Assuming we arrange people in a perfect grid, and each person stands on 0.66 square meters...
Assuming you are interested in the sound intensity at the origin in a two dimensional infinite flat grid and you have people whose individual power is $P_0$ standing at grid locations other than the origin, then the intensity at the origin is:
$$
I = \sum_{(i,j) \neq (0,0)} \frac{P_0}{4\pi a^2(i^2 + j^2)}\;,
$$
where $a^2$ is 0.66 square meters. (And where we have assumed that our two-dimensional grid of people is embedded in our usual three-dimensional flat space.)
Clearly $I$ increases as you increase the number of people. It increases "slowly" but increases nevertheless.
Whether or not the sum increases without limit depends on the dimensionality of the grid.
For a two-dimensional grid, you can look at the behavior far from the origin in the continuum limit. In this limit assume the intensity at the origin due to a patch of area $\delta A$ is proportional to $\frac{\delta A P_0}{a^2 r^2}$. Then:
$$
I \to I_{\text far} \sim \int_{\text far} \frac{\delta A P_0}{a^2 r^2}
$$
$$
\sim \int_{\text far} \frac{r dr d\theta P_0}{a^2 r^2}\;.
$$
In this two-dimensional case we have the "far" contribution as
$$
I_{\text far}\sim\frac{2\pi P_0}{a^2} \int_{r=R}^\infty \frac{dr}{r}\;,
$$
where $R$ is some large distance (large enough that we are fine with the continuum approximately). The integral diverges logarithmically due to the upper limit of integration being infinity. The integral diverges worse for a three-dimensional grid. The integral doesn't diverge for a one-dimensional grid (e.g., a line of people embedded in three-dimensional flat space).
Per the comments, OP now also wants to ask a similar question about coherent audio sources.
Note first, however, that a bunch of people screaming do not produce coherent audio. There is no clear mechanism by which a bunch of people screaming would coordinate the waves coming out of all their mouths.
But anyways. Let's assume the audio sources are can all be specified at an exact wavelength $\lambda$ and phase offset $\phi$. In this case the amplitude at the origin of the entire grid of coherent screamers is:
$$
A = \alpha_0\sum_{(i,j)\neq (0,0)}\frac{e^{i\left(\frac{2\pi a}{\lambda} \sqrt{i^2 + j^2}\right)+i\phi_{ij}}}{a\sqrt{i^2 + j^2}}
$$
and the intensity is:
$$
I = |A|^2 = |\alpha_0|^2
\sum_{(i,j)\neq (0,0)}
\sum_{(k,\ell)\neq (0,0)}
\frac{e^{i\left(\frac{2\pi a}{\lambda} (\sqrt{i^2 + j^2} - \sqrt{k^2+\ell^2}\right)}}{a^2\sqrt{i^2 + j^2}\sqrt{k^2+\ell^2}}
e^{i(\phi_{ij}-\phi_{kl})}\;.
$$
The average intensity (averaging over the site-specific phase) is:
$$
<I> = |\alpha_0|^2
\sum_{(i,j)\neq (0,0)}
\sum_{(k,\ell)\neq (0,0)}
\frac{e^{i\left(\frac{2\pi a}{\lambda} (\sqrt{i^2 + j^2} - \sqrt{k^2+\ell^2}\right)}}{a^2\sqrt{i^2 + j^2}\sqrt{k^2+\ell^2}}
<e^{i(\phi_{ij}-\phi_{kl})}>\;,
$$
where we recover the incoherent result when $<e^{i(\phi_{ij}-\phi_{kl})}>=\delta_{(i,j),(k,\ell)}$ and we see that $|\alpha_0|^2\sim P_0$.
OTOH, if the site-specific phases are the same (e.g., all $\phi=0$) then:
$$
I = |A|^2 =
{\left|
\alpha_0\sum_{(i,j)\neq (0,0)}\frac{e^{i\left(\frac{2\pi a}{\lambda} \sqrt{i^2 + j^2}\right)}}{a\sqrt{i^2 + j^2}}
\right|}^2
$$