Tomsguide tech journalist Ryan Morrison has had initial experiences with Gemini Advanced to answer the question many users have: Can Google’s latest AI creation compete with ChatGPT?
Google has recently introduced its newest artificial intelligence model, Gemini Ultra 1.0, claiming it can directly compete with OpenAI’s most powerful AI model, GPT-4-Turbo, in most tasks. If proven accurate, Gemini Ultra will mark the first true competition with OpenAI’s ChatGPT since the AI generative explosion began.
In a somewhat confusing move, Google also decided to change the chatbot’s name from Bard to Gemini, aligning both the AI model and the model access service under the same name.
In light of this, Tomsguide tech journalist Ryan Morrison had initial experiences with Gemini Advanced – the paid version of Gemini priced at around $19.99/month, which includes more features and processing power than the free version. Morrison then provided preliminary insights to answer the question: Can Google’s latest AI creation compete with ChatGPT, especially ChatGPT Plus?
Programming Capability
When encountering any new chatbot, Morrison always checks its programming capabilities, a skill that artificial intelligence should naturally perform. Google has promoted the programming capabilities of the Gemini Ultra AI model quite well. To verify this, Morrison requested Gemini Advanced to assist in designing a website.
First, the Tomsguide journalist had an image of an anime and pixel-art-style scene of a cat sitting on a fence. Morrison then sent this image to Gemini, along with the idea of the cat playing the role of a “cat spy” named Cat-astrophe in a fictional TV show. Next, Morrison crafted a prompt asking Gemini to design a StyleSheet (including style rules to inform the browser how to present a desired document) based on the colors in the image.
Although the result was not perfect due to selecting a limited number of colors, Gemini successfully created a usable stylesheet. Morrison then requested the chatbot to create a website using the stylesheet to introduce the fictional TV show about the spy cat Cat-astrophe. Surprisingly, Gemini Advanced performed this task well. All that was needed from the user was to provide the correct image path.
Image Generation Capability
In terms of image creation, Gemini Advanced does not perform this task itself but relies on the Imagen 2 model developed by Google DeepMind. However, the enhanced inference and understanding capabilities of Gemini Advanced promise to facilitate smoother collaboration between users and chatbots, thereby producing higher-quality images.
To test Gemini Advanced’s image generation capability, journalist Ryan Morrison requested Gemini Advanced to create an image with the theme “Cat on the Moon,” along with the request to add the text “Cat on the Moon” below the image. The result was two images as requested.
However, because it had achieved a similar task with Bard before, Morrison wanted to challenge Gemini Advanced by issuing a more complex command: “Place the cat in a space suit and place it on the surface of the Moon with the Earth in the background, while maintaining the image style from the previously created image.” Gemini Advanced successfully fulfilled the request but forgot to add the text as requested. When asked to add the text in the bottom left corner, Gemini Advanced created a new image and placed the text in the top right corner. Although it did not achieve the desired result of the user, this is understandable as other AI platforms also struggle to maintain consistency when generating images.
This result shows that while Gemini Advanced demonstrates great potential in supporting image creation, it still requires adjustments and improvements to achieve consistency and accurately meet user requirements, especially when compared to the flexibility and understanding capabilities of other AI platforms like ChatGPT.
Poetry Composition Capability
In the field of poetry, artificial intelligence (AI) has begun to excel, especially in mimicking or learning the styles of famous personalities. To explore this capability, journalist Ryan Morrison requested Gemini Advanced to compose a Shakespearean sonnet on the theme of the battle between gaming systems like Xbox and PlayStation. In the first attempt, although the sonnet created by Gemini Advanced read fairly well, it ‘forgot’ to mention specific gaming systems.
To further challenge it, Morrison asked Gemini Advanced to focus on the idea, and the result was an impressive poem. This demonstrates that while composing poetry is no longer a significant challenge for AI, using common poetic words to create an appealing line is not straightforward.
Morrison expected creative and humorous results using common words like “of,” “is,” and “or,” but did not receive the desired outcome, forcing multiple attempts. In contrast, ChatGPT captured the user’s request accurately from the first attempt.
Conclusion: Good but Still Needs Improvement
Overall, while Gemini Advanced shows remarkable progress compared to previous versions with the ability to provide logical and structured answers to complex queries, there are still limitations in generating responses as expected by users, according to journalist Ryan Morrison’s conclusion.
Clearly, Gemini Advanced is still in a phase that requires further optimization and refinement, not to mention additional features that Google needs to add. On the other hand, the quality of responses (when executed correctly according to user requests) from Gemini Advanced is quite good and can rival ChatGPT Plus.
The remaining question is how Gemini Advanced will be improved as Google continues to develop this chatbot in the future, based on its own adjustments and direct feedback from users.
See more information of AI in here.
