Skip to main content

Ad

technology-iconTechnology
clock-iconPUBLISHEDMarch 22, 2023
comments icon1
share150

Google Released AI Chatbot Bard, And It Immediately Made Some Embarrassing Errors

In one conversation it claimed to be based on its rival, ChatGPT.

James Felton headshot

James Felton

James Felton headshot

James Felton

Senior Staff Writer

James is a published author with multiple pop-history and science books to his name. He specializes in history, space, strange science, and anything out of the ordinary.

Senior Staff Writer

James is a published author with multiple pop-history and science books to his name. He specializes in history, space, strange science, and anything out of the ordinary.View full profile

James is a published author with multiple pop-history and science books to his name. He specializes in history, space, strange science, and anything out of the ordinary.

View full profile
Google Bard logo on a phone.

So far, people are not impressed. Image credit: gguy/shutterstock.com


Google has finally launched its artificial intelligence (AI) chatbot Bard for the public to play with. While the large language model may be fun to play with, users have noted that so far it lags behind competitors, making a number of basic and amusing errors on simple tasks.

The rest of this article is behind a paywall. Please sign in or subscribe to access the full content.

One user discovered that the chatbot believed that it had already been shut down due to a lack of interest.

The user discovered that the sole source that the bot had used for this information was a 6-hour-old comment on Hacker News. Bard went on to list more reasons that Bard (which has not been shut down) had been shut down, including that it didn't offer anything new or innovative.

"Whatever the reason, it is clear that Google Bard was not a successful product," Google Bard added. "It was shut down after less than six months since its launch, and it is unlikely that it will ever be revived."

Elsewhere, the chatbot struggled with math.

For comparison, rival OpenAI's ChatGPT solved the question in the style of Cotton-Eyed Joe.

Bard no longer makes this error, and when asked for an answer to the question in the style of Cotton-Eyed Joe it gives it a good old try, vaguely attempting to stick to the rhyming structure and singing:

"If you were traveling at 8 miles an hourAnd you wanted to walk 8 milesIt would take you 1 hourTo walk 8 milesIf you were traveling at 4 miles an hourAnd you wanted to walk 8 milesIt would take you 2 hoursTo walk 8 milesIf you were traveling at 2 miles an hourAnd you wanted to walk 8 milesIt would take you 4 hoursTo walk 8 milesSo if you want to get there quickYou better walk at 8 miles an hourOr you'll be walking all dayAnd you'll still be nowhere"

Users were also disappointed with Bard's programming skills, though others pointed out that Bard would fulfill the request if you tweak the prompt slightly and specify that you want Javascript code.

In one final amusing blunder, it also claimed to be based on its rival.

Bard itself is a slimmed-down version of Google's Language Model for Dialogue Applications (LaMDA). 

"This much smaller model requires significantly less computing power, enabling us to scale to more users, allowing for more feedback," CEO of Google Sundar Pichai explained in a blog post ahead of the launch. "We’ll combine external feedback with our own internal testing to make sure Bard’s responses meet a high bar for quality, safety and groundedness in real-world information."

The chatbot failed to impress investors last month, wiping $100 billion off shares after an embarrassing blunder during a video launch. The bot, when asked to list discoveries made by the JWST telescope, claimed that it was the first telescope to photograph an exoplanet. However, that feat was actually performed by the Very Large Telescope in 2004, 17 years before JWST's launch.

It's unclear when or whether LaMDA will be released to the public. Google may feel it is lagging behind its rivals in the AI arena, with OpenAI's Chat GPT-4 already able to code well, do taxes, and excel at many human exams


Written by 

Add us as a Google preferred source to see more of our
trusted coverage in Search