I dont have access to the article, but I imagine 'high' is a relative term. In general, the density of primes arround n is roughly 1/log n; if you could do sign

I dont have access to the article, but I imagine 'high' is a relative term. In general, the density of primes arround n is roughly 1/log n; if you could do significantly better than that say 1/sqrt(log n) you could call that high accuracy. Meanwhile, the actual probability of predicting a prime would still be very low. Note that I am only speculating here.