How Good is Google Bard's Visual Understanding? An Empirical Study on Open Challenges