Imagine a text file with a million different numbers in random order. If you read the file from top to bottom it appears you’re reading a list of random numbers. Simply put this is how random numbers work in computers. When you ask the computer for a random number it’ll give you the next one in the list.
By default it’ll start reading from the top of the file but you can “seed” it by telling it which row to start on. It is almost always seeded with the current time, down to milliseconds, so even though it’s only reading numbers off a list, since that list is so big and you’ve told it to use the current millisecond (something that will be different every time you use it) as a starting point the result is random enough for everyday use.
On the other hand if you use seed values you know, you also know what sequence of random numbers you’ll get. This is often used for things like procedurally generated maps in games. Everyone who uses the same seed value will always get the exact same map generated.
I don’t know a good example of when this would be useful in particles but that’s what it does anyway.