# AI, Marketing & Content

This week, we’re diving into the latest in AI advancements that are redefining what's possible for marketers and creators.

?? AI learns to use computers

?? Motion capture for the masses

?? Image editing, levelled up

?? Self-learning AI from Meta

1. Claude 3.5 Sonnet Introduces “Computer Use” Feature in Public Beta

• Overview: Anthropic's Claude 3.5 Sonnet now has a beta feature allowing it to interact with computers like a human—navigating screens, clicking, and typing.
• Usage: Brands like Asana and Replit are testing this feature for complex, multi-step tasks within their applications.
• Limitations: While groundbreaking, the feature is still in its experimental phase and can be error-prone.

Takeaway: AI that can "use" computers might unlock a whole new level of automated workflows, but it’s still a bit like training a toddler to use a mouse—promising, yet chaotic.

2. Runway’s “Act-One” Brings Realistic Facial Animation to AI Video

• Feature: Runway's new Act-One tool transfers human facial expressions to AI-generated characters, eliminating traditional motion capture.
• Accessibility: Works with standard cameras (even smartphones), making high-quality animation more accessible to smaller creators.
• Safety: Includes safeguards to prevent misuse of public figures' faces and voices.

Takeaway: Act-One is a huge win for indie creators—think high-end facial animation, no mocap studio required. Just your face and a camera.

3. Ideogram’s “Infinite Canvas” Expands AI Image Editing Flexibility

• Tools: Ideogram’s Infinite Canvas lets users resize, merge, and edit AI-generated images with features like Magic Fill and Extend.
• Customization: Users can upload their own images for even more personalized content creation.
• Market Position: By offering proprietary models, Ideogram aims to set itself apart from open-source competitors like Stable Diffusion.

Takeaway: Infinite Canvas is like Photoshop for AI images—giving creators more control and fewer headaches when refining generated content.

4. Meta's Self-Taught AI Evaluator Learns on Its Own

• Tech: Meta’s “Self-Taught Evaluator” uses AI to improve its own responses without human oversight, breaking down complex tasks into manageable steps.
• Applications: Initially focused on fields like science and math, with broader implications for AI training processes.
• Open Source: Meta is sharing this model and other new tools with external developers despite potential risks.

Takeaway: Meta’s self-learning AI means we’re one step closer to machines that teach themselves—helpful, but maybe also a tiny bit sci-fi ominous.

Follow me at - https://www.dhirubhai.net/in/fergusdyersmith/ for more on AI in marketing.

Visit - wooshii.com for the AI products built for enterprise marketing teams and their agencies