Listen to the Podcast:
Google, a big company based in Mountain View, didn’t want to be left out of a Silicon Valley arms race, so it came out with its own chatbot AI, called Bard, to compete with GPT and Claude.
When trained on a lot of text, these big language models can come up with some pretty cool stuff. OpenAI’s GPT-4 has shown some amazing coding skills, most famously by turning a napkin sketch into a working website.
Specialized AI coding systems like Github’s Copilot (which is driven by OpenAI) and Replit’s Ghostwriter are already available to the public and show a lot of promise for making programmers more productive.
In recent updates, Google Bard has included a suite of coding capabilities that, in addition to offering explanations of what the code does, are aimed to assist in the management of a variety of software and program development chores.
In the AI wars, Microsoft-backed OpenAI and even smaller rivals like Anthropic have put Google in an unusual situation. As of last month, when TechCrunch asked Google Bard, GPT-4, and Claude to write some simple code head-to-head, Bard just said no. GPT-4 and Claude, on the other hand, wrote code that worked, even if it had some bugs.
Now, Google has released a set of coding features for Bard that are meant to help with different software and program development tasks and explain what the code does. This may be especially helpful for new programmers or programmers who are trying out a new language or application.
“Since we launched Bard, our experiment that lets you work together with generative AI, coding has been one of our users’ top requests,” Paige Bailey, group product manager at Google Research, wrote in a blog post announcing the new features.
From internal letters, Insider learned that Google’s own engineers had been testing the AI before it was released.
Bailey said that Bard can now code in more than 20 different computer languages. TechCrunch said that the computer polyglot can work with C++, Go, Java, Javascript, Typescript, and Python. It can also translate from one language to another. Also, the AI can now write functions for Google Sheets. When it uses open source code, it will also give credit to the source.
Bard can not only write code, but also help it work better.
“In many cases, Bard can also help you optimize your code by making it faster or more efficient,” Bailey said. “Just answer Bard’s first output with “Could you please make that code faster?” or “Find error handling sections you might have missed.”
AIs are not perfect. Anyone who has asked ChatGPT to do some easy math, quote a scientific study, or give them a jailbreak hack knows this. They “hallucinate,” which means that they just make things up. They also come to wrong conclusions with the faith of a high school quarterback in a soft conference.
Bailey said, “Bard is still a new experiment, and it may sometimes give wrong, misleading, or false information in a confident way.” “When it comes to coding, Bard may give you code that works but doesn’t do what you want it to do, or code that isn’t the best or isn’t complete.”
Bailey said that before adding the chatbot’s code, you should check it twice and look closely for mistakes or security holes. Bard can fix bugs in code, even its own. Just tell it something didn’t work and ask it to try to fix it, and the AI will do its best.
Bailey said, “Now that we have new coding skills, we’re excited to use generative AI to speed up software development, spur innovation, and help people solve hard engineering problems.”
