Open AI’s ChatGPT VS Google’s Gemini

A little over a year ago, OpenAI introduced ChatGPT, and it immediately became a major player in the world of AI. This momentous launch had a seismic impact on Google, a company that has long been a pioneer in the foundational technology of the ongoing AI revolution. Despite branding itself as an "AI-first" organization for nearly a decade, Google found itself visibly unprepared for the exceptional capabilities of ChatGPT and the swift ascent of OpenAI's technology in the industry.

The skyrocketing popularity of ChatGPT chatbot posed a serious threat to Google's business. In response, the company initiated substantial investments to catch up with the generative AI trend. Rumors suggest that a "code red" was declared in December 2022, although Google's CEO Sundar Pichai has denied this claim. Despite maintaining a global search market share exceeding 90 percent, Google announced Project Gemini during its I/O conference in May 2023. This move was a hurried attempt to integrate generative AI into its search functionality. It aimed to counteract ChatGPT's popularity and mitigate the risk of OpenAI's technology significantly boosting Microsoft's Bing search engine.

After an entire year since the introduction of ChatGPT, Google is now ready to make its move. On December 6, 2023, Google officially announced the launch of Gemini, its highly anticipated response to ChatGPT, aiming to challenge ChatGPT's dominance in generative AI. 

Gemini version 1.0 represents the advancement of Google's previously developed large language model PaLM 2 and comes in three variants tailored for various devices and computing capabilities: Gemini Nano, Gemini Pro, and the upcoming Gemini Ultra, set to launch next year.

Gemini Nano, designed for mobile phones on Google's Android system, is the smallest version in the Gemini series. Since December 6, 2023, Gemini Nano has already been working on Google Pixel 8 smartphones, powering the auto-summarization feature in the Recorder app and the Smart Reply in the Gboard keyboard, particularly in WhatsApp, with Google indicating plans to expand its integration into more apps next year. It operates directly on the device, functioning without an internet connection. While currently exclusive to Pixel 8 Pro, Google is actively pursuing broader integration by exploring ways to incorporate Nano into the entire Android system.

Gemini Pro is a medium size version of Gemini, now powering Bard Chatbot. It supports English queries in over 170 countries. Starting December 13, developers and enterprise customers can access Gemini Pro through Google Generative AI Studio or Vertex AI in Google Cloud.

Gemini Ultra, the most powerful version of Gemini, is set to be integrated into Bard and will be accessible through a cloud API in 2024. According to Google, Gemini Ultra has demonstrated superior performance compared to human experts across 57 subjects, spanning disciplines such as math, physics, history, medicine, law, and more.

Gemini vs ChatGPT Comparison: Language Models

ChatGPT runs on a language model architecture created by OpenAI called the Generative Pre-trained Transformer (GPT), primarily focusing on text-based interactions and responses. OpenAI has released other models and systems beyond ChatGPT, such as DALL-E or Codex. DALL-E, for instance, was an AI system specifically designed to generate images from textual descriptions, showcasing the capabilities of AI in creative tasks.

The Bard chatbot platform is powered by the Bard LLM (primarily PaLM 2 for general functions and LaMDA for dialogue tasks) and the Gemini Pro LLM for English queries. From the beginning, Google Gemini was designed as a multimodal model capable of analyzing text, images, video, audio, and code. Right now, Gemini's Pro models are text in and text out. Still, more powerful models like Gemini Ultra are expected to accept and create images, audio, and video in addition to just text. 

Initially, Google was using Gemini's ability to generate text and images to differentiate it from ChatGPT, but on September 25, 2023, OpenAI announced that users could enter voice and image queries into ChatGPT.

Gemini vs ChatGPT Comparison: Data Sources 

ChatGPT's data model is derived from publicly available sources on the Internet, including Wikipedia, books, news articles, and scientific journals. The training data for ChatGPT-3.5 is limited to information up to September 2021 or earlier, restricting its accuracy to that timeframe. Despite ChatGPT-4 sharing a similar limitation, it excels in adapting and responding to current information obtained through ChatGPT prompts. Following the recent release of ChatGPT-Turbo and updates to GPT-3.5 and GPT-4 models, these language models have begun responding their data is constrained and extends only until January 2022.

