Details on an AI extension for Google search were not expected until next Wednesday, but initial information about Microsoft Bing-ChatGPT has apparently made Google change plans: via blog post, CEO Sundar Pichai announces the AI extension “Bard” today.
Bard is based on the LaMDA language model
Bard is based on Google's own LaMDA language model, so it has nothing to do with the OpenAI Large Language Model (LLM) used by ChatGPT.
The development of the Language Model for Dialogue The group started Applications two years ago. Bard is the first “experimental AI service” based on this model. For the time being, like Bing-ChatGPT, it should be open to “trusted testers”, but it will be made available to a “broader public” for testing “in the coming weeks”. Google has not yet given any details.
According to Pichai, Bard is initially using “a simplified model version of LaMDA” that requires less computing power and can therefore be used by more people.
According to initial information, like Bing-ChatGPT, Bard will be able to use complex search queries or tasks based on current information on the Internet. As a first example, Google cites the search query “What recent discoveries from the James Webb Space Telescope can I tell my nine-year-old child?”, which Bard knows how to answer, taking into account the typical interests of a nine-year-old child.
Pichai promises that in the future Bard will be able to provide an overview of the search results found for a question from Google Search – not compiled in a comparatively simple manner as is currently the case, but summarized in a high-quality manner based on the AI.
When people think of Google, they often think that we can give them quick, factual answers, like “How many keys does a piano have?”. But more and more people are turning to Google when they need deeper insight and understanding, such as “Is the piano or guitar easier to learn and how much practice does it take?”. Finding out exactly what you really need to know is often a little tedious, and you often want to get to know different opinions and perspectives.
AI can help in those moments by consolidating insights for questions that don't have a single right answer. You'll soon see AI-powered features in search that translate complex information and multiple perspectives into easy-to-understand formats so you can quickly see the big picture and learn more from the web: whether you're looking for additional perspectives, how for example blogs from people who play both piano and guitar, or want to deepen a related topic such as steps to get started as a beginner. These new AI features will soon be integrated into Google Search.
Sundar Pichai via Bard
LaMDA via API
Google states that it does not want to use the LaMDA model alone in the future. As early as March, the group wants to start connecting individual developers, creators and companies “so that they can try out our API for generative language models, which is initially based on LaMDA and will later be expanded to include a number of models”. This announcement and the reference to the fact that the group is aware of its responsibility in the field of AI development should be aimed in particular at skeptics and critics of the latest advances in the field of artificial intelligence. Almost four years ago, Google set up an AI ethics council to deal with ethical issues in AI development – but it got off to a rocky start.
< p class="p text-width">Whether we use AI to radically transform our own products or make these powerful tools available to others, we will continue to be bold and responsible. And that's just the beginning – in the coming weeks and months we will do even more in all these areas.
Sundar Pichai on Google responsibility
With more details of the ChatGPT or Bing ChatGPT competitor is calculated on Wednesday, February 8th. Google will be streaming on YouTube that day.