Google Unveils Gemma: A New Open-Source AI Model to Rival Competitors

Google announces new open AI model Gemma 2 2B; Aims to beat other generative AI models

Google claims that the Gemma 2 2B has outperformed its big competitors including ChatGPT 3.5 Google announced its new AI model Gemma 2 2B, a new open-source AI model under the Gemma 2 series. Google says that Gemma 2 2B is a compact yet more efficient language model. The model is designed to perform tasks such as text generation, translation, and answering questions. According to an official blog Google mentioned that it has released a trio of new, "open" generative AI models. It further claimed that these generative AI models are "safer," "smaller" and "more transparent" than others. Google introduces new AI model According to sources Google's Gemma series of models can be different from its Gemini models. But how is Gemma different? Google describes it in a blog post, "Gemma Scope is made up of] specialized neural networks that help us unpack the dense, complex information processed by Gemma 2, expanding it into a form that's easier to analyze and understand." The company further adds that by studying these expanded views, researchers can gain valuable insights into how Gemma 2 identifies patterns, processes information and ultimately makes predictions. Google also plans to foster goodwill within the developer community with the Gemma series, just like Meta is attempting to do with Llama. Google has optimised the Gemma model for NVIDIA GPUs, making it accessible to a broader developer community. Additionally, the AI models are also compatible with popular AI frameworks such as Keras, JAX and Hugging Face, among others. Other updates The release of the new Gemma 2 models comes just after the U.S. Commerce Department endorsed open AI models in a preliminary report. The report stated that open models have the potential to broaden generative AI's availability to smaller companies, researchers, non-profits and individual developers. It further highlighted that there is a need for capabilities to monitor such models for potential risks. Google is continuously in the process of updating itself. Recently on Wednesday, Google announced a feature, called About This Image,in an official blog. It also said that it will work in collaboration with 'Circle to Search' for phones that support that feature and more broadly in 'Google Lens' for iOS and Android. Additionally, the search results could also include how other news outlets or websites have used the photo, the photo's metadata (information from inside a photo file that may include details about the file). Another update that Google came up with is allowing Chrome to offer a "Tab compare" function, using AI to create comparison tables for products across multiple open tabs. This feature aims to simplify the online shopping experience.

Google DeepMind releases Gemma 2 2B, its most compact, open-source AI model

Google DeepMind has released a new, compact, open-source AI model, Gemma 2 2B, on Thursday. Containing just 2.6 billion parameters, the language model surpasses rivals like OpenAI's GPT-3.5 and Mistral AI's Mixtral 8x7B, which are much bigger in size. The AI model is also deliberately nifty so as to fit on to a wider range of devices including smartphones while offering GPT-3.5's level of performance. LMSYS, an independent AI research organisation that tested Gemma 2 2B said that it achieved a score of 1130 in evaluation which placed it slightly ahead of GPT-3.5 Turbo-0613 and Mistral-8x7B. The latter models have more than ten times the parameters. Google I/O Connect | Updates on developer centric project; tie-up with MeitY Google's Developer Blog stated that Gemma 2 2B had been built using distillation techniques which effectively reduces computational requirements while distilling knowledge from larger models into smaller ones. (For top technology news of the day, subscribe to our tech newsletter Today's Cache) Earlier in June, Google DeepMind had announced the Gemma 2 9B and 27B models, but has moved to building even smaller and more efficient models as the market towards mobile and edge-based AI is expected to grow. Developers can access the model on the Hugging Face platform and can be implemented via PyTorch and TensorFlow. Read Comments