Google’s announcement about its new language model, Gemini, I unraveled the pace of curiosity and skepticism. Although the video promotion is an outstanding promotion, although it performs very well There are significant differences between the quantity shown and the actual capacity of the system. There is a very eye-catching font, your potential is lower in terms of explanation.

The real result of Gemini’s trailer: steaming hot and just arriving?

During Gemini’s latest rollout, Google went public with a video showing the alleged advanced capabilities of the AI’s new linguistic model. However, not all media have their first beat, so Bloombergone starts Verifying Prazioni’s authenticity.

In the video, it is seen that Gemini can instantly interpret a number of moves, including sasso, card and scissors play. But take turns to get correct answers from Gemini, and the rarest movements and movements required and provide specific guiding questions. Questo suggests that the effective process is more complex and less intuitive than initially represented.

I claim that the real capacity of neuronal nerves is “abbellito” Transparency and honesty at Google nel le proprie presents innovations. According to some Google employees who preferred to remain anonymous, the video demonstration did not fully reflect Gemini’s true capacity. This point of view is further confirmed Eli CollinIn Google DeepMind, there is a disegno dell’atra, as well as all the functional features, Sono is a live research and integrated with the product in a way that is not fully integrated.

Despite these statements, Google maintains that the video will be a simplified representation of Gemini’s capabilities for entertainment and clarity reasons. Description of the video on YouTube, more shortened presentation period according to the scope of research.

During a test with Gemini on the chatbot Google Bard, journalists also took notes the capacity of the new model IA does not differ significantly from that of GPT-3.5. The technology offered by Gemini compared to its current model needs to be questioned effectively.

Three varieties of Gemini

There are three versions of the neural network: Ultra, Prof. to nano. The first is designed to work in data centers and business applications; the second focuses on Bard and other services that Google and Nano have designed to work offline on Android. Last mission and announcement made Pixel 8 Pro. As an example of its use, Google showed a summary of information obtained from audio recordings.