Google's Bard Vs. ChatGPT: Who Wins the AI Battle?

On November 30, 2022, a company in San Francisco, USA, debuted its most ambitious AI project that is ChatGPT. And within five days of its launch, ChatGPT got 1 million users worldwide. This feat is what Silicon Valley products like Facebook or Instagram could achieve in the past, as even Instagram took 2.5 months to get a 1Million user base.

ChatGPT, developed by OpenAI, became a massive hit and seems like an overnight success. But the research and development powering the GPT model that powers it took decades, going back to the 1980s when people in the field of Artificial Intelligence started working on recurrent neural networks (RNNs) - a sort of AI that can process sequential data such as text and human speech. The breakthrough came in 2017 when some scientists working at Google Brain came up with the Transformer Architecture, which became the basis of OpenAI’s GPT (Generative Pre-Trained Transformer) after GPT-1, GPT-2, GPT-3, we got ChatGPT Chatbot, a generative AI chatbot capable of understanding dialogues and generating creative text content when prompted, poems, speeches, and letters even.

On the other hand, Google, the company with the hegemony in web search over the past two decades, has been actively working on AI since the early 2000s with its Google Brain project and has since made several notable achievements and breakthroughs in the field of AI. As a prominent player in AI, Google has developed numerous AI-driven products and services, which include Google Translate, Google Autocomplete, Google Voice, Google Photos, Google Cloud Natural Language API, Google Cloud Speech API, and the AI chatbot Bard. Beyond this, Google also leads cutting-edge AI research and development and publishes several papers in top AI conferences and journals.

In 2016, a few months after becoming CEO of Google, Sundar Pichai made a sweeping proclamation that Google, whose name had become synonymous with search, would now be an “AI-first” company. After taking the company reins, Mr. Pichai’s announcement was his first order of business at Google’s annual I/O developers’ meet, but believe it or not, Google is not an 'AI-first' company at present, and it is playing to catch up with OpenAI, and Microsoft Inc.

On February 6, just a day before Microsoft announced its new Bing Chatbot powered by OpenAI’s ChatGPT, Google’s CEO, in a blog post, announced Bard, an experimental conversational AI service powered by LaMDA (Language Model for Dialogue Applications). The announcement of Bard appeared to be a pretty rushed one. And a factual error about James Webb Space Telescope (JWST) discovering exoplanets (It was the European Southern Observatory's large telescope that took the first pictures of exoplanets in 2004 and not JWST) in a promotional video for Bard made Google’s parent company Alphabet lose $100 billion (9%) in market value.

So, even though Google has been working on its AI and deep learning developments for quite some time now, it has to cover a lot of ground for generative AI and this AI battle. With so many companies like Microsoft, IBM, Apple, etc., investing heavily in AI and ML research and development, Google certainly has a lot of competitors, and Microsoft’s Bing AI is leading from the first when it comes to challenging Google’s Search Platform.

ChatGPT vs Bard AI - Key Highlights
GPT-3.5 and GPT-4 - The Large Language Model Powering ChatGPT and ChatGPT Plus
LaMDA - Google's Language Powering Bard
Which AI Chatbot is Better - ChatGPT or Google's Bard?

ChatGPT vs Bard AI - Key Highlights

ChatGPT was created by OpenAI— using a specially tweaked version of OpenAI’s Generative Pre-training Transformer 3 (GPT-3) or GPT-4 (depending on the version)— and Bard AI was created by Google. It was announced in February 2023, just a day before Microsoft announced its new Bing Chatbot powered by OpenAI’s ChatGPT, while ChatGPT was launched in late November 2022. Bard AI uses Google’s Language Model for Dialogue Applications (LaMDA).

If we talk about data sources, ChatGPT has trained on a massive dataset of text, including Wikipedia, books, documents, articles, and content taken from the open internet. However, its sources end in 2021, limiting the latest world events and research. And Bard AI trained on Infiniset, a blend of data (set) including Common Crawl, Wikipedia, documents, and conversations and dialogues from the web. Google claims that Bard will search the web in real-time to find the most recent answers to questions and the latest research. Bard combines the breadth of the world’s knowledge with the power, intelligence, and creativity of Google's large language models.

