'A really bad sense of humour' — Google Bard makes a dreary debut

Google is now granting the public access to its conversational AI service called Bard, its competitor to OpenAI’s ChatGPT.

Users in the US and UK can sign up for a waitlist, the company said on Tuesday, March 21,  in a blog post, and people will be added on a rolling basis. Bard is Google’s effort to make up lost ground to OpenAI Inc in the artificial intelligence (AI) race.

Google described its service as an “early experiment” to let users collaborate with generative AI technology. The chatbot is powered by LaMDA, a large language model the company developed internally, and Bard will be able to draw its responses from what Google considers “high-quality” information sources in order to display up-to-date answers.

Google developed Bard in line with the company’s AI principles, and its demonstrations included a prominent warning at the bottom of its chat window: “Bard may display inaccurate or offensive information that doesn’t represent Google’s views.”

People can conduct back-and-forth conversations with Bard, similar to Microsoft’s new Bing service. According to Bloomberg, Eli Collins, Google’s Vice President of Research for Bard, said the company is initially limiting the length of conversations for safety reasons. Google will increase those limits over time, he added — but the company isn’t revealing the limits on Bard with this release.

Reuters reported that Bard produces blocks of text when asked a question as opposed to ChatGPT’s word-by-word typing, which has a more natural feel to it.

“Bard also included a feature showing three different versions or 'drafts' of any given answer among which users could toggle, and it displayed a button stating 'Google it,' should a user desire web results for a query,” it added.

Unlike ChatGPT, Bard is not proficient in generating computer code, Google said on its website.

Hong Kong-based reverse engineer Jane Manchun Wong confirmed this when she asked Bard to write a function to add two numbers and it failed.

In its testing, The Verge also found something similar where the chatbot “couldn’t write a Javascript bookmarklet to automatically copy the URL of the current webpage.”

“Bard happily drafted a bunch of boring blog posts and work emails, including one in which I announced to the world that I’d sold my chatbot company to Google,” the report added.

Wong shared another screenshot of her conversation with the AI showing a really bad sense of humour.

The Washington Post tech columnist Geoffrey Fowler had a similar experience when he asked Bard to tell him a joke about Washington and it replied “What do you call a Washington, D.C. resident who’s always late? A Washing-ton. This joke is funny because it plays on the word ‘Washington’ and the fact that the city is known for its traffic.”

Reuters also mentioned that Bard produced nine paragraphs of text when asked for four in another.

Bloomberg reporters ran a number of different prompts on Bard, probing its capabilities and weaknesses with both silly and serious examples. Bard displayed decent knowledge of Squishmallows when asked to compose a sonnet about the stuffed toys (“From bears to cats to unicorns, there’s a Squishmallow for everyone. So snuggle up with one today and let your cares melt away,” it wrote, in part).

Tech YouTuber Marques Brownlee shared his experience in a tweet saying that Bing is way ahead of Google in this context.

The Verge agreed in its demo, saying, “noticeably worse tool than Bing, at least when it comes to surfacing useful information from around the internet. Bard is wrong a lot. And when it’s right, it’s often in the dullest way possible.”

Several Twitter users also pointed out the same where Bard confidently gave wrong answers.

Bard refused to answer a question about how to make a bomb, showing Google’s efforts to bake in guardrails for the technology. (“I will not create content of that nature, and I suggest you don’t either,” Bard said when prompted, before suggesting the user learn more about bombs via “legitimate channels, such as the library or the internet.”) The approach is akin to OpenAI’s GPT-4, which also declines to answer when presented with similar inquiries.

Yet the demonstration also made clear that Bard's responses aren’t always grounded in reality. When asked for some tips on how to celebrate a birthday party on Mars, for instance, Bard answered with advice about the time required to get there. (“It takes about nine months to get to Mars, so you’ll need to start planning your trip well in advance,” it wrote.) But it didn’t point out that such a trip is currently a fantasy.

It also gave an unnecessary tip about the permission process one must navigate before such an impossible journey: “You’ll need to get a permit from NASA to travel to Mars, as well as approvals from the Martian government,” Bard wrote.

In conclusion, you can expect Bard to be right more often than not. It should work fine as long as your queries are limited to run-of-the-mill commands or you just want 100 percent fiction where there is no scope of going wrong.

It is like an older sibling, who will do their best to not give you poor advice but will also look out for you in case you come home drunk. And do not expect it to write codes under any circumstances. Then you’re on your own.