Why use symbolic when you alredy have a closed form expressions for the differential equations? Usually, the Symbolic Math Toolbox is used to derive these expressions. But you don't need to derive them, so no need to use the symbolic stuff. Besides that, there were a couple of issues in the code related to the initial condition, the definition of odeFunc, and the call to ode45.
odeFunn needs to take two inputs, t and x, where x in this case is 4x1 vector, and return a 4x1 output of derivatives. The call to ode45 should not include the @(t,y), which is used to define the function, not to use at as function handle input argument to ode45.
Alternatively, it might be clearer to define odeFunc as an m-function, as opposed to an anonymous function as done here.
Anyway, check out the modified code below, which at least runs w/o error ann I think implments the intended equations.
t_span = linspace(0,20,600);
d2theta1 = @(x) (m2*L2*x(4)^2*sin(x(3)-x(1))-(m1+m2)*g*sin(x(1))+m2*cos(x(3)-x(1))*(L1*x(2)^2*sin(x(3)-x(1))+g*sin(x(3))))/(L1*(m1+m2)-m2*L1*(cos(x(3)-x(1))^2));
d2theta2 = @(x) (-cos(x(3)-x(1))*(m2*L2*x(4)^2*sin(x(3)-x(1))-(m1+m2)*g*sin(x(1)))+(m1+m2)*(-L1*x(2)^2*sin(x(3)-x(1))-g*sin(x(3))))/(L2*(m1+m2)-m2*L2*cos(x(3)-x(1))^2);
odeFunc = @(t,x) ([x(2) ; d2theta1(x) ; x(4) ; d2theta2(x)]);
[t,y] = ode45(odeFunc, t_span, x0);