The news that the Google Bard chatbot and the underlying new PaLM 2 linguistic model are introducing on Web searches and on the entire ecosystem, both for private users and for professionals and companies.
During the Google I/O 2023the Mountain View company has renewed its willingness to draw heavily on solutions based on artificial intelligence also for what concerns searches on the Web.
Predictably, the chatbot Google Bard has become one of the central topics that the company led by Sundar Pichai addressed during the event reserved mainly for developers and supporters of Google technologies. The company first confirmed that Bard leverages the new large language model (LLM) PaLM 2 and noted that it had removed the “waiting list” for access to its features.
Point of reference for try Google Bard it is the home page of the chatbot which, however, unless you use a third-party VPN service, still does not allow you to access the service from Europe. Google spokespersons have in fact added that soon Bard will embrace support for 40 languages, Europen included.
PaLM 2 has been trained on huge volumes of multilingual texts: in this way, Google engineers could improve the ability to “understand”, generate and translate nuanced texts, including idioms, poems and riddles; also passes the exams of advanced language proficiency at the level of “mastery”.
The new language model unveiled by Google demonstrates improved capabilities in logiccommon sense reasoning, and mathematics as it was trained using a huge set of scientific documents and web pages that contain mathematical expressions.
Skills that have to do with warfare are also significantly improved over Bard’s preview code generation programming: PaLM 2 can count on an immense set of publicly available source code data. It can therefore excel in development with i programming languages popular ones like Python and JavaScript, but it can also generate specialized code in languages like Prolog, Fortran, and Verilog.
As the main competitor Microsoft Bing is already doing, Bard will soon allow create images generated by artificial intelligence, responding to specific requests made by users. The system will rest on Adobe FireflyAdobe’s family of generative AI templates: users will be able to quickly and easily transform creative ideas into high-quality images that can then be further modified and improved as well as inserted into projects managed with Adobe Express.
The images will also be used in the answers to the questions (prompt) of users in order to better argue what is offered by chatbot Bard. Google engineers have also shown a preview of a tool that will allow you to create documents of all types and for any need, with the possibility of exporting them directly to Gmail or Google Docs for final changes.
As early as next week, Bard will begin to cite sources precisely: when the chatbot inserts a code block or refers to content created and published by third parties, simply click on the side note to learn more and deepen your research.
The button Exportinstead, it allows developers to export the code generated by Bard and run it directly: the mechanism combines the skills of the popular cloud-based integrated development environment (IDE) al (short for Google Colaboratory) with the features made available by the partner Replit.
Replit is itself an IDE environment that allows users to create, share and collaborate on programming projects in different programming languages, such as Python, JavaScript, Ruby, C++, and many others. It is similar to Colab, but also offers features for building web applications, managing databases, and hosting applications. Replit was designed to make team programming easy, with real-time collaboration capabilities that allow multiple users to work on a project simultaneously, with built-in chat and the ability to comment on code.
During Google I/O, the company announced more 25 new products e functionality based on PaLM 2: this means that the language model is becoming the backbone of Google products bringing tangible benefits to people, including consumers, developers and companies of all sizes.
In recent days there was already talk of Sec-PaLM and the use of artificial intelligence in the field of information security. With a brief examination, Google talks about the use of the PaLM 2 model for productivity, therefore in Workspace, in the medical field and in the development segment (projects Vertex AI e Duet AI for Google Cloud).
Powering Search with Generative AI
Thanks to Bard, the PaLM 2 linguistic model therefore fully enters searches on the Web and in the search engine Google. Leveraging new features based ongenerative AIonline searches can become easier by enabling users to understand a topic more quickly, discover new points of view and insights.
Google managers give a concrete example: let’s take a question like “For a family with children under three and a dog, Bryce Canyon or Arches National Park is better?” Typically, you should break this question down into smaller ones, search through the vast amount of information available, and start putting the pieces together yourself. With Generative AI, the Google search can do a large part of this work, without any effort on the part of the user. The AI will generate an overview of the key information to consider, with a series of links to learn more.
With Generative AI in Search, Google can offer help in search too Online shopping making even the most delicate and complex purchasing decisions quicker and easier. When searching for a product, the web search provides an overview of important factors to consider and the list of objects and services that can meet the user’s expectations.
Google offers product descriptions with a variety of relevant and up-to-date reviews, ratings, prices and images. The new shopping experience with generative AI is indeed based on Shopping Graph by Google, which contains more than 35 billion product listings. A number that configures the Google solution as the one based on the most complete set of data in the world between products, vendors, brands, reviews and continuously updated inventories. More than 1.8 billion ballots are updated every hour in the Shopping Graphto offer up-to-date and reliable results.
Not to disappoint publishers and content creators alike, the new Google Search is designed to highlight and draw attention to content on the web, making it easier to dive into any topic outside of the search engine and Bard. “While integrating Generative AI into Search, we are committed to continuing to drive valuable traffic to websites“, assures Google.