AI is a powerful tool. It can help us in many ways. In this article, I will compare Meta.AI, Claude 3 Opus and Chat GPT4 in the term of keeping context.

AI Comparison
AI Comparison

First prompt - Bootstrap DB

I like to write the code. AI can do this too. The main difference is that AI can lose the context immediately. Let’s take a simple prompt:

1

prompt > create mysql database using docker with demo.sql file

I’ll remove all that AI explanation crap and here is the main results:

Claude: Dockerfile:

1
2
3
4
5
6
7
8

FROM mysql:latest

# Set environment variables
ENV MYSQL_ROOT_PASSWORD=root
ENV MYSQL_DATABASE=mydb

# Copy the SQL file to the container
COPY demo.sql /docker-entrypoint-initdb.d/

demo.sql:

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11

-- Create a table
CREATE TABLE users (
  id INT PRIMARY KEY,
  name VARCHAR(50),
  email VARCHAR(50)
);

-- Insert sample data
INSERT INTO users (id, name, email) VALUES
  (1, 'John Doe', 'john@example.com'),
  (2, 'Jane Smith', 'jane@example.com');

And a bunch of instructions:

1
2
3

docker build -t mysql-demo .
docker run --name mysql-container -d -p 3306:3306 mysql-demo
docker exec -it mysql-container mysql -uroot -proot

Emm, who will manually would like to run this?

1
2
3

USE mydb;
SHOW TABLES;
SELECT * FROM users;

ChatGPT: Pretty similar to Claude

1
2
3

FROM mysql:5.7
ENV MYSQL_DATABASE demo
COPY ./demo.sql /docker-entrypoint-initdb.d/

version 5.7 - WAT?

And ChatGPT provided docker-compose file:

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13

version: '3'
services:
  mysql:
    build: .
    environment:
      MYSQL_ROOT_PASSWORD: rootpassword
    ports:
      - "3306:3306"
    volumes:
      - mysql_data:/var/lib/mysql

volumes:
  mysql_data:

And here is an instructions on how to run them:

1
2

docker-compose build
docker-compose up

MetaAI: MetaAI decided to go another way. It provides just a bunch of instructions.

1
2
3
4

docker pull mysql/mysql-server:8.0
docker run --name mysql-demo -p 3306:3306 -e MYSQL_ROOT_PASSWORD=<your_password> -d mysql/mysql-server:8.0
docker cp demo.sql mysql-demo:/tmp/
docker exec -it mysql-demo mysql -uroot -p<your_password>

run pretty the same sql queries as Claude.

1
2
3

mysql> CREATE DATABASE demo;
mysql> USE demo;
mysql> SOURCE /tmp/demo.sql;

Firs step overview

Every AI provided different approach with the same instruction. I do not like Meta, because it provides just a bunch of instructions. Claude used mysql:latest which is kind better than ChatGPT mysql:5.7.

Second prompt - use docker compose

I like to use docker-compose. It is easier to maintain and run.

1

prompt > use docker compose and store mysql data in ./data folder

Claude:

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12

version: '3'
services:
  db:
    image: mysql:latest
    environment:
      MYSQL_ROOT_PASSWORD: root
      MYSQL_DATABASE: mydb
    volumes:
      - ./data:/var/lib/mysql
      - ./demo.sql:/docker-entrypoint-initdb.d/demo.sql
    ports:
      - "3306:3306"

ChatGPT:

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11

version: '3.1'  # Using version 3.1 for compatibility with volume mapping

services:
  mysql:
    build: .
    environment:
      MYSQL_ROOT_PASSWORD: rootpassword  # Change the password as needed
    ports:
      - "3306:3306"
    volumes:
      - ./data:/var/lib/mysql  # Maps the local 'data' directory to the container's MySQL storage

MetaAI:

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12

version: '3'
services:
  mysql:
    image: mysql/mysql-server:8.0
    restart: always
    environment:
      MYSQL_ROOT_PASSWORD: <your_password>
      MYSQL_DATABASE: demo
    volumes:
      - ./data:/var/lib/mysql
    ports:
      - 3306:3306

Conclusion

The best AI for this task is Claude. It provides the most accurate result understanding my end goal.