Conference Presentation, Panel
AI Unleashed : Navigating the Future with Next Gen Applications | RAISE Summit 2024 | Paris
- Animage aims to establish a franchise comparable to "Paw Patrol" by developing proprietary data sets and applications trained on niche requirements such as cartoon body proportions, while planning to wait several months for news on off-the-shelf capabilities before advancing proprietary models.
- X-Plane has transitioned from the demo phase to production with real clients, automating the public tender process to reduce response times from one to one-and-a-half months down to a couple of days and enabling clients to extract three times more information with five less time consumption.
- General Gen AI trends indicate a shift toward verticalized "agents" and integrated ecosystems rather than standalone generic tools, with major competitors like Microsoft, Google, and SAP focusing on embedding assistants into existing workflows, though current enterprise offerings remain underwhelming.
- Production viability for AI video and image generation faces significant barriers, including high costs (potentially reaching 10 euros per user daily), latency issues where 40-second delays hinder usability, and the need for improved motion controllability and output fidelity to match Hollywood standards.
- Animage plans to integrate off-the-shelf AI with proprietary text-to-motion and animatics-to-motion tools within traditional 3D environments like Unity or Maya to facilitate manual motion correction, while the industry anticipates the release of the first fully AI-generated movie in the coming months.
- L'Oréal is developing the "Beauty Genius" app to achieve ultra-personalization with a goal of limiting product recommendations to 5-6 items for a fluid shopping experience, with potential new features to be announced around Vivatech, pending management decisions.
- The animation sector is projected to realize 90% savings in production costs, shifting value from fabrication to creativity, though traditional studios face a difficult workforce transition as 80% of staff rely on methods that may not integrate with AI.
- Future AI applications must be designed to function despite daily model errors, requiring robust architectures that allow users to easily verify, source, and edit information rather than relying on perfect language models.
- Public procurement represents 12% of global GDP and is targeted for simplification, with the long-term vision being the quality improvement of all public tenders rather than merely increasing volume, though current high costs for top-tier models like GPT-4 necessitate the use of lower-quality alternatives in production.
- Cultural adoption requires addressing the anxiety of artists and designers by ensuring interfaces demonstrate empowerment rather than replacement, while the surge in content generation may paradoxically increase human workloads regarding review, editing, and personalization.
- The market remains uncertain regarding whether the future will be dominated by vertical solutions, next-generation generalist apps, or proprietary company-built assistants, with early feedback on beauty apps indicating a need to refine empathy and language consistency to avoid robotic interactions.