Bard is so stupid that I tried the take a picture feature, asked a question about a picture, then when I asked a followup question about the picture it said "Sorry I can't look at pictures."
I do expect it to get much smarter by the end of next year.
I asked bard to make a comparison chart of OpenAI, Claude, and PaLM's specs (size and context window) and it repeatedly gave me a nice looking table with obviously wrong values.
I do expect it to get much smarter by the end of next year.