ChatGPT, meet the competition
Google has invested heavily in AI-powered features like Magic Eraser, Photo Unblur, and Google Assistant for its Pixel range of phones. Those extensive R&D efforts touch all aspects of its business, like Search, Adsense, chatbots, and even just some fun demos. With Microsoft-backed ChatGPT’s sudden rise to fame in the last few months, Google is now on edge, and the industry has been curious to see how it will respond. A new report based on insider information, staff memos, and samples of development work shows the company scrambling to create a ChatGPT equivalent integrated with Search.
Google is working on a two-pronged response to ChatGPT’s popularity, CNBC reports. First off, the corporation is internally testing a chatbot called Apprentice Bard. Anonymous staffers say the implementation resembles ChatGPT — the bot provides detailed answers to user prompts and questions. However, instead of using the Microsoft-developed GPT-3.5 code underneath, Bard uses LaMDA, Google’s Language Model for Dialogue Application. You might remember that as the same tool that grabbed headlines a few months ago for getting people talking crazy about sentience. The company is reportedly testing this tool, and in recent weeks, Google has asked more employees to lend a hand as the AI improves.
At some point, Google plans to enhance Search with Bard’s association. Images leaked from the company’s testing processes show a redesigned interface in the works, incorporating Apprentice Bard responses that attempt to feel like there’s an actual person responding to your Search queries.
One of the key factors differentiating Bard from ChatGPT is the former’s ability to respond based on current data, staying up-to-date with news and trends. The bot cited data from as recently as January 2023 during internal testing. Meanwhile, ChatGPT engineers confess they halted the AI model training in 2021.
Google’s been pitting Bard against ChatGPT for some head-to-head comparisons, including asking both how they might end up replacing human jobs. Both tend to offer similar answers (they swear that programmers’ jobs are safe), but ChatGPT provides lengthier and more detailed responses when prompted to get creative, like with a movie script. However, the Microsoft-backed effort fails in spectacular fashion when answering logical riddles, unlike LaMDA.
Google’s reluctant to deploy an AI that’s factually inaccurate, and ChatGPT has been woefully wrong at times. At a recent meeting, Google’s AI chief Jeff Dean said the company has a bigger reputation to safeguard, and is therefore progressing “more conservatively than a small startup.”
That said, Dean expressed interest in sharing the company’s work publicly soon, but didn’t specify a timeline. Supposedly, the LaMDA team has instructions to place other projects on the back burner and focus on Bard’s development. Simultaneously, Google’s cloud division is also developing a “code red” response to ChatGPT called Atlas, details about which are rather scant. These pointers suggest Google is rather close to delivering on company CEO Sundar Pichai’s earlier tease of releasing a ChatGPT-like product sometime this year. Hopefully, we don’t need to wait much longer.
