Google Bard gives wrong answer during Google's presentation
Google's artificial search assistant Bard is of to a rocky start after Google's announcement of the feature earlier this week. Alphabet's stock, Google's parent company, dropped significantly hours after yesterday's Live From Paris event.
Bard is a language model that is optimized for dialogue. It is Google's answer to ChatGPT, which Microsoft favors. Unlike ChatGPT, which is already available for testing in the open, Bard is currently only in testing internally at Google.
Google and Alphabet CEO Sundar Pichai heralded Bard as a tool that combined "the breadth of the world’s knowledge" with the "with the power, intelligence and creativity" of the company's language models.
When Google demonstrated Bard's capabilities in an animated image during the feature presentation, it became clear quickly to some that one of Bard's responses to a demo query was incorrect.
Google asked Bard "what new discoveries from the James Webb Space Telescope can I tell my 9 year old about?". A good question, especially since it highlighted that Bard could give answers to more recent events, as the NASA telescope was made operational in December 2021. The public version of ChatGPT, which most users interested in the field are familiar with, is unaware of information and events after 2021.
Bard's answers included that the telescope spotted a number of galaxies nicknamed green peas in 2023 and captured images of galaxies that are over 13 billion years old. The third answer claimed that the James Webb Space Telescope captured the first pictures of a planet outside of mankind's solar system. The answer, unfortunately for Google, is incorrect.
Space enthusiasts were quick to point out that the Very Large Telescope in Chile was the first to capture a planet outside of the solar system in 2004.
Google, just like OpenAI and Microsoft, have disclaimers attached to their artificial intelligence helpers that inform users that answers are not infallible. Google, for example, states that "Bard may give inaccurate or inappropriate information" in Bard's interface.
Google did not expect Bard to provide an inaccurate answer during its presentation. The impact on the company's stock was significant, as it dropped by more than 7% during trading hours on Wednesday.
For now, it seems that Microsoft has the advantage in the race to integrate artificial intelligence into company products. Microsoft launched a preview of OpenAI's technology in Bing already and plans to expand it to more users in the coming weeks and months. The company has also announced plans to integrate AI into its web browser Microsoft Edge.
Google plans to roll out Bard to a limited number of public users in the coming weeks as well, but wants to test Bard internally first before it does so.
The company seems to have been caught on the wrong foot when Microsoft announced the investment in OpenAI and the integration of OpenAI's technologies into company products. Microsoft climbed more than 11% in the past seven days.
https://xkcd.com/1838/
It probably picked info worded like “first exoplanet discovered by jwst” meaning first one by jwst but not in general but it parsed it as first ever discovered by jwst and not by someone else.
And if that’s the case it’s not its fault as it was poorly written.
Yes, there is other info about earlier exoplanets and it could be combined to figure out real meaning but it is language model. All perceived “thinking” is not really there. It’s not made for that. It’s Google’s fault if they present it differently.
Imagine being Zuckerberg and betting the house on VR rather than AI
It’s called changing history, and given the trends in the western world “Bard” does it very well…
Whether Bard is at fault or not depends where it collect it data and whether the data collected are truth or fact (look up the difference yourself).