DEV Community

Cover image for A Research On The 2 Popular AI Language Models
Simon Ugorji
Simon Ugorji

Posted on • Updated on

A Research On The 2 Popular AI Language Models

#ai

In this ever-evolving world of artificial intelligence, there is a groundbreaking technology taking center stage: language models. These remarkable AI models are revolutionizing how we humans interact with computers and navigate through massive volumes of text.

Powered by advanced deep learning algorithms, these language models possess an extraordinary capability to comprehend, generate, and respond to human language in a manner that closely emulates human intelligence.

Currently, there are 2 popular AI language models;

ChatGPT is a chatbot developed by OpenAI and released in November 2022. It is trained to follow instructions in a prompt and provide a detailed response.

Bard AI, which is developed by Google and released in March 2023. It is a creative and helpful collaborator to supercharge your imagination, boost productivity, and bring ideas to life.

In as much as Bard is experimental, it still helps to perform most of the tasks and provide detailed responses wherever and whenever it is needed.

In this article, I will be sharing some differences and features between these 2 great language models.

Some differences & features Between BardAI and ChatGPT

Session Management

When using ChatGPT, each chat is treated as a separate session and this means that any information or data used in that session only remains in that session.

session_management_1

This is particularly useful if you want the AI to store information or data that will be used later in the conversation. For example, you can prompt the AI to address you by a particular name throughout the entire conversation.

session_management_2

And you cannot request the same information in a new chat session.

session_management_3

However, with Bard AI, there's no session management and this means that you can ask the AI to recall information that was shared earlier as long as the previous chats are still there. But if you happen to reload the page (or reset the chat), then the AI tries to answer your question using information that was shared by someone else 😕.

session_management_4

However, this language model is still in its experimental state and I hope they improve on the ability for it to recall information.

I can say ChatGPT passed this test ✅

Programming

We all know that programming is not an easy job, especially during the phase when you are trying to find a solution to a problem.

With these 2 great language models, you have an edge over other programmers because you can easily generate, debug and review your code and this can help improve your productivity.

However, there are a few things that you need to keep an eye on when using any of these 2 language models in programming.

Code Generation

When generating a code, think of the compatibility of that particular code that was generated and then ascertain if it can be executed on a different platform.

Here's a prompt for ChatGPT to generate a code that uses the speech-recognition API to listen and transcribe the spoken words.

programming_1

In the image above, you can confirm that the AI generated the code that was needed and on the first line of this code, you will notice that it had placed a check to determine if the code (the speech recognition API) is compatible with the user's browser.

On the other hand, when using Bard AI to generate a similar code, it generated the code in Python programming language and I had to explicitly tell it that it had to generate it on JavaScript.

programming_2

Still, the response is quite poor because the API does not check for compatibility and it's amazing that this code cannot be executed on Google Chrome🙂. The reason for this is that it is not compatible with the browser.

programming_3

So I guess, ChatGPT passed the code compatibility test ✅.

Code Explanation

Both language models were able to explain the code it had generated, line by line and I can say that they both have good code explanations.

I can check that off my list and this means that they passed the code explanation test 😅

programming_5

Code Debugging

We all know how relieving it is for us to debug our code almost immediately as the bug is spotted.

Consider the bugged code below;

(function() {
  (function() {
    console.log('Hello')
  })()

  (function() {
    console.log('World!')
  })()
})()
Enter fullscreen mode Exit fullscreen mode

Credit: https://victorzhou.com/blog/a-javascript-bug-i-had-once/

When I used BardAI to debug this code, it was able to rewrite the code in such a way that you wouldn't have to experience the bug in the first place 😅 and this is not helpful.

programming_6

However, ChatGPT was able to correctly debug this code and it was also able to explain why there was a bug in the first place.

programming_7

So I can say that ChatGPT passed this test ✅

I can go on and on on programming cause it is my line of business but I would leave it to you to further explore these 2 language models.

Internet Access

We all know how important it is to have access to current information on the Internet.

It is quite unfortunate that ChatGPT is not able to action any request that involves accessing current information as it has a knowledge cut-off date (this is a specific point in time until which an AI or information system has been trained or has access to information).

However, with BardAI, you can get access to current information on the Internet and if this information involves the name of a person, please do well to use an underscore to separate the name of this person as it would have trouble looking up that information.

Prompt: look up simon_ugorji on Google

internet_Access_1

With BardAI, you can look up information on social media such as Instagram as well.

Below, I prompted the AI to look up the Nigerian chef, Hilda Baci's account on Instagram (Hilda Baci is a passionate chef that cooked for 100 hours straight in a bid to beat India’s record in Guinness World Record).

Internet_Access_2

From the information above you can see the number of followers she has, the number of people she's following, etc.

I must admit it, this is a very powerful and useful feature.

internet_Access_4

Mathematics and Logical Questions

You can as well prompt these 2 AI language models to answer mathematics and logical questions and I can say that they passed this test ✅

I think the reason why some developers are afraid of being laid off due to the advancement of AI in this field is because they are scared to use it. A fun fact is that I was one of those developers and right now you cannot imagine how happy I am to have it as my companion.

I listed only a few differences and features of these 2 great AI language models and I'm sure there's more especially when you use it to generate a writing or an article.

However, there's always room for improvement. If you are a researcher for any of these popular AI language models, I believe you've read about something in this article that you need to work on 😉

Thank you for reading this little piece of mine about AI.

Psst! I'm open to web development and Technical writing roles

Image credit: Alex Knight on Unsplash

Original Article: https://simon-ugorji.medium.com/a-research-on-the-2-popular-ai-language-models-79d83781f865

Top comments (0)