Neither is correct.
Likelihood is a probability, and hence it is a proper fraction. So, its log-likelihood cannot possibly be positive, and hence it cannot be 0.12. I haven't seen the video, but almost certainly it is not the log likelihood, rather the density of the \mathbb{N}(32,2.5^2) at the point x=34.
The normalF function doesn't calculate the exact log-likelihood. While finding the maximum likelihood estimators of \mu and \sigma^2, you don't need the constant part involving \pi. Hence, it removes that part and returns only what is necessary. So, it's output is not the exact log-likelihood.
And, a short note is that the function is rather misleading, as though the comments say it expects parvec[2] to be standard deviation, it actually needs variance. That's why the -0.5 part is present and the denominator of the 2^{nd} term is not squared. If you want to verify this, do the derivation yourself, or see the last equation of the post.
If you want the exact likelihood, I'll say this is a far easier way:
get_normal_log_likelihood <- function(observations, mean_parameter, sd_parameter)
{
log(x = prod(dnorm(x = observations,
mean = mean_parameter,
sd = sd_parameter)))
}
get_normal_log_likelihood(observations = 34,
mean_parameter = 32,
sd_parameter = 2.5)
#> [1] -2.155229
Created on 2019-10-12 by the reprex package (v0.3.0)