Survey of Vulnerabilities in Large Language Models Revealed by Adversarial Attacks