Name: GptPanda
Uploaded: 2024-12-14T12:21:30.315Z
Duration: 2 min 33 s
Description: An AI-powered Slack integration that brings ChatGPT's capabilities directly into your workspace, enhancing team collaboration and productivity.

gptpanda

An AI-powered Slack integration that brings ChatGPT's capabilities directly into your workspace, enhancing team collaboration and productivity.

GptPanda

So let's see how we use GptPanda. You start by signing up. You can start for free and you will get 1000 credits that you can start experimenting with. 

So here I would usually see my models and my projects. I don't have any, so I hit create a new project. 

So here's our first project. Let's say my project is to get insights from startup news. So that's my project. I will just rename it startup those insights. 

And here we should describe what the model should do. So it can be as detailed as you wish. And in fact, if you can be more details, you'll probably get better results. But for this example, I'm gonna stay big. Like who has time for details? I will just say get insights from startup. 

News. Let's see go to the next step. So here you can see that GPT Panda is generated for you the inputs and outputs. 

So here what you see are the labels for typos. So imagine later you want to use the model later or maybe you want to share it with someone. So if the end user need to know what to expect need to know what type of input to enter and what output they will get. 

So here these inputs and outputs were suggested automatically by GPT Panda. So you enter the article text and then you get the startup name funding investor sectors you found it. Whatever you can add you can you can like you can delete and add different labels. But for me I think I'm happy with this one. So we

Click the next step. So here if you have trending examples. So let's say you have a different style of getting the insights. Maybe you want to generate some summary um in your specific way. So here that will help very much if you have training examples. 

So you can you enter the text and then you enter the outputs and then you save the example um so the model learns later to do something very very similar. So wait a little bit to see how you can generate amazing examples. You can also upload the CSV file if you have a lot of data that you want to upload. But that's for later. 

So now let's just keep it simple. We don't have examples. We go to the next step and that's it we're basically. 

Done GPT Panda already generated for us two models. One is based on 3.253.5. One is on four. The fine tune model is not active simply because we didn't provide training examples. Like you would need at least 20 training examples to fine-tune the model. But I'll show you how to do that. 

We have a very easy way to generate training examples. Let's first test these two models and see which one is performing better. So we have here the input label which is article text. That's input label we defined before. And here all I have to do is just enter some article text which I copied from TechCrunch about one startup. I would hit test. Amazing! Now I have the results. So I have the results.

From both models they both extracted the data that I required or requested. Um the the data are quite similar to be honest between two models. Um which is not always the case. Like in some in some use cases in some tasks you can see a huge difference between GPT-4 GPT 3.5. So it depends on your use case. 

But that's why you have GPT Pandas so you can compare. Uh you can also compare the cred the the the cost here. So here is two GPT Panda credits here is almost like a hundred. It's a huge difference. 

So in this case like if I'm if I'm happy with the output it's clear that I'm going to go with 3.5. However let's say I'm not really happy. So what can I do? I can go to edit project or I can add. 

More details to the uh to the to the task. Um like the first step or I can do even something better. I can use those outputs as the training examples. So if I hit here on Save example they have this input and output already entered for me. I can edit them like if I'm not happy with the out with all the outputs. I can edit some and I have saved them as examples. 

And then I go back to the models and then the new models will actually be implementing those examples that I saved. So here very easily I can just run a couple of tests and generate lots of examples. And maybe at some point um if you find yourself like okay like your examples are quite different then maybe fine tuning could be. 

Great. So then you'll be able to fine tune and then compare the three models. Not only not only the The Prompt models but also the fine-tuned one. In this case, it's um we'll keep it simple like it's super easy. It's super easy task. 

So I'm going to publish the 3.5 model. So here I can see my published models. I can edit them to edit any labels, uh add description as name. I can share the link and I can test them. 

So that's the model that we trained which is a startup insights model. And that's that's a dedicated page that you can share with others or you can use for yourself. 

And here again all you need to do is enter the input. So um paste into your um article this time and let's see how it.

Works yeah you see again different articles. We get also extraction and the user here can edit immediately. So they can read it if they are not happy with some results. And then they can copy it. Um every single output can be added. It can be copied and used somewhere else. And of course they can share the link with others if they wish. 


Note, that there may be slight differences in text and video. Watch video for the original content.

GptPanda Video