ChatGPT is open for free to all users, ChatGPT Plus, the newer, upgraded version of ChatGPT, based on the GPT-4 model, is available for use at a $20/month subscription to include access to more capabilities, access during peak times, longer content, visual inputs, faster response times, and priority access to new features. Google’s Bard AI is free to use, but the access is not publicly available. However, you can join the waitlist, and soon access will be provided.

Now coming to knowledge and accuracy, ChatGPT is ignorant of events after 2021 due to its training data cut-off. But the Bard AI can search the internet for current information and use that to generate responses. However, Bard AI can also make hard-to-spot errors, including the error mentioned below.

When asked about Google’s AI efforts and progress, the company claimed that it launched the GPT-2 and GPT-3 models, which isn't the case— OpenAI developed and launched them.

GPT-3.5 and GPT-4 - The Large Language Model Powering ChatGPT and ChatGPT Plus

The AI chatbot of ChatGPT works by using natural language processing (NLP) and machine learning algorithms to understand and generate human-like responses when prompted by users. It is based on OpenAI's Generative Pre-training Transformer’s third version (GPT-3) and uses an extensive neural network trained on a massive dataset of text and codes.

When a user inputs a prompt or query, ChatGPT processes the input using its NLP algorithms to understand the meaning and intent behind the user's message. It then uses its machine learning algorithms to generate a response based on its training data and the context of the conversation. The response is thus shown to the user in natural language.

GPT-4 vs GPT 3.5: 5 Key Differences Explained

OpenAI claims its new upgraded version GPT-4 is much more capable than GPT-3.5. Let’s see the 5 key differences between the GPT-4 and its predecessor in this article.

StartupTalkyAbira Dutta

GPT-3, the powerful and advanced language model with 175 billion parameters, can generate coherent and contextually appropriate language. GPT-3.5 is an update to GPT-3 that includes new capabilities and improved performance, and GPT-4 is expected to have an even larger vocabulary size and more advanced deep-learning concepts. GPT-4 is capable of analyzing visual inputs and logical interpretation of images. It also accepts up to 25000-word for better context and creates longer, creative content than ChatGPT.

LaMDA - Google's Language Powering Bard

It was the annual Google I/O event in 2021 that Google announced LaMDA (Language Model for Dialogue Applications). It is an advanced AI language model developed to engage in free-flowing conversations across endless topics and can mimic human dialogue. During its training, LaMDA picked up on several nuances of open-ended conversation to ensure sensible and impressive replies. It can prioritize helpful suggestions and witty retorts, avoiding generic responses such as "Okay" or "I'm not sure."

To guarantee factual accuracy, Google integrated LaMDA with third-party information sources to source facts and avoid generating conflicting or false information— as concerns were raised over the misinformation generated by existing chatbots.

LaMDA is powered by Google's Transformer (deep learning) architecture (developed in 2017), which enables the model to "read" multiple words at once and analyze their relationships with one another. Even OpenAI's GPT models rely on the same Transformer architecture.

Which AI Chatbot is Better - ChatGPT or Google's Bard?

ChatGPT is the most successful AI chatbot at the moment. It took just five days since its launch on November 30, 2022, to get 100m users, and by January, it had 100m users worldwide. While it is too early to conclude which AI Chatbot— ChatGPT or Google’s Bard— is the better one and who wins the AI battle, it is worth commenting that Bard is still an experiment as claimed by Google and makes factual errors like the one mentioned above. And ChatGPT is limited by its training data set that remains anchored to the past, 2021.

That being said, ChatGPT is now upgraded to the newer version, ChatGPT Plus, and it is much more advantageous when it comes to generating creative content, writing codes, or doing some research work.

It is also worth mentioning that the use of these two AI Chatbots depends upon user needs and how one uses these tools, though a careful approach and fact-checking of some facts, once in a while, is certainly needed.

FAQs

What are the major differences between Google's Bard and ChatGPT?

• ChatGPT was created by OpenAI, whereas Bard AI was created by Google
• ChatGPT's data sources end in 2021, however, Bard claims that it searches in real-time
• ChatGPT's knowledge and accuracy are till 2021 due to its training data cut-off. But the Bard AI can search the internet for current information and use that to generate responses

What are the language models in which ChatGPT and Google's Bard work?

The AI chatbot of ChatGPT works by using natural language processing (NLP) and machine learning algorithms to understand and generate human-like responses when prompted by users.

While Google's Bard works by using LaMDA (Language Model for Dialogue Applications).