Also Google takes the field with Bard, an artificial intelligence chatbot, to counter Microsoft and OpenAi. In fact, CEO Sundar Pichai himself announced it on February 6, 2023, in the official blog of the Mountain View company. In the long post, Pichai explains that Bard is currently on an experimental basis while it is available for “trusted testers”. In any case, he promises support soon also for developers with specific APIs, as done by ChatGTP.
What is Google Bard and how was it born?
Sundar Pichai is very enthusiastic about the long way artificial intelligence has come so far. In fact, he cites advances in supporting the work of doctors, or in allowing people to access information in their own language. (Perhaps a possible reference to what was also done by big G, with google translate?)
And perhaps the reference instead to what the company has done in the last six years is not a coincidence; talking about what we can see today in generative AIs. From his study Attention Is All You Need, passing through Google AI and DeepMind, the stated goal is to help people, transforming searches into concrete products.
Google Bard starts with and is based on LaMDA, Language Model for Dialogue Applications, in English. AI is the latest step in a journey that began two years ago, an experimental conversational artificial intelligence service. If this reminds you of ChatGTP, it’s because, in practice, this AI also does something similar. But, as Pichai assures, Bard employs something more.
Indeed, “seeks to combine the breadth of world knowledge with the power, intelligence and creativity of our large language models“. Again as reported in the blog. Before making it available to everyone in the near future, however, there will be a test phase, as we anticipated in the intro.
The history of Google in the artificial intelligence of human language
The forerunner of the models on which Bard is based is the one proposed under the name BERT. According to Google itself, BERT represented a revolution in the understanding of complex human language. His successor, MUM, now two years old, is a thousand times more powerful and supports more languages; even able to identify key moments in videos and provide crucial information.
What Big G offers today, i.e. AI services such as LaMDA, PaLM, Imagen and MusicLM, have roots precisely in these aforementioned systems. New technologies that, from research, have entered the world of information, creating completely new ways of interacting with it.
How Does Google Bard Work?
Sundar says Bard can be an outlet for creativity and a springboard for curiosity. And he gives a simple example: He can help explain the new discoveries of NASA’s James Webb Space Telescope to a 9-year-old boy. Or to learn more about the best forwards in football right now, and to do exercises to improve our skills in a particular field.
In practice, the system draws information from the web to provide answers that are always new, updated and of high quality. But that’s not all that can be done thanks to Google’s new artificial intelligence. Among many, one of the most interesting opportunities is the way it can deepen the understanding of information and transform it into useful knowledge.
What Explained means: from simple question-and-answer requests provided by the traditional Google search engine, we move on to something more “semantic”. The example that the page makes of the piano is clear in this sense. We usually ask how many frets this instrument has; where doubts afflict us such as, “is it easier to play guitar or piano?”.
To this, which might seem like a difficult question for an AI system, Google Bard answers almost as if it were a human. In fact, he could tell us that some find it more difficult to do one thing rather than the other, and he would provide different opinions and insights. Among which, moreover, the different options taken from the various blogs on the net, or even the steps to start playing as a beginner.
The API question
With a view to greater use and participation in these advances, Goolge will start involving individual developers, creators and companies next month. In order to test the Generative Language API, initially supported by LaMDAthen with a series of models.
But developers also know how computationally expensive it is to build reliable AI applications. Which, for example, many startups without high computing power need. Google has therefore also thought of them, through the Google Cloud partnerships with Cohere, C3.ai and Anthropic.
The direction that Big G wants to take
In the press release, THE CEO proclaims it as clearly as possible:
It is vital to bring experiences rooted in these patterns into the world in a bold and responsible way. Here’s why we’re committed to developing AI responsibly: In 2018, Google was one of the first companies to publish a set of AI principles. We continue to provide training and resources to our researchers, collaborate with governments and external organizations to develop standards and best practices, and work with communities and experts to make AI safe and useful
Sundar Pichai, CEO Google and Alphabet
While on the ethical side Goolge therefore puts the word into it, on the purely technical side it tends to limp a bit. Indeed, recently, many have noticed the error in the promotional video shown. The bad impression cost the company a lot of money, passing under the judgment of many experts and astronomy enthusiasts.
In particular, this is the wrong answer given by Bard, on the James Webb Space Telescope mentioned above. In this regard, the AI allegedly said Webb responsible for the first photos in history of a planet located outside the solar system, or exoplanet. When instead these images are to be attributed to other less recent equipment.
However, Google responds by saying that the problem underlines the need to continue with new updates to the Bard system. After all, it’s in the hands of testers for a reason.
