10 mins read

ChatGPT 4 vs. Bard’s Gemini Pro – What’s The Difference?


When talking about AI, two big names are making news lately: ChatGPT and Google’s Gemini. In this article we’ll compare OpenAI’s ChatGPT-4 and Google’s Bard with Gemini Pro, highlighting how they’re changing the way we interact with technology.


Gemini Pro and GPT4 are both large language models (LLMs) that have been trained on massive amounts of text data. They can generate text, translate languages, write different kinds of creative content, and answer your questions in an informative way. However, there are some key differences between the two models.

A Quick Overview

Latest updates from Google

Gemini Pro is a new LLM from Google AI that was introduced in December 2023. It is based on the same architecture as ChatGPT4, but it has been trained on a larger dataset of text data and it has some new features, such as the ability to generate different creative text formats of text content, like poems, code, scripts, musical pieces, email, letters, etc. Bard Gemini used to be called LaMDA, but it got a big upgrade with the help of a new tool called Gemini. This upgrade helped Bard Gemini become even better at understanding and using language.

  1. Bard Extensions: These allow Bard to access and use data from various Google tools like Gmail, Docs, Drive, Maps, YouTube, and Google Flights and Hotels. This integration is one of the strengths of Google and its models, and it is meant to offer more tailored and useful responses by pulling relevant information from these services​​​​​​.
  2. Improved “Google it” Feature: This feature enables users to fact-check Bard’s responses against information available on the wider web. It helps in verifying the accuracy of the AI’s outputs​​​​.
  3. Modify Answers: You can easily modify an answer using 5 presets provided: Shorter, Longer, Simpler, More Casual, More Professional
  4. User Privacy and Trust: Bard Extensions operate based on explicit user opt-in, and permissions can be revoked at any time. Personal data from Gmail, Docs, and Drive is protected and not used for Bard’s learning or exposed to human reviewers​​.

About OpenAI Models

ChatGPT4 is an LLM from OpenAI that was introduced in November 2022. It is also based on the Transformer architecture, but it has been trained on a different dataset of text data. ChatGPT4 is known for its ability to generate human-quality text, and it has been used for a variety of tasks, such as writing articles, translating languages, and creating creative content. It’s even better at generating coherent and informative text, especially when it comes to open-ended or challenging prompts. ChatGPT4 is also a pro at creative writing and at coding, at least when it doesn’t become too lazy.

  1. DALL·E 3 Integration: Allows for the creation of unique, detailed images based on user descriptions during conversations. It emphasizes responsible content generation and is available to ChatGPT Plus and Enterprise users​​​​.
  2. Voice Integration: Features a voice-chat capability for more natural interactions, particularly on mobile devices. This feature is powered by OpenAI’s Text-to-Speech and Whisper models and is available for Plus and Enterprise users​​.
  3. Image Input Capability: ChatGPT can now process image inputs, enhancing its utility for a range of applications. This feature is available to Plus users across all platforms​​.
  4. Custom GPTs: Users can create custom GPTs for specific tasks like creative writing or trip planning without needing coding skills. This is part of the Plus plan​​.
  5. Assistants API: This API allows developers to build AI experiences within their applications, offering a range of functionalities like Code Interpreter and Retrieval​​.
  6. Code Interpreter: In beta, this feature enables Python code execution and file access, useful for data analysis and complex problem-solving. It’s available to ChatGPT Plus users​​.
  7. Enhanced Messaging Limits: The Plus plan now offers increased message limits for more extensive interactions with GPT-4​​.
  8. Browsing and Search on Mobile: Provides Plus users with real-time information and comprehensive answers beyond the model’s original training data​​.
generate images with dalle-3
DALLE-3 can generate images with ChatGPT Plus

