The world of artificial intelligence (AI) is rapidly evolving, and AI-language models have become a critical component in this space. As AI continues to become more advanced, language models like ChatGPT and Bard are leading the charge in understanding natural language and generating human-like responses. Both of these models have gained significant attention in recent months, with many people wondering which is the better AI language model. In this article, we’ll take a deep dive into ChatGPT and Bard, comparing their strengths and weaknesses to determine which one is the reigning champion of AI language models, as well as how this impacts Google and Microsoft over the long term.
Introduction
ChatGPT
ChatGPT is essentially an app built on GPT-3, which is an AI-language model developed by OpenAI, designed to generate human-like conversational responses to the input it receives. Generative Pre-trained Transformer 3, GPT-3, which is an autoregressive language model released in 2020, uses deep learning to produce human-like text. Given an initial text as a prompt, it will produce text that continues the prompt. It has been trained on a massive amount of text data from the internet, allowing it to generate text in a wide range of styles and formats, including storytelling, summarization, translation, content creation, and more.
Using ChatGPT is quite straightforward as seen in the image above. All a user has to do is type in his/her questions or any of the tasks discussed above into the chat box and ChatGPT will quickly process your demands and provide you with a human-like output. The amazing part about ChatGPT is the fact that you can have a full conversation with it since it remembers and uses the existing conversation context such as things you asked/discussed with the bot in the chat. Within the first 2 months of its launch, ChatGPT has reached more than 100 million users and has more than 13 million daily visitors as of 2023.
Bard
Bard is a state-of-the-art AI-language model designed by Google that can generate human-like text responses to prompts. This model is based on the GPT (Generative Pre-trained Transformer) architecture and has been trained on a massive amount of text data from various sources.
One unique feature of Bard is its ability to generate text that stays on topic for a longer period than previous language models, similar to ChatGPT as mentioned above. Bard can also perform a wide range of language tasks, including language translation, summarization, and question-answering. Additionally, it has been trained on specific domains such as Wikipedia, meaning it can generate highly accurate responses to questions and prompts in these domains.
Bard’s design is focused on improving coherence and accuracy in language generation, which has been a challenging task in the past for AI-language models. Its impressive ability to understand the context of the text has made it a promising model for tasks such as content creation, dialogue systems, and customer service chatbots.
Failed Launch
On the 7th of February, Google had their “Live in Paris” event which was supposed to announce the new Google Maps and Google Search features which have been in the works for a while now. On top of this, they actually launched a sneak preview of Bard, which turned out to be a disaster.
In a Twitter advertisement, Bard was asked a question: “What new discoveries from the James Webb Space Telescope can I tell my 9-year-old about?” In response to this simple question which was intended to exhibit how Bard can make complicated things simpler and hence the average users’ lives easier. Bard gave three responses.
Albeit the first two responses were correct, unfortunately, the last response: “JWST took the very first pictures of a planet outside our own solar system.” was factually incorrect. This response caught the eye of astronomers and experts in the field and soon the internet was swarmed by individuals pointing it out. As the fact is that the first pictures outside our solar system were actually taken by the European Southern Observatory’s Very Large Telescope, which NASA confirms.
Now the underlying reason behind this mistake was that Bard relies heavily on Google Search and when you search up the same question on google you do find a link that puts these two words together leading to the mistake. In a nutshell, Bard is currently incapable of realizing that the fact that two words are found together does not guarantee that the answer is indeed correct and hence needs some refining in that aspect.
What Happened To Google After Bard’s Failed Launch?
The stock market is a nasty place and can be very unforgiving to mistakes such as these. Hence, when the debut/blunder was revealed, it was time for action. In response to the blunder, Google’s stock nosedived by 7.7% on the same day (Wednesday). And lost a further 4.4% on Thursday when it fell to 95.01 at the day’s close. On the last day of trading (Friday) however, the stock managed to finally press the brakes as it ended the week closing at 94.69. On Wednesday alone, this caused Google’s parent company, Alphabet, to lose approximately $100 billion in market value.
Despite all this, I feel that the rapid selloffs were merely an overreaction by the market due to a poor ‘debut’ (not even a launch). Bard is definitely still in its development stage and has a lot more improvements to be made before it can be rolled out officially. However, the success of Bard is critical to the success of Google moving forward, and here’s why.
Microsoft is a prominent investor in OpenAI (creators of ChatGPT) and Microsoft’s search engine (Bing) has always lagged behind Google. Microsoft is currently gearing towards the opportunity created by ChatGPT to finally cover their gap with Google as they plan on integrating ChatGPT with their search engine, Bing. Finally giving users a noticeable incentive to switch from Google to Bing. Hence, the power struggle you see between ChatGPT and Bard is essentially a cold war between Alphabet and Microsoft.
ChatGPT vs Bard
Well, both AI-language models are indeed life-changing innovations for mankind. Even though their services are quite similar as both of them are programmed to produce human-like responses in response to the queries they are asked, they do come with their own set of differences.
#1 Underlying Model
ChatGPT currently uses GPT 3.5 for dialogue applications which helps it generate content in response to a single text input. It employs deep learning techniques to understand the context and meaning of a given input, and then generates a response that is appropriate for the context.
Bard, on the other hand, derives its responses from the internet and employs Google’s Language Model for Dialogue Applications (LaMDA). With the use of an open-source network, LaMDA was developed to understand human-like language. Instead of looking for individual words to create dialogue, it has been trained to seek for patterns in phrases and the spaces between words.
Hence, it seems that Bard will be able to provide more in-depth answers than the typical Google search through this language model as it produces more information in longer chunks.
#2 Training Data
Secondly, as mentioned above, Bard is trained with data coming straight from the internet. In comparison, ChatGPT is only trained with data up till 2021. It is not officially confirmed if the model is trained with updated data sources from 2021 and 2022. Being able to obtain the most recent data off the internet, will put Bard ahead of ChatGPT as Bard will have access to real-time data which could be more accurate based on the user’s needs.
#3 Financial Support
As many would know, training AI models tend to be very costly as you need to gather billions of data points, pre-process them and then train the model against this data set over a long period of time, over several epochs. As such, these models require extensive amounts of investments in Research and Development (R&D), and hence, strong financial support is critical to their success.
When it comes to financial support, it may seem like Google with its humongous resources should be able to surpass ChatGPT in no time. However, that might not be the case because ChatGPT cannot be underestimated either as the company that created it, OpenAI, was not only founded by rich and influential individuals like Elon Musk but Microsoft is also a major shareholder of the company and announced a multibillion-dollar investment in OpenAI last month.
Final Thoughts
When we compare the technology and underlying model, I would favor ChatGPT over Bard due to 1 simple reason. ChatGPT has launched for quite a while now, with over 100 million active users at the moment, asking billions of questions every day. Each question and generated response can be funneled back into ChatGPT’s training pipeline, allowing it to learn from its mistakes and improve upon itself. This gives them a huge pipeline of data points to train against, ultimately improving its accuracy overtime. On the flip side, Bard is still in development, which could hinder its progression as it lacks real life user interaction.
Regardless which AI-language model you favor, we cannot deny the fact that these 2 products will be a huge cash cow for their respective companies. Microsoft has already started to roll out implementation across some of its product suites and we can definitely expect more to come. As for Google, we might have to wait a while longer before we see more of Bard in action.
I want to leave you with 1 last interesting thought before I end off this article. What if Bard and ChatGPT merged into 1? Just imagine, with ChatGPT’s AI-language model and Bard’s access to the internet, it could really create a monster of a product. Perhaps we might see this over the next few years as these 2 tech giants merge their respective AI-language models into 1.