Using chat GPT 3.5, I cracked the code to make my local LLM smarter. By analyzing user input and using Python code, I was able to enhance the response to questions. With a mix of local models and APIs, I even tackled complex queries like Bitcoin prices. And with anthropic CLA 3 Opus, I added a new feature in just minutes. This has been an epic journey of boosting intelligence! 🚀
Testing the Intelligence of LLM
Okay, so this is a custom test I used to check the intelligence of LLM. We are running it on chat GPT 3.5. The answer calculated by the system was 32 apples, but the correct answer is 36. By using the API, we were able to create a Python code that fixed this error and resulted in the accurate output of 36 apples.
Key Takeaways
| Test Outcome | Conclusion |
|---|---|
| Corrected Output | 36 apples |
| System Used | Chat GPT 3.5 with API |
| Results | Accurate |
Implementing CLA 3 Opus Features
The logic behind this system involves analyzing user input queries to determine if code is necessary to improve the answer. If yes, the system will write and execute the code, then use the context to enhance the response of the LLM.
Code Analysis Logic
| Function | Description |
|---|---|
| should use code | Determines when to use Python code |
| Code Output | Stored output from executed code |
| Natural Language Response | Provides helpful answers based on context |
Testing Different Models
We started by using a local model on the system, switching out all functions related to open AI for our local model. Then, we tested using an API to find out the price of Bitcoin today and the upcoming weekend weather forecast for London.
Price of Bitcoin Today
Price of Bitcoin: $72287
Upcoming Weekend Weather Forecast
Weather Forecast: Overcast clouds, temperature of 13° Fahrenheit, humidity of 81%
Experimenting with CLA 3 Opus for Coding
We attempted to use an API to find out the cost and worth of one Bitcoin one year ago and compare it with today’s value. After encountering errors, we switched models and successfully obtained accurate results.
Percentage Gain Calculation
- Cost of Bitcoin One Year Ago: $24,178
- Current Cost: $71,000
- Percentage Gain: 195%
Adding New Features with CLA 3 Opus
We demonstrated how easy it is to add new features to existing code using CLA 3 Opus. By simply providing instructions, we were able to prompt users for permission before executing code and generate alternative code options.
Updated Feature
Before executing code, users are prompted to confirm with "y" or "n". If "y" is pressed, the code will execute; if "n", a different code will be suggested.
In conclusion, CLA 3 Opus offers a seamless way to enhance the intelligence of LLM through advanced coding capabilities. It has proven to be efficient in handling various tasks and adding new features effortlessly. If you’re interested in accessing the code used in this demonstration, consider becoming a member for full access. Have a great day!
EN_GB