Google just launched Bard, its chatbot of generative artificial intelligence, in the 27 countries of the European Union and Brazil. The tool is available in 40 new languages including Spanish, Portuguese, Chinese, Arabic and Hindi. Like the popular ChatGPT, it is free and provides multiple answer versions for the same question. In addition to his ability to compose texts, Bard can read aloud, similar to the feature that already exists in the popular Google translator, and his responses can be shared via Gmail or a public link. To interact with Bard, you need to have a Google account.
Despite having some improvements compared to ChatGPT, the OpenAI tool, Google’s intelligent chatbot also has problems: it makes mistakes when writing web links, it misidentifies images, and it occasionally provides biased or out-of-date information.
Incorporates new information, but context is lost
Unlike ChatGPT, whose database is limited to 2021, Bard includes recent information from the internet. For example, you have been able to explain that today (July 13) is the last day to vote by mail for the general elections in Spain. Below the answer there is an icon that directs to the Google results of the question asked.
But their answers are still opinionated and imprecise. When asked who is the most influential politician in Spain, Bard flatly answers that he is Pedro Sánchez. In one of his answers, he defines it as “a charismatic and popular politician who has managed to unite a divided country”. In another, he names paul married as the second most influential. And in another, he puts in the list to Albert Rivera Díazformer leader of Ciudadanos, who left politics a few years ago.
Lets share answers
A notable improvement over ChatGPT is the ability to listen to responses. The bot uses a male voice that sounds smoother and less artificial compared to the familiar Google Translate assistant. In addition, Bard allows you to export your content to other company products. By clicking the share button below each answer, you can send an email directly from Gmail or generate a document in Google Docs. It is also possible to generate a link with the answer and share it on social networks, for example.
Make mistakes when identifying photos
Google has announced its intention to integrate Google Lens, a tool that allows you to upload images to obtain information about them, such as identifying what they represent or writing a description. This feature is only available in English, but it doesn’t seem to work well at all. To use it, you need to generate a photo link from the Google Photos app and send it to the chatbot. When sending a photo of Casa Batlló, in Barcelona, Bard made a mistake: all three of his answers pointed to different places in the world, none of them indicated the iconic Catalan place.

confused links
When Bard was asked to provide a list of the most relevant organizations in cancer research, the chatbot presented a list with different names, almost all from the United States, despite claims that the answers are based on the geolocation of the user. However, by asking you for the web addresses of these institutions, Bard he was wrong once again, presenting repeated links for the different institutions. Chatbots often have difficulties with web addresses on the internet, but ChatGPT 3.5, the free version, has adequately answered the same question.
More concerned with privacy
Bard’s delay in reaching the European Union – the chatbot already worked in 180 other countries – is due in large part to the authorities’ concern about the protection of user data. When you first open the page, Bard presents the terms and conditions that include a list of “things to know.” First, that Bard uses the location and past conversations to “provide the best answer”; that the information may be inaccurate or inappropriate; that you should not be relied on for medical, legal, financial or professional advice; and that the user should not include confidential information. In addition, each time the chat is opened, there is a warning that “humans can process conversations” so “sensitive information” should not be included.
Users have the option to select the duration for which they want Bard to retain their data. By default, Google saves Bard activity for up to 18 months, but you can set it to three or 36 months, or turn this option off entirely.
The company has stated that it will soon be able to expand its availability to other languages in Spain, such as Catalan, Basque and Galician. According to Google’s director of product management, Jack Krawczyk, Bard already has the ability to respond in these languages, but “they need to train him to provide responsible responses.”
You can follow THE COUNTRY Technology in Facebook and Twitter or sign up here to receive our weekly newsletter.
Subscribe to continue reading
Read without limits