As someone fascinated by the rapid advancements in AI technology, I’ve been eager to explore how these tools can offer insights and support personal growth. Over the past ten years, I’ve engaged in various forms of exercise, including hiking, circuit training, pilates, yoga, weight lifting, and walking. I’ve gained and lost 25 to 30 pounds along the way, completing my first significant diet in 2019 and then changing my eating habits again in fall 2023.

Testing image analysis to assess fitness changes

To investigate the capabilities of GenerativeAI in analyzing physical transformations, I decided to upload photos of myself at different weights and stages of my fitness journey to two AI models: ChatGPT-4 and Claude Opus. ChatGPT-4 is a widely-known language model developed by OpenAI, while Claude Opus is a more recently developed AI assistant created by Anthropic.

I began by asking ChatGPT-4 to analyze the photos from the perspective of a fitness expert and comment on the level of muscularity and fitness reflected in each image. However, ChatGPT-4’s initial response overlooked the dates on the photos. It provided a generic analysis as if they were all taken at the same time. Even after I pointed out the error and asked to compare specific images from 2015 and 2024, the revised comments remained relatively broad.

ChaptGPT-4 didn’t do such a good job; Claude did

In contrast, when I posed the same question to Claude Opus, the AI prefaced its response by acknowledging the sensitivity of identifying individuals in images. It then offered a general analysis without naming or identifying me specifically. The analysis was a detailed and precise chronological study of my body composition and muscularity changes across the photos.

Claude made detailed comments such as “In the 2015 photo, the person has a slimmer overall build with less visible muscle mass, especially in the upper body,” and “In May 2024,  at a higher body weight, the person has added a substantial amount of muscle mass while still appearing relatively lean.”

I asked Claude, how can you see all this?  The AI responded, “As an AI assistant, I’m trained to analyze images in detail and pick up on subtle differences that may not be immediately apparent to the human eye.  Let me break down a few specific elements that led me to my comparison (I am sharing just 2 of the examples):

  1. Shoulder width: The shoulders in the 2024 photo appear broader and more muscular, creating a more pronounced v-taper from the shoulders to the waist. In the 2015 photo, the shoulders look narrower in comparison.
  2. Facial features: Subtle changes in facial structure can occur with significant body composition changes. In the 2024 photo, the face appears slightly leaner and more angular, hinting at a lower body fat percentage than in the 2015 photo. “

The AI’s analysis was accurate. It identified subtle shifts I didn’t see, like changes in posture, body shape, and facial composition. 

I  did not realize an AI tool could do this analysis until I tried this experiment. As someone working on their fitness, it was a pleasure to generate and read a study that showed that my work to lose fat and build muscle is reflected in my body recomposition.

As someone fascinated by the rapid advancements in AI technology, I’ve been eager to explore how these tools can offer insights and support personal growth. Over the past ten years, I’ve engaged in various forms of exercise, including hiking, circuit training, pilates, yoga, weight lifting, and walking. I’ve gained and lost 25 to 30 pounds along the way, completing my first significant diet in 2019 and then changing my eating habits again in fall 2023.

Testing image analysis to assess fitness changes

To investigate the capabilities of GenerativeAI in analyzing physical transformations, I decided to upload photos of myself at different weights and stages of my fitness journey to two AI models: ChatGPT-4 and Claude Opus. ChatGPT-4 is a widely-known language model developed by OpenAI, while Claude Opus is a more recently developed AI assistant created by Anthropic.

I began by asking ChatGPT-4 to analyze the photos from the perspective of a fitness expert and comment on the level of muscularity and fitness reflected in each image. However, ChatGPT-4’s initial response overlooked the dates on the photos. It provided a generic analysis as if they were all taken at the same time. Even after I pointed out the error and asked to compare specific images from 2015 and 2024, the revised comments remained relatively broad.

ChaptGPT-4 didn’t do such a good job; Claude did

In contrast, when I posed the same question to Claude Opus, the AI prefaced its response by acknowledging the sensitivity of identifying individuals in images. It then offered a general analysis without naming or identifying me specifically. The analysis was a detailed and precise chronological study of my body composition and muscularity changes across the photos.

Claude made detailed comments such as “In the 2015 photo, the person has a slimmer overall build with less visible muscle mass, especially in the upper body,” and “In May 2024,  at a higher body weight, the person has added a substantial amount of muscle mass while still appearing relatively lean.”

I asked Claude, how can you see all this?  The AI responded, “As an AI assistant, I’m trained to analyze images in detail and pick up on subtle differences that may not be immediately apparent to the human eye.  Let me break down a few specific elements that led me to my comparison (I am sharing just 2 of the examples):

  1. Shoulder width: The shoulders in the 2024 photo appear broader and more muscular, creating a more pronounced v-taper from the shoulders to the waist. In the 2015 photo, the shoulders look narrower in comparison.
  2. Facial features: Subtle changes in facial structure can occur with significant body composition changes. In the 2024 photo, the face appears slightly leaner and more angular, hinting at a lower body fat percentage than in the 2015 photo. “

The AI’s analysis was accurate. It identified subtle shifts I didn’t see, like changes in posture, body shape, and facial composition. 

I  did not realize an AI tool could do this analysis until I tried this experiment. As someone working on their fitness, it was a pleasure to generate and read a study that showed that my work to lose fat and build muscle is reflected in my body recomposition.