Faced with the threat posed by ChatGPT and, above all, its association with Microsoft, Google is pulling out the heavy artillery with Bard, its conversational agent, just unveiled. The AI war is truly declared!
Google and Microsoft have been neck and neck in their war to rule the internet and the AI world ever since ChatGPT, the conversational AI developed by OpenAI, rocked the internet when it was released in December 2022. Quickly amazed by its unique, complex, precise answers formulated in natural language – which, moreover, gives teachers a lot of trouble – many saw it as a potential competitor to Google – opinion that does not is however not shared by the AI itself. Even if the tool is far from perfect – it often makes mistakes and spreads fake news with great conviction – Microsoft was quick to see the potential of the technology and has already already planned to integrate it with its Bing search engine as well as its Microsoft 365 office suite in order to attack Google on its own territory. No wonder the Mountain View firm, yet a pioneer in the field of AI, is really worried.
Even if it alone holds 90% of the sector’s market share and that, in front of it, the Redmond firm, yet number two, seems very tiny, Google must not waste time, because more ChatGPT is used by Internet users, the better it gets. The company has therefore launched a “red code” and reorganized various departments to move forward on its AI projects, which it is about to unveil to the public. Of the sources cited by CNBC also mentioned a few days ago a chatbot called Apprentice Bard and the integration of a new AI on the search engine that Google would currently have its employees test. The new announcements from Sundar Pichai, CEO of Google, tell us more while waiting for the conference that the firm will hold on February 8.
Google Bard: an AI based on LaMDA
In a press release published on his blog entitled “A major step in our progress in AI“, Google presents its new Bard chatbot, intended to compete with ChatGPT. It is based on the LaMDA language model (Language Model for Dialogue Application, or language model for dialog applications in French), already presented in 2021 and which is comparable to GPT-3 from Open AI – it became known in particular because one of the Google researchers thought that AI had a soul. Unlike many language models, LaMDA was trained on dialogue, not text. It is based on a neural network developed by Google which mimics – in a simplified way – the functioning of the human brain with its multiple connections between neurons and synapses.
“Bard’s ambition is to combine the breadth of knowledge of the world with the power, intelligence and creativity of our great language models. It draws on information from the web to provide up-to-date, high-quality answers.” explains Sundar Pichai, slightly tackling ChatGPT, whose knowledge stops in 2021. Because Bard should benefit from real-time access. Thus, it will be possible according to the firm to ask him to “explaining new discoveries from NASA’s James Webb Space Telescope to a 9-year-old“.
Google Bard: a chatbod soon available
Until now, Google had never dared to open AI to the general public, but the firm announces that it will be the case “in the coming weeks” in a lesser version, which “requires less computing power, allows us to deploy to more users, and process a larger volume of feedback.” In the meantime, the chatbot is reserved for “trust testers”. It also announces that its AIs will be integrated directly into Google Search and that “several AI-powered functions will synthesize information for questions for which there is no single correct answer”.
Obviously, we all wonder about the form that Google Bard will take. A box at the top of the search results? An application ? A function of Google Assistant? The Mountain View company offers some visual examples, with an interface that is very similar to that of the search engine. Thus, one can ask Bard “Between the piano and the guitar, which is the easiest to learn, and how long does each take to practice?” and get a clear answer in natural language.
Google recalls the crucial role of DeepMind – a subsidiary of Alphabet, the parent company of Google, specialized in artificial intelligence – in the field of AI because it is she who is at the origin of the neural network Transform, on which LaMSA is based but also… GPT-3, and therefore ChatGPT. However, the digital giant takes a much more cautious approach than OpenAI in order not to tarnish its reputation – because of its reputation, Google has no room for error. He must take many precautions in order to avoid the excesses of the AI, in particular the production of violent, offensive or false content, as well as political, philosophical or religious questions. Within the next month, developers will be able to use Google’s LaMDA model to integrate AI into their projects. An API is under development and access to it will be chargeable. In parallel, other online services managed by the company should benefit from the arrival of Bard in the coming weeks.
Google AI: see you on February 8 to discover the new tools
More news on Google’s intelligence advancements will be coming shortly. Indeed, the digital giant will not wait for its big Google I/O conference which is held in the spring to talk about its progress in artificial intelligence. He announced a surprise press conference to be held on February 8 from 2:30 p.m. French time – it will also be broadcast live on YouTube. An announcement that comes shortly after Sundar Pichai, the CEO of Google, claimed that the company would soon allow “to interact directly” with its “latest and most powerful language models as a research companion”. We obviously think of Sparrow, a chatbot quite similar to ChatGPT (see our article).
In description of the announcement video, Google explains that “We are reimagining the way people search, explore and interact with information, making it more natural and intuitive than ever to find what you need. Join us to discover how we are opening up better access to information for people people around the world, through search, Maps and beyond”. We can therefore expect improvements to Google Maps but also, given the thumbnail of the video, Google Lens and Google Shopping. We will have to wait until February 8 to see our curiosity satisfied!
But these announcements have visibly prompted Microsoft to respond. Indeed, the Redmond firm has sent invitations to the press for a conference directly in its HQ, in the United States. The event will take place on Tuesday, February 7 from 7 p.m., but without a live broadcast. The confirmed presence of Sam Altman, the president of OpenA, suggests the presentation of GPT-4. Hoping that the haste does not have a negative impact on the safeguards to be installed with such technology. Either way, war is well and truly declared!