Today, Google Bard is getting a little smarter with the addition of math and logic capabilities. Google employee Jack Krawczyk announced the change on Twitter, saying “Bard will now better understand and respond to your prompts for multi-step verbal and math problems, with coding coming soon.”
Logic questions were a big drawback when Bard showed up dozens of days ago, and some of the answers made Bard seem especially stupid to early testers. In one example from last week, Bard repeatedly stated that one plus two equals four. Today, Google’s current AI chatbot models can now correctly say that the answer is three. So at least there have been some changes. He can also list the months of the year correctly instead of coming up with names like “Maruari”.
However, the Bard is still baffled by really simple logical questions. HowToGeek’s Chris Hoffman asked Bard the following day, “Which is heavier, five pounds of feathers or one-pound dumbbells?”Google Bard responded with the ridiculous statement that “there is no such thing as 5 pounds of feathers.”ChatGPT’s responses were no better, stating that five pounds of feathers and one-kilogram dumbbells “weigh as much as five pounds.”
With today’s update, Google Bard now gives the same incorrect answer as ChatGPT: “Five pounds of feathers and a one-pound dumbbell weigh the same.”This may be a common mistake with these types of language models (they all seem very bad with facts and figures), but it’s interesting given that Google was accused (and denied) of teaching Bard with the release of ChatGPT.
In addition to being a major gap in Bard’s capabilities, logic has also been artificially limited to not attempt to answer programming questions, so it’s good to hear from Krawczyk that those capabilities are coming soon. ChatGPT is known for being able to pump out tons of code in any language and style you like, and at times the code even works!
Krawczyk added: “We always balance new Bard features with efficiency. And this update is just one example of the many improvements we’re making to Bard every week.”
Weekly improvements would be great. Wall Street has squashed Google for taking a slow approach to its AI releases, but it still seems like the company is taking a slow approach with Bard. The first release is labeled “Experimental”, is not part of Google search, and is isolated on its own little site, bard.google.com. The service is also only available in the US and UK.
In an interview with The New York Times published today, Google CEO Sundar Pichai acknowledged that Google is still holding back on its best AI, saying, “We clearly have more powerful models. Pretty soon, maybe when this goes live, we’ll update Bard to some of our more functional PaLM models, so they provide more options, be it reasoning, coding. They can answer math questions better. So you’ll see progress over the next week.”