Mistral launches Pixtral 12B, its first Multimodal model

Here is the compilation of the five biggest and juiciest news from the world of AI and technology during last 24 hours.

1)

Mistral launches Pixtral 12B, its first Multimodal model

French AI startup Mistral, which is known for its open-source AI models, on Wednesday unveiled a new AI model Pixtral 12B. The special thing about Pixtral 12B is that it is first-ever multimodal model from Mistral’s stable. For all those who are not aware, multimodal model has the capability of processing both images and texts. This model is not yet publicly available. However, you can access its source code on the Hugging Face and GitHub. Pixtral 12B will bring Mistral in line with its competitors like OpenAI and Anthropic, which already have models with text as well as image processing capabilities.

2)

Google will now embed Internet archive links to search results

In a welcome move, search giant Google on Wednesday announced a new feature that will incorporate the internet archive’s Wayback Machine within the search results. This feature will bring back the cached web page feature that was deactivated earlier this year. Users can access this new feature in the ‘about this page’ that can be accessed by clicking the three dots next to the search results. The ability to access the archived versions of the webpages directly from Google search interface is certainly a very handy feature. For all those who are not aware, the Internet Archive is a non-profit organization that preserves the snapshots of the websites through its wayback machine.

3)

Google’s AI note-taking app NotebookLM gets an audio overview feature

Google’s experimental AI tool NotebookLM, which helps in taking notes and research, has until today used text to summarize and explain source material. However, starting from today it can do this task by speaking loud. That’s because Google’s NotebookLM is getting a Audio overview feature. To activate this feature, you need to open an existing notebook, go to Notebook guide and click on the generate button on the right-hand side. Currently, it can generate response only in English language and would take roughly 3-to-5 minutes for generating response.

4)

Adobe will unveil its AI video creation tool later this year

Adobe on Wednesday announced that will launch its own generative AI video creation tool in limited release later this year.  This tool will be available in Photoshop’s existing line of Firefly image generating applications. Currently, It can generate video clip up to five minute long from a single prompt and can understand both text as well as image prompts. Adobe has also clarified that its AI video model hasn’t been trained by using its customer data but rather it has used licensed content. Although Adobe’s AI video tool won’t be available as a standalone tool, it will still help the company to compete with dozens of AI image generating platforms like Luma, Stability AI and OpenAI’s Sora.