Today, Google unveiled several new search capabilities and a demo of its AI chatbot, known as Bard, a day after Microsoft announced plans to inject AI into its Bing search engine and Edge web browser.
Google’s spate of AI announcements come after the tech giant reportedly issued a ‘code red’ to rapidly accelerate its AI strategy, following the viral success of rival ChatGPT and Microsoft’s US$10 billion investment in the product.
Bard is conceptually similar to ChatGPT, which, since its launch two months ago, has garnered over 100 million active users. Amateurs and experts alike are dazzled by its capabilities that keep extending, going from producing essays and poems to passing MBA exams.
But Google did take the credit for pioneering the technology that ChatGPT relies on and boasted of its ambitions for responsible AI.
“We made significant contributions to the scientific community, like developing the Transformer (Google’s machine learning model), which set the stage for much of the generative AI activity we see today, and we will continue to bring these technologies to the world in a responsible way that benefits everyone,” said Prabhakar Raghavan, senior vice president responsible for search, geo, ads, commerce products and more at Google.
Bard, released currently in a lightweight version to “trusted testers”, is powered by Google’s LaMDA large language model trained on dialogue to allow free-flowing, open-ended conversations. Google also plans to integrate LaMDA into its search results to generate narrative responses to queries that do not have one clear or right answer.
Microsoft’s Bing, which currently occupies only a small fraction of the search engine market, has similar plans to roll out an intelligent chatbot alongside Bing’s search results that summarizes and synthesizes web pages and different sources, as well as composing emails.
With investors flocking to AI stocks right now, Microsoft expects every percentage point of share it gains will bring in another US$2 billion in search advertising revenue, Reuters wrote.
“This technology is going to reshape pretty much every software category,” said Microsoft chief executive officer, Satya Nadella, to reporters during a briefing at the company’s headquarters in Redmond, Washington.
Microsoft’s new Bing search engine is live in limited preview on desktop computers, and will be available for mobile devices in coming weeks.
During the livestream event this morning, Google also sneaked in a few other AI-powered capabilities including:
- Addition of 33 new languages, including Corsican, Latin and Yiddish, to Google Translate’s offline mode
- New capabilities added to Google Lens to facilitate visual search across the web
- Immersive View AI that uses 2D images and information from Google Maps to generate a 3D representation of the entire context of a location. Rolling out today in London, Los Angeles, New York, San Francisco and Tokyo.
- APIs that will allow developers, creators and enterprises to build applications using its large language model, starting with LaMDA.
- Eco friendly routing in Europe, designed to help users choose the most energy-efficient route for electric vehicles, including suggesting the best charging stop, fast charging stations if the driver is in a rush, or supermarkets that have charging stations in case the user needs to run errands.
With all these announcements, the only limit to search, Google said, is our imagination. But there’s another limit: inaccuracy, and that did not take long to manifest.
This morning, on Twitter, Google published an advertisement for Bard in the form of a short GIF of it being prompted to respond to a query, to which it confidently delivered – a wrong answer.
The prompt was “What new discoveries from the James Webb Space Telescope (JWST) can I tell my 9-year old about?”, to which Bard responded with a number of answers, including the statement that JWST was used to photograph the very first pictures of a planet outside the Earth’s solar system (an exoplanet), which is grossly inaccurate. The first pictures of exoplanets were taken by the European Southern Observatory’s Very Large Telescope (VLT) in 2004, according to NASA.
The mistake was spotted hours after the livestream event, and the criticisms and derision inevitably flooded in. At the time of writing, the advertisement has been viewed over 1.2 million times.
Microsoft has also warned that its AI may still produce factually inaccurate information, known as a hallucination, but hopes that users’ feedback will gradually improve the technology.