Gemini AI is different—it's trained on real-time data from the Internet, allowing it to answer questions with the latest info. This massive model handles text, code, images, audio, and other data sources, making it even more powerful than ChatGPT. It is supposed to enable it to generate intricate and nuanced text while handling more challenging tasks.

Gemini vs ChatGPT Comparison: Availability

ChatGPT chatbot has been available to the public on its dedicated website since its debut in November 2022. It boasts extensive country coverage and supports more than 50 languages. However, its proficiency may vary depending on the language, and it may perform best in English due to the training data's focus. ChatGPT usage can be limited in certain countries for various reasons, including internet censorship laws and regulations, such as North Korea, Iran, China, Cuba, and Syria. However, with a VPN and phone number from the supported countries, users can also use ChatGPT in these countries.

Gemini Pro is now available through the Bard chatbot website. While Bard is currently available in more than 40 languages and over 230 countries and territories, only the English version of Bard utilizes the "Gemini Pro" technology. And users can access it with English queries in all covered countries.  

Gemini vs ChatGPT Comparison: Pricing

ChatGPT offers a straightforward pricing structure. Users who sign up for a free account can access the free tier, specifically ChatGPT-3.5, by providing their email address and phone number. For those seeking more advanced features with ChatGPT-4, there's ChatGPT Plus, a paid tier available at $20 per month. This subscription includes an enhanced language model and the capability to integrate ChatGPT plugins, which are all currently free. 

Explore our articles comparing ChatGPT versions and showcasing the best ChatGPT Plugins. Additionally, there's the Enterprise ChatGPT tier for enterprise-level usage, which features on-demand pricing.

As for Gemini, you can use it for free currently. Bard, the generative text AI platform, is the easiest way to get to know the tool, but you can also use Gemini with Pixel 8. What is important to consider is that Google already incorporates various AI versions into its widely used free products like Search Engine, YouTube, Gmail, Google Maps, Google Play, and Android. Since Gemini is seen as an added value for Google, there's a possibility it might be gradually integrated into these services. We can assume that the complete integration of Gemini into all Google products will be gradual, likely combining previous language models with the newer, more efficient Gemini ones. It could mean you won't need an extra monthly subscription, as Gemini is expected to become part of numerous products. However, whether it will be a paid add-on or what enterprise services will be available remains uncertain yet.

Gemini vs ChatGPT Comparison: Performance

Google claims that the "Ultra" version of Gemini can outperform ChatGPT. Initial reports suggest that Ultra surpasses ChatGPT-4 in various categories, showcasing superior accuracy in most tests except for one—HellaSwag, designed to evaluate AI's understanding of real-world context and logic, essentially a common-sense test. It's important to note that these are preliminary tests, and a comprehensive assessment requires further independent evaluations, especially in addressing the tendency of all models to hallucinate.

Bard currently utilizes the Pro model, and the Ultra version is not expected until somewhere in 2024. Consequently, so far, comparing only Gemini Pro with ChatGPT's versions is more fair. When comparing Bard Pro with GPT-3.5, whose performances are quite similar, the results of Bard Pro are not that promising. When comparing it with ChatGPT-4, the latter outperforms Gemini Pro in all tests. Thus, while ChatGPT has its imperfections, Google needs to demonstrate that Gemini can genuinely impress ChatGPT users, which, so far, seems to be a challenge.

Source: https://storage.googleapis.com/deepmind-media/gemini/gemini_1_report.pdf

So, Gemini vs ChatGPT: Which is better? It will take time and more real-world examples to determine if Google has truly entered the competition and if Gemini lives up to its on-paper promises. The ultimate goal is to develop a fully-fledged Artificial General Intelligence (AGI), a complex challenge in AI research due to the broad range of capabilities required for machines to comprehend, reason, learn, and adapt across various scenarios. Who will achieve this, whether it's Google, OpenAI, or someone entirely different—is still uncertain and remains everyone's best guess.

Reading time: 5 min 48 sec