They can be. I know J has libraries for databases, image, audio processing, etc. I've written an assembler in J and translated it to K. The array languages are pretty decent for prototyping machine learning stuff. Often you don't need any libraries at all.

Here's a 2-layer perceptron, first in Python+NumPy:

    import numpy as np
    X = np.array([[0,0,1], [0,1,1], [1,0,1], [1,1,1]])
    y = np.array([[0,1,1,0]]).T
    np.random.seed(1)
    w0 = 2*np.random.random((3,4)) - 1
    w1 = 2*np.random.random((4,1)) - 1
    for j in range(10000):
        l1 = 1/(1+np.exp(-(X @ w0)))
        l2 = 1/(1+np.exp(-(l1 @ w1)))
        l2_error = y - l2
        l2_delta = l2_error * l2 * (1 - l2)
        l1_error = l2_delta @ w1.T
        l1_delta = l1_error * l1 * (1 - l1)
        w0 += X.T @ l1_delta
        w1 += l1.T @ l2_delta
    print(np.round(l2,3))
    [[0.007]
     [0.991]
     [0.992]
     [0.01 ]]

And again in J:

    input =: 4 3 $ 0 0 1  0 1 1  1 0 1  1 1 1
    target =: 4 1 $ 0 1 1 0
    dot =: +/ .*
    sig =: {{ %>:^-y }}
    train =: {{
        'ignore_me w0 w1' =. y
        l1 =. sig input dot w0
        l2 =. sig l1 dot w1
        l2_error =. target - l2
        l2_delta =. l2_error * l2 * 1 - l2
        l1_error =. l2_delta dot |: w1
        l1_delta =. l1_error * l1 * 1 - l1
        w0 =. w0 + (|:input) dot l1_delta
        w1 =. w1 + (|:l1) dot l2_delta
        l2;w0;w1 }}
    5j3":0{::train^:10000 {{<:+:?.y$0}} each 1 ; 3 4 ; 4 1
    0.009
    0.990
    0.990
    0.011

Here's that assembler (for the Nand2Tetris assembly language) in K. It's intentionally golfed.

    L:({x^"\r "}'*'"/"\'0:0)^,""
    T:(("R",'$!16)!!16),(($`SCREEN`KBD`SP`LCL`ARG`THIS`THAT)!16384,24576,!5),((-1_1_)'L@&i)!{x-!#x}@&i:"("=*'L
    T,:n!16+!#n:?(({$[("@"=*x)&^`I$1_x;1_x;`]}'L)^!T)^`
    C:(" "\"D&A D+A A-D D !D D-1 -D D-A D|A D+1 0 A !A A-1 -A A+1 -1 1")!0 2 7 12 13 14 15 19 21 31 42 48 49 50 51 55 58 63
    D:$`" "`M`D`MD`A`AM`AD`AMD
    J:$`" "`GT`EQ`GE`LT`NE`LE`MP
    B:{,/$(x#2)\y}
    P:{c:*d:|"="\*j:";"\x;"111",/(B.7,(64*|/"M"=c)+C@"A"/"M"\c;B.3,D??d 1;B.3,J??1_j 1)}
    `0:({(P;{"0"^-16$,/$2\(`I$1_x)^T@1_x})[2/"@"=*x;x]}'L)@&~i