This month, Google unveiled its latest effort to dethrone ChatGPT from its reigning position as the top generative AI chatbot.

Bard – now renamed Gemini – was released in early 2023 following OpenAI’s groundbreaking launch of its LLM-powered chat interface. Initially, it seemed like Gemini was playing catch-up.

From the start, Bard benefited from internet access thanks to its integration with Google’s search technology. In contrast, the launch version of ChatGPT was limited to the knowledge it was trained on.

However, OpenAI soon enhanced ChatGPT by adding connectivity and the ability to access external information through a partnership with Microsoft’s Bing. Despite these improvements, the consensus remains that ChatGPT is more versatile and useful for a wider range of language processing tasks.

Now, Google

is pulling out all the stops, rebranding Bard with the name of the powerful language model driving its functionality: Gemini. They are also offering access to its Advanced service through a subscription, competitively priced to go head-to-head with ChatGPT.

So, is it ready to step into the ring and go toe-to-toe with the undisputed champion? In this overview, I’ll highlight the key differences between both platforms, helping you decide which one best suits your needs.

Comparing Key Language Models

First, it’s important to note that both Gemini and ChatGPT are built on incredibly vast and powerful large language models (LLMs), far more advanced than anything previously available to the public.

Remember, ChatGPT is merely the interface through which users interact with the language model – GPT-4 for ChatGPT Pro subscribers and GPT-3.5 for free users.

In Google’s setup, the interface, now known as Gemini (previously Bard), facilitates communication with the language model, also named Gemini (or Gemini Ultra for those using the Gemini Advanced service).

It’s crucial to recognize that while both are referred to as chatbots, they serve slightly different purposes. ChatGPT is geared towards facilitating conversational problem-solving, akin to chatting with a knowledgeable individual. On the contrary, Gemini appears tailored to streamline information processing and task automation, aiming to enhance user efficiency.

In terms of technical prowess, the size of LLM models is often gauged by the number of parameters within the neural network. While reports suggest GPT-4’s networks boast around a trillion parameters, concrete details regarding Gemini’s parameters remain elusive. Nevertheless, the sheer power of both models is undeniable.

As highlighted by AI professor Subbarao Kambhampati from Arizona State University, the quantitative metrics of LLMs are becoming increasingly similar. Hence, what truly matters is how these models are fine-tuned, trained, and presented to aid users in problem-solving tasks.

The Ultimate Verdict: Crowning the Top Language Model

After engaging in numerous conversations covering diverse topics using both platforms, it’s evident to me that ChatGPT maintains its superiority as a chat interface, primarily due to the robust capabilities of GPT-4. However, Gemini is steadily narrowing the margin!

Following multiple discussions across a spectrum of subjects with both platforms, it’s apparent that ChatGPT remains the stronger conversational tool, attributed largely to the formidable support from GPT-4. Nonetheless, Gemini is making strides in catching up!

Having conversed extensively on various subjects using both platforms, it’s apparent that ChatGPT retains its edge as the premier chat interface, chiefly due to the impressive capabilities of GPT-4. However, Gemini is steadily bridging the divide!

Advanced Methods for Finding Relevant Data

One advantage of Gemini lies in its default consideration of a vast array of information sources, including the internet, Google’s extensive knowledge graph, and its training data.

In contrast, ChatGPT often relies solely on its training data to answer questions, occasionally leading to outdated information. While you can prompt ChatGPT to search the web for the latest data, this introduces an additional step that Gemini has demonstrated is unnecessary.

In my experience with both platforms, Gemini proves slightly more proficient than ChatGPT in online searching and integrating found information into responses.

However, when ChatGPT does search online, its responses tend to lose dynamism, often relying on a single web search and source of information rather than conducting a comprehensive analysis of available data.

For instance, when seeking an overview of a company or its products/services, ChatGPT often regurgitates marketing material from the website. In contrast, in my limited testing, Gemini provides a more nuanced summary, attempting to offer a balanced overview of features.

In this aspect, Gemini holds a slight edge over its rival. However, ChatGPT excels in intelligently parsing its training data to formulate responses, making it the winner in that regard.

And The Winner Is…

In assessing these AI models, it appears that neither ChatGPT nor Gemini emerges as the clear victor. Gemini demonstrates superiority in formulating answers from internet-based text sources, while ChatGPT excels in addressing queries that do not require internet access. Ultimately, the choice between these models depends on the specific use case and user preferences.

Exploring Cross-Modal Functionalities

Multi-modal AIs have the capability to process various types of data. In earlier iterations, ChatGPT solely handled text but gained multi-modal abilities with the upgrade to GPT-4, allowing it to process visual and audio data. In contrast, Gemini was inherently multi-modal, although not all features were immediately activated.

ChatGPT utilizes the DALL-E model developed by OpenAI for image generation, while Gemini employs Google’s Imagen 2 engine. Both are powerful, generating impressive results. However, ChatGPT tends to be more consistent in creating images closely aligned with the user’s prompts on a same-prompt basis.

Some have noted that Imagen 2 and Gemini excel in producing photorealistic, highly detailed images. Conversely, ChatGPT shines in managing spatial relationships between objects and creatively interpreting prompts.

Both platforms understand and write computer code across various programming languages, albeit with slight differences. The beauty lies in the fact that proficiency in programming is not required with ChatGPT or Gemini at your disposal.

ChatGPT’s superior conversational skills offer significant advantages in code-related tasks. It excels in generating clear guidance, suggestions, and tips, making it beneficial when uncertainties arise regarding code functionalities or integration methods.

And The Winner Is

I’ll hand this victory to ChatGPT once more. While Gemini excels in creating photorealistic images, ChatGPT prevails in generating images that precisely align with user prompts. Although Gemini shows some prowess in technical code generation, it falls short compared to ChatGPT’s conversational interface, particularly during the building and experimentation phases.

(Quick note: Gemini’s image generation is not yet available to users in Europe – hopefully, it will be added soon.)

The Ultimate AI Showdown: Determining the Champion

While both platforms have their flaws, they’re far from perfect. They still suffer from hallucinations and occasionally provide incorrect information. For instance, Gemini erroneously claimed that OpenAI’s DALL-E 2 doesn’t utilize diffusion model technology, and ChatGPT falsely stated that Gemini lacks image generation capabilities.

However, if I had to choose just one to subscribe to, my money would be on ChatGPT Pro for now.

There are a few caveats to consider – if you heavily rely on Google’s ecosystem, Gemini’s integration with Gmail and Google Docs might be a major draw for you. Similarly, if you’re an experienced coder and coding is your primary focus, Gemini is worth exploring (though Microsoft’s Co-Pilot is also worth a look).

For tasks like writing, document creation, summarization, general-purpose image generation, and learning through conversations, I believe ChatGPT currently holds the edge. For these reasons, it maintains its position as the top choice available.