NEWS IN BRIEF: AI/ML FRESH UPDATES

Get your daily dose of global tech news and stay ahead in the industry! Read more about AI trends and breakthroughs from around the world

Enhancing Visual Intelligence: Next-Token Prediction and Video Diffusion

MIT researchers propose Diffusion Forcing, a new training technique that combines next-token and full-sequence diffusion models for flexible, reliable sequence generation. This method enhances AI decision-making, improves video quality, and aids robots in completing tasks by predicting future steps with varying noise levels.

Climate Solutions: Use Them or Lose Them

Tech barons push for new technologies to solve climate crisis, but fixes already exist. Former Google CEO advocates for AI over climate goals, sparking debate on tech's role in addressing climate change.

Newsom's veto leaves Californians questioning AI safety

AI lobbyists resist regulation due to bipartisan voter distrust; Governor Newsom opposes SB1047 for excluding smaller AI models, advocating for comprehensive regulations instead. Journalist Garrison Lovely questions the decision, highlighting the need for proper oversight in the AI sector.

AI Co-Pilots: Revolutionizing Healthcare

Ambience Healthcare, founded by Mike Ng MBA ’16 and Nikhil Buduma ’17, offers an AI-powered platform embedded in EHRs to automate tasks for clinicians, saving 2-3 hours per day on documentation and improving patient care. Used in 40+ institutions like UCSF Health, Ambience helps clinicians focus on patients, not paperwork, leading to lower burnout and better relationships.

Streamlining Migration with Amazon Bedrock

Harnessing generative AI and Amazon Bedrock, organizations can simplify, accelerate, and scale cloud migration assessments. By utilizing Amazon Bedrock Agents and Knowledge Bases, a migration assistant application rapidly generates plans, R-dispositions, and cost estimates for AWS migrations, significantly speeding up the planning phase.

Secure AI Workflows with Amazon Verified Permissions

Amazon Bedrock Agents facilitate generative AI applications by orchestrating tasks, calling APIs, and applying fine-grained access controls. Verified Permissions integrates with agents to provide contextually aware access controls for secure application workflows.

Revolutionizing Vision Tasks with Florence-2

Florence-2 by Microsoft, a compact Vision-Language Model, excels in image annotation tasks with zero-shot capabilities. Pre-trained on FLD-5B, it supports tasks like captioning, object detection, segmentation, and OCR in a single model.