Try Gemma, Google's new artificial intelligence

Google’s favourite artificial intelligence is no longer Gemini, but Gemma. Gemma is a family of open-source language models created by Google based on the same technology used in the powerful Gemini models. These are lightweight and versatile models, designed to enable developers and researchers in other technical fields to experiment and create new artificial intelligence applications.

Although Google has been years ahead of its competitors in AI research, it is only now that it has launched the Gemma models, which stand out for being specifically designed for mobile phones.

Thus, Gemma 2, a model recently announced by Google, is designed to integrate seamlessly into compact smartphones without compromising on performance. A digital challenge worthy of a tech giant like Google!

What does Gemma 2 offer compared to the models from Meta and OpenAI?

Competition amongst the tech giants to develop more advanced language models is becoming increasingly fierce. Gemma 2, Meta’s Llama 3.1 and GPT-3.5 OpenAI’s three products are among the leading contenders in the race for artificial intelligence. However, when we compare the three, we see that Gemma 2 offers an undeniable competitive advantage:

Size and efficiency

One of the key features that sets Gemma 2 apart is its compact design; in other words, it It offers greater efficiency whilst taking up less space. Although it is smaller than its competitors, it achieves performance comparable to GPT-3.5, making it a more efficient option in terms of computing resources.

Accessibility

Both Gemma 2 and Llama 3.1 are designed to be more accessible than OpenAI’s models. Both offer open-source versions so that developers can customise and adapt the models to their specific needs. In contrast, OpenAI’s models tend to be more restricted and require access to the cloud.

Applications

All three models have a wide range of applications, from generating creative text to machine translation. However, Gemma 2 stands out for its focus on portability, designed for mobile apps and IoT devices. Llama 3.1 also stands out for its efficiency, whilst OpenAI’s models are better known for their ability to generate high-quality text.

Gemma family: types of applications

As we shall see, the Gemma models form part of a very large and varied family:

• Gemma 2B and Gemma 7B: These are the base models in the family. They differ mainly in their size (2 billion and 7 billion parameters respectively), which affects their capacity and complexity. They are highly flexible models that can be adapted to a wide variety of tasks, from text generation to translation.

• CodeGemma: As its name suggests, this model specialises in programming-related tasks. It can generate code, autocomplete lines of code and even explain programming concepts. It is a very useful tool for developers looking to automate certain tasks or learn new programming languages.

• RecurrentGemma: Like Gemma, this open-language model is designed to answer questions, summarise texts and reflect on a wide range of topics. However, it takes up less space in the computer’s memory, offers greater processing power and works faster.

• PaliGemma: It is a vision and language model capable of processing both images and text. It is suitable for a wide range of applications, such as image description, recording, answering questions in graphical format, facial recognition and object detection.

• ShieldGemma: It is designed as a filter that blocks harmful content such as hate speech or pornographic and violent material, to ensure that users’ interactions with the AI are safer and more respectful.

• GemmaScore: It is a model designed to evaluate and optimise the performance of other models in the Gemma family. This enables developers to continuously improve their AI systems.

How can I try out Gemma 2?

Google launched its new artificial intelligence model at the end of June, so It is now available to everyone. Whilst we can expect Google to release updates and improved versions of this new AI model, you can already get a taste of its full potential in the Google AI Studio.

Simply click on the link and select the model you want to use. In this case, you can choose between the Gemma 2 2B, the Gemma 2 9B and, the latest model, Gemma 2 27B. What are you waiting for? Why not give it a go?