ChatGPT vs. Google Bard

Kiriti Amaravadi
4 min readMay 24, 2023

In and out of ChatGPT —

The above two mentioned are the sensations that arose recently. The word chatGPT can be expanded as Chat Generative Pre-trained Transformer. This expansion means that, this is an artificially intelligent model that is trained using data. Any model in AI needs to be trained using some data which is then used to make any sense out of that. ChatGPT was produced by OpenAI. OpenAI is an artificial intelligence research laboratory and company founded in 2015. Its main purpose is to present all the benefits of AI to humanity. Concerning about the privacy and security issues, the data that is specifically used to train this AI model was not disclosed to public. ChatGPT is designed in such a way that, it is trained with the data from 2017 to September 2021. Most of the information that was available in that timeline was used to train the model. This gives a disadvantage to chatGPT in being updated with the current day trends. The prompts that are produced when a question is presented to the model are in such a way that, they cover almost every possible detail regarding that question.

HISTORY —

Previously, there were versions of the GPT by OpenAI.

  1. GPT: This was first designed in 2018. The first attempt of presenting an AI that produces answers by taking prompts as inputs was not completely successful, but opened most of the doors that can give chances to succeed. Internet data at that time was used to train the model. Number of parameters used was 117 million.
  2. GPT-2: This version of GPT was released in 2019. This is the updated version of GPT, which was more larger than the previous version. Its text and answer generative capabilities were astonishing. Everyone was shocked by the improvements that were taking place in the field of AI. Also, the researchers and scientists were fearful about the potential misuse that can take place when this version will be released to the public. This was the major reason that stopped OpenAI from releasing GPT-2 to the public. But later, the scientists were convinced somehow that this might help in enhancing the solutions for research and development issues in those days. This made the official public release of GPT-2 in November 2019. Number of parameters used was 1.5 billion.
  3. GPT-3: This version of the GPT model released in 2020, turned out to be another improved version of GPT-2. GPT-3 marked one of the significant milestones of the Natural Language Processing domain at that time. The model was able to produce more relevant answers to the prompts that were presented to it. As the time passed, scientists and researchers were able to bring out the best out of technology. Number of parameters used was 175 billion.
  4. ChatGPT: ChatGPT is the updated versions of all the GPT models that were previously produced. The relevance of information that will be produced using chatGPT will be the highest amongst all the preceding models.

One of the important common thing about all the versions is each version was the biggest and most powerful AI tool that was put into practice at that time. Also, the number of parameters that are used to train the model varied as the version changed. The complicated nature of the model also increased as the version changed.

Number of parameters ∝ Complexity of GPT model

Complexity of GPT model ∝ Relevant output

What is Google Bard?

Google Bard is the AI that is developed by Google. It is a Large Language Model (LLM) chatbot. LLM is a part of AI model that is used for whole purpose of text input and text output. Unlike LLM, AI models can be of various types such as Image Recognition Models, Speech Recognition Models and what not. But, LLM is purely based on text. The prompts that are put into Google Bard are processed by considering the data that is used for its training.

Data used for its training is not limited to a specific time period such as for ChatGPT. Data used for its training is constantly updated with the current day trends. This can show that, Google Bard’s output can be more relevant to the current day reality unlike the factual information produced by ChatGPT.

Google Bard is currently limitedly released to a few people, but chatGPT is released to everyone. The updated version of which is released as a paid version that has more relevance of answers to various prompts presented.

Insights that can be personally drawn —

The purpose of usage of each model can vary. If at all we need to get factual information straight out, then chatGPT should be the first preference. But, when it comes to more relevant information to the current day reality, then Google Bard can handle the situation really well.

ChatGPT — https://chat.openai.com/

Google Bard — https://bard.google.com/

Thanks!

Kiriti Amaravadi.

--

--

Kiriti Amaravadi

Passionate tech enthusiast and CS master's student at UNC Charlotte. Exploring AI, data science, and innovation to make a positive impact. Let's connect!