side in the sidebar, you’ll notice an option for “Explore GPTs,”ย
(and yes, you will see the word GPT a lot of times now)
- To add your very own GPT, in the top right-hand corner, let’s click on the button that says “Create.”
- This now opens up an interface where we can build our very own custom GPT. Over on the left-hand side, we have two different options. You can click on “Create,” and this opens up a conversational interface that you can use to build your AI assistant. You can also jump into “Configure,” and here you can manually go through and fill out all of these different fields to construct your AI assistant.
For Example – Let’s say we need to make a chatbot that helps people with baking cookies.
- We type that in and it comes back with a name, which isn’t actually bad so let’s take that. We can also ask for a picture for our chatbot and the DALL-E AI integrated into it which does image generation and gives us a pretty cute profile picture.
On the right side, you can see the Preview of our chatbot simultaneously.
- On the right side, you can see the Preview of our chatbot simultaneously. Following that, GPT will ask us further questions to gather as much information as possible to make this custom GPT useful and personalized. Whatever we enter will also automatically be summarised and passed to the configure tab too as instructions.
- Over on the right-hand side, let’s now test out our new AI assistant, the Cookie Helper. Right down below in the text field, letโs type in, “I want to bake my favorite cookie, chocolate chip cookies,” and see what it comes up with. Remember we have enabled โweb browsingโ for our chatbot so it can go on the internet and search for our queries.
We now get a recipe back for chocolate chip cookies and looking at this, it has all the ingredients, and here are all the instructions. I might have to make some chocolate chip cookies tonight by looking at this.
- Let’s try asking a question that it won’t know anything about. Here at the Kevin Cookie Company, we have a cookie called Clippies Scrumptious Cookies. Let me ask the AI to see if it knows how to bake these. Let’s send that through. Turns out it doesn’t really know about it.
Over on the left-hand side, there is a section for knowledge, and this is where you can contribute additional files to this GPT. This section right here is what makes custom GPTs so special as in this section we can put in our data and make the BOT work specifically on that.
As you can see, we can upload our data in the BOTs knowledge section and itโll analyze that and work accordingly. For example: Imagine that you’re a student in school. You could build a custom GPT for every single class that you’re taking, and imagine that you upload the textbook as knowledge into your GPT. Then you could ask questions against your textbook. The use cases of this thing are infinite.
- Next up, we have something called Actions, through which we can create a new action. With actions, you can talk to external services and their APIs. You have APIs that can get the weather, stock prices, sports scores, or maybe your organization has its own set of APIs. You can now interface with all of that data and those APIs using conversation via your GPT, this makes data very accessible.
Let’s enter an API of the Kevin Cookie Company. This API allows users to find cookie stores in a specified location, and then it’ll provide back that list.
- Next,ย if you want to publish your GPT. In the top right-hand corner, click on the button that says, “Create.” This now opens up a prompt, and you can decide who you want to share your GPT with.
Who can use these :-
- Students can upload their textbooks and papers in the GPT and then ask just about anything to it, from math problems to coding questions, itโll follow all the instructions and data given to it.
- Finance enthusiasts can use it to analyze the market and make decisions accordingly, they can put in the past trends and ask for the probable future outcomes in the financial world.
- The health industry can put all the medical guides and procedures into it to quickly resort to it in case of emergencies and complications.
A lot of people can just use it for niche personal uses like we just did to illustrate the setup of the BOT and enjoy AI to its heights.
Also Read