The current situation


ChatGPT-4 stands out for its multimodal capabilities thanks, allowing it to process text and images, along with its integration with DALL·E 3 for creative image generation and voice interaction features. CustomGPTs allow also for more customization and personalized use cases. Google Bard, powered by the Gemini Pro model, focuses on seamless integration with Google’s applications, providing contextually relevant responses by accessing services like Gmail, Docs, and Maps. While ChatGPT-4 excels in creative and interactive tasks, Google Bard leverages its integration with the Google ecosystem to deliver enhanced, application-specific assistance​.

gemini ultra benchmarks
Gemini Ultra’s Benchmark

Something to keep in mind: Gemini Ultra is the most powerful mode introduced by Google, but it has not yet been officially released. It has already surpassed ChatGPT4 in numerous benchmark tests, demonstrating its superior performance, at least on paper. However, the version currently accessible in Bard is Gemini Pro, a smaller iteration of the model, which exhibits abilities that exceed ChatGPT 3.5 but still lag behind ChatGPT 4.

Practical Comparisons

Google Bard: Enhancing User Experience with Versatile Features

Google Bard is renowned for its user-friendly interface, which significantly enhances the overall experience. It allows users to effortlessly modify prompts, integrate responses with Google Docs and Gmail, and share conversations. This smooth integration with Google’s ecosystem makes the tool an interesting assistant for many daily tasks.

Another notable feature is Bard’s ability to vocalize responses. However, the robotic tone of its voice leaves room for improvement. Despite this, Bard’s rapid web search capabilities, available to all users at no cost, are a considerable advantage. This feature is especially useful for quick information retrieval.

Bard also excels in fetching images from the web, broadening its utility. However, it’s not without its drawbacks. While Bard does provide external links and sources upon request, the reliability of these sources can sometimes be questionable. Furthermore, Bard’s experience is somewhat isolated, with limited integration options beyond Google’s own suite of applications.

ChatGPT: Excelling in Text Generation and Collaboration

ChatGPT 4, on the other hand, shines in generating text-based content, such as long-form articles and emails. Its ability to create AI-generated images with DALLE-3 is also one of the best features, especially in terms of following instructions and generating logos or specific art styles.

The voice capabilities available with ChatGPT’s mobile app are also impressive: it can listen to your speech and respond with a very convincing human-sounding voice, making pauses and even clearing its throat at times.

Despite these strengths, ChatGPT does have its limitations. Users have reported that web searches can be slower and less reliable compared to Bard. Additionally, while sharing conversations is possible, ChatGPT doesn’t support sharing them with images, which might be a drawback for visual content creators. Also, in order to get access to all features, a subscription is needed, unless you are happy with the less performant ChatGPT 3.5, which is still great for some use cases like translations or proofreading.

Shared Challenges: The Risk of Inaccuracies

Despite their difference, given the nature of their models, both Google Bard and ChatGPT have the tendency to produce plausible-sounding but sometimes inaccurate responses, known as hallucinations. They might give a convincing response that is completely false, or they might not be able to understand whether they are right or wrong in case you ask them if they are sure about something. Integrations with internet browsing can help reduce these hallucinations.

Prompt Challenges

Let’s make some requests to see how the results differ between the two models: I will be asking the same question and report the results.

Describe an input image

ChatGPT easily described the image in details, whereas google answered that it cannot help with images of people; pretty disappointing, but at least it understood that it was a face.

chatgpt4 image recognition

Trying with another image, also bard gave a good response:

bard's image recognition

Creative Writing

Let’s ask the to compose a short poem about the changing seasons.

Bard’s one is a bit too long maybe, but they both show good creativity.

Information Retrieval

The responses are definitely different; I wonder why Bard felt the need to add these images of the airlines, they are a bit out of context. Also it is not clear if a layover is necessary from Bard’s response.


Conclusion

In conclusion, both Google Bard and ChatGPT offer a range of interesting features. Bard excels in user experience, speed, and integration with Google services, while ChatGPT stands out for its text generation capabilities, image and voice capabilities and customization with CustomGPTs. However, we should always be aware of their limitations, especially concerning the accuracy and reliability of the information they provide.

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.