ChatGPT vs. Gemini: Visual Recognition Capabilities Put to the Test
A comprehensive comparison of how two leading AI models interpret the visual world.
1The Rise of Multimodal Capabilities
AI models like ChatGPT and Gemini are now being tested not just on text processing but also on understanding visual data. With users increasingly utilizing camera features for object identification, both models have recently upgraded their multimodal abilities. These tests aim to measure which chatbot better comprehends the real world through visual inputs, marking a significant shift in AI evaluation criteria beyond pure text analysis.
2Handwriting Recognition Challenge
In the first test, a passage from Arthur Conan Doyle's 'The Hound of the Baskervilles' was handwritten and presented. While ChatGPT largely misinterpreted the text, Gemini performed nearly flawlessly. Gemini not only transcribed the text accurately but also analyzed handwriting characteristics like open loop formations in letters y, g, and f. This test clearly demonstrated Gemini's superior handwriting analysis capabilities compared to its competitor.
3Chess Position Analysis
Using a chess position photo from Esports World Cup, both models' move suggestions were compared. Gemini provided a lengthy, elaborate analysis while ChatGPT took over 4 minutes to respond, mentioning its attempt to locate a Stockfish engine. Interestingly, another AI named Claude reviewed both responses and concluded ChatGPT's answer was more appropriate, though without expert verification, this test's outcome remains inconclusive.
4Venus Flytrap Identification
A photo of a special Venus flytrap (towering giant variety) tested the models' ability to explain why some traps were turning black. Both chatbots correctly identified the primary reason (attempting to digest overly large insects). While ChatGPT offered more detailed care suggestions, Gemini failed to precisely name the cultivar. Due to its comprehensive response, ChatGPT won this particular round of testing.
5Movie Art Recognition Test
When presented with handmade paintings depicting the Sanderson sisters' mouths from 'Hocus Pocus,' Gemini easily identified the 1993 Disney film, characters, and even the actresses. ChatGPT could only describe the images as custom diamond-shaped artwork without recognizing the source material. This test highlighted Gemini's superior cultural reference recognition capabilities.
6Response Time and Detail Comparison
Throughout the tests, Gemini generally provided faster responses, while ChatGPT sometimes delivered more detailed analyses. However, Gemini showed more consistent performance in source recognition and visual comprehension. ChatGPT's technical details, like its attempt to access chess engines, revealed interesting aspects of its processing methodology that differed from Gemini's approach.
7Choosing Based on Use Cases
For handwriting analysis and visual recognition tasks, Gemini appears superior, while ChatGPT excels in scenarios requiring detailed explanations and care instructions. Gemini's extensive cultural database proves impressive for film and literature references. Users should select their AI model based on specific needs, as each demonstrates unique strengths in different multimodal applications.
8The Future of Multimodal AI
These tests demonstrate rapid advancements in AI's visual comprehension abilities. Gemini's source recognition skills and ChatGPT's analytical depth offer distinct advantages for various scenarios. As multimodal capabilities continue evolving, AI's role in our daily lives will likely expand significantly, with visual understanding becoming as crucial as text processing in AI applications across industries.
Original article: CNET — ChatGPT ve Gemini Görsel Algı Testinde Karşı Karşıya