I’m going to answer these in reverse, as it’ll make more sense that way.

All the other values have a 1 as their left most bit.
Not exactly. This is true if we simply write out the values, but it is false if we look at each number in the context of it being a 4byte (32bit) value. For example:
i=1 1073741816 = 111111111111111111111111111000 (30 bits to store number)
If we view the entirety of the 32bit space this number can occupy, we see this:
Here we can see the entirety of the 32bit space, with the binary representation of your number in the bits highlighted in green. There are two leading zeroes at the start of the number, in yellow. The leftmost bit is a zero, hence the number is positive.
 Why does it become negative?
In Java, the int
type is signed. Everything to a computer comes down to bits  1
or 0
. There is no concept of a negative sign 
, so if the computer needs to be aware of negative values we need some way of telling it about them.
Java is seeing 32
as the following.
As the most significant (left most) bit is a 1
, the number is a negative. This is by convention; the computer does everything in 1
s and 0
s  there’s no concept of encoding a negative sign, so a bit is reserved to indicate this instead.
 Why does it become 32 in particular?
Java uses a method of representing negative numbers which is called “Two’s complement”. I’ll link to a fuller description of that at the end, but for now I’ll just cover the basics.
Let's change 32 to 32 by converting to Twos complement representation
00000000000000000000000000100000 # 32 in binary
11111111111111111111111111011111 # invert (flip) each bit.
11111111111111111111111111100000 # add one.
The value represented by 11111111111111111111111111100000
is essentially the same as taking (2^31)
 the leftmost bit  and then adding the positive value of each subsequent 1
, like so:
2147483648
+ 1073741824 + 536870912 + 268435456 + 134217728 + 67108864 + 33554432 + 16777216
+ 8388608 + 4194304 + 2097152 + 1048576 + 524288 + 262144 + 131072 + 65536 + 32768
+ 16384 + 8192 + 4096 + 2048 + 1024 + 512 + 256 + 128 + 64 + 32
= 32
You end up with 32
.
Hope that clears it up a bit.
Also, here’s the link for more info on Two’s complement representation.