The long-awaited Bard update brings Google’s chatbot closer to ChatGPT. Here’s how it works and what you need to know: A few months after the launch of Bard, Google is supporting its strategy in the race to find the best generative artificial intelligence with Gemini, a new AI model capable of working with texts, images, and videos.
Suppose ChatGPT paved the way in the sector by indicating the way forward. In that case, Gemini wants to be the arrival point of this technology to bring Big G to leadership in the industry – also thanks to the problems that the OpenAI management is running into.
The first version of the new algorithm will be the new operational brain of Bard, and a more compact version of it will also be installed inside Google Pixel 8 and 8 Pro to help users reply to messages; then, gradually, it will be integrated into all the products of the Mountain View company by the end of 2024 when the arrival of Gemini Ultra is expected, the super-powerful version capable of beating even the GPT-4 model which had surprised everyone this year.
The absolute novelty of Gemini is that it is “natively multimodal,” as Demis Hassabis, CEO of Google DeepMind, told Wired. What does it mean? If, until now, we proceeded with a separate approach to the different modalities (therefore texts, videos, audio, and so on) and then connected them at a later time, Google’s AI would be able to do this right from the start, elevating its class-leading capabilities in almost every area.
It seems trivial, but the ability to “digest” everything simultaneously allows the algorithm to pick up nuances and interpretations in language that have never been seen in artificial intelligence until now. It is, therefore, no coincidence that it has obtained results comparable to those of humans in a series of tests and benchmarks that measure reasoning abilities.
What Is Gemini For?
Google’s goal is to transform it into the next universal search engine that will not only return links or video responses – as Google Search has done so far – but precisely the information we need. So, to give an example, when searching for the best pizzeria in Milan, you will no longer find a simple site that has been able to place itself better on that search keyword, but the place that is right for you will appear based on what Big G knows about it.
You, what is written online, the reviews, the videos, and everything Gemini can develop. If this is an absurd and banal example, we can hint at its potential by applying it on a large scale to medical research, the scientific method, or the world of finance.
What Are The Risks We Run With Gemini?
In a context like this, AI would risk making the future unpredictable from every point of view. From job losses to the out-of-control flood of fake news and propaganda, the main risk is “losing control of our civilization”: at least these are the words of the letter published by the non-profit organization Future of Life Institute, which launched a petition to ask for a six-month pause in the development of systems more powerful than the fourth generation of chatGPT such as Gemini.
Google has conducted extensive security tests to understand how far Gemini can go, pushing it to behave incorrectly and expose its vulnerabilities. Timely and precise control will be needed to avoid problems. At a regulatory level, the European Union is at the forefront, thanks to the approval of the AI Act. This ongoing regulation will establish the rights and duties of those who use or develop artificial intelligence solutions.
It will take two years to draw it up concretely. Still, until then, some critical limits have already been established: the ban on the use of AI for biometric recognition and the classification of individuals based on sensitive data, as well as for recognizing emotions or evaluating people based on personal characteristics or political, religious, and sexual orientations. The goal is to protect citizens’ freedom and privacy and prevent mass surveillance systems.
Read Also: What Challenges For Cybersecurity In 2024?