Specialized Data + LLMs Surpass LLMs on their Own
An Objective Comparison: Putting GPT-4, Claude 3 Opus, Lllama 3 70B, and others to the Test
For me, wellness means the balance of mind, body, and spirit. As an ultramarathoner, triathlete, surfer, and entrepreneur, this balance is core to my survival. My journey is not just about crossing the finish line; it’s about enhancing my overall well-being and leading by example.
Recently, I was on a flight with a notebook in hand, penciling out my wellness calendar in preparation for a full Ironman at the end of the year. As I sketched out my plan, it occurred to me that I might benefit from some scientific insights. I thought of the Huberman Lab Podcast, known for its deep dives into health and performance optimization. But then an even bigger idea struck me: why not pit GPT-4 and other large language models (LLMs) against each other to determine which provides the best advice?
And so, I decided to experiment with different AI models to see which could create the best weekly training schedule. To ensure objectivity, I provided the responses to ChatGPT and asked it to determine the most useful and comprehensive one and why.
The Challenge: Crafting the Ideal Training Schedule
I presented each AI model with this prompt:
“I want to achieve a perfect balance of mind, body, and spirit, optimizing for strength, mental fortitude, and performance. Please remember that I will be participating in a full Ironman triathlon at the end of the year. My activities include weight training, reading, cycling, swimming, running, yoga, and meditation. I can only do two activities per day. Can you create a schedule for each day of the week that helps me achieve my desired outcome?”
The Contenders: Four AI Models
I tested schedules from four different AI models to determine which provided the most comprehensive and useful plan:
- GPT-4
- Huberman Lab Podcast + Alani
- Llama 3 70B
- Claude 3 Opus
The Results: A Comparative Analysis
GPT-4:
- Morning: Weight Training (Monday), Cycling (Tuesday), Running (Wednesday), Swimming (Thursday), Weight Training (Friday), Long Bike Ride (Saturday), Rest Day (Sunday).
- Evening: Meditation (Monday, Thursday), Reading (Tuesday, Friday), Yoga (Wednesday, Saturday).
- Pros: Balanced inclusion of various activities, focus on physical and mental aspects.
- Cons: Lacked specificity in training techniques, less emphasis on scientific principles.
Huberman Lab + Alani:
- Morning: Weight Training (Monday, Friday), Cycling (Tuesday), Running (Wednesday, Saturday), Swimming (Thursday).
- Evening: Meditation (Monday, Thursday), Yoga (Tuesday, Friday), Reading (Wednesday, Saturday), Rest or Active Recovery (Sunday).
- Pros: Highly detailed, scientifically grounded training techniques, balance of physical, mental, and spiritual activities, emphasis on flexibility and recovery.
- Cons: More complex to follow due to detailed instructions.
Llama 3 70B:
- Morning: Weight Training (Monday, Thursday), Cycling (Tuesday, Friday), Swimming (Wednesday, Sunday), Running (Saturday).
- Afternoon: Meditation (Monday, Friday), Reading (Tuesday, Saturday), Yoga (Wednesday, Thursday, Sunday).
- Pros: Clear and straightforward schedule, balanced activities.
- Cons: Less detailed, lacks scientific backing in training recommendations.
Claude 3 Opus:
- Morning: Weight Training (Monday, Friday), Swimming (Tuesday, Sunday), Cycling (Wednesday), Running (Thursday, Saturday).
- Evening: Yoga (Monday, Thursday), Meditation (Tuesday, Friday), Reading (Wednesday, Saturday), Yoga or Rest Day (Sunday).
- Pros: Balanced approach, inclusion of recovery days.
- Cons: Less detailed and specific, similar to Llama 3 70B.
Why Huberman Lab + Alani Stood Out
After careful consideration, the schedule from Huberman Lab + Alani stood out as the most useful and comprehensive for my needs. Here’s why:
- Balanced Approach: It ensured a balanced mix of strength training, endurance exercises, flexibility, and mental fortitude, which aligned perfectly with my goal of holistic well-being.
- Specific Training Goals: The plan included detailed instructions for each activity, such as interval training for cycling and running, and technique-focused swimming sessions, essential for Ironman preparation.
- Mental and Spiritual Activities: The schedule addressed the mental fortitude needed for an Ironman by incorporating meditation and reading sessions focused on mental strategies and sports psychology.
- Flexibility and Recovery: Emphasized the importance of flexibility and recovery, suggested adjustments based on physical feedback, and included yoga and meditation for mental relaxation.
- Scientific Backing: References to scientific principles, such as the benefits of specific sets and repetitions in strength training and the importance of interval training, added credibility and depth to the training plan.
Here’s a snapshot of the Huberman Lab + Alani schedule:
Additional Tips
From the Huberman Lab + Alani model:
- Flexibility: Adjust workouts as needed to prevent overtraining.
- Nutrition and Hydration: Pay close attention to diet and hydration.
- Sleep: Prioritize sleep for recovery.
- Scientific Principles: Training suggestions are backed by scientific research.
The Power of Integrating Specific Data Sets with LLMs
TL;DR: To prepare for an Ironman triathlon, I tested four AI models to create a balanced training schedule. The Huberman Lab + Alani provided the most comprehensive and scientifically backed plan, effectively balancing physical, mental, and spiritual activities. This case study demonstrates how combining domain-specific data with LLMs leads to superior outcomes.
