AI roundup [April 23]: Microsoft launches compact AI model, Adobe to launch full AI image generation tool and more

Microsoft launches compact AI model for content creation; Adobe to launch full AI image generation in photoshop, and much more today.

Know what’s happening in the AI universe today, April 23. (Pixabay)

Microsoft launches compact AI model for content creation; Adobe to launch full AI image generation in photoshop; India leads global GenAI adoption, investment surge expected: Report; Scientists develop 'Toxic AI' to train safe chatbots- this and more in our daily roundup. Let us take a look.

1. Microsoft launches compact AI model for content creation

Microsoft unveils Phi-3-mini, a compact AI model designed for content creation and social media posts. Despite its smaller size, it surpasses larger models in language, coding, and maths tasks. Tailored for companies with limited resources, Phi-3-mini aids in summarising documents and extracting insights. Available on Azure, Hugging Face, and Ollama platforms, it expands accessibility to AI solutions, according to a report by Reuters.

2. Adobe to launch full AI image generation in photoshop

Adobe plans to introduce full AI image generation in Photoshop this year. Its upcoming tool, Firefly Image 3, will utilise AI to create images from scratch, leveraging user-uploaded images as references. This move aims to streamline the workflow for creative professionals amidst rising competition from startups like OpenAI, Midjourney, and Stability AI, according to a report by Reuters. 

3. India leads global GenAI adoption, investment surge expected: Report

India leads globally in GenAI adoption, with 81 percent of organisations already implementing the technology, according to a report by Elastic. Surveying 1,200 in the Asia Pacific and Japan region, including 300 from India, the report indicates a significant expectation of increased investment in GenAI over the next 2-3 years, despite concerns about data processing and usage, PTI reported. 

4. Scientists develop 'Toxic AI' to train safe chatbots

Scientists have developed a controversial AI training method called curiosity-driven red teaming (CRT), where an AI generates harmful prompts to filter out toxic content in AI chatbots. This approach aims to prevent dangerous and discriminatory responses. The research, uploaded to arXiv, suggests CRT could revolutionise AI training to ensure safe user interactions, amidst the proliferation of large language models, according to a Live Science report.

5. Government prepares AI framework to combat misuse

Union Minister Ashwini Vaishnaw announces plans for a robust AI framework to combat synthetic content misuse during elections. Public consultations will commence post-election, aiming to balance AI's public good and prevent misuse. The draft framework addresses deepfake concerns, focusing on economic growth while setting protective measures. Internet platforms must ensure compliance with IT Rules, with AI-specific advisories for accountability, according to a report by IANS.