Beyond the Algorithm: What’s Next for AI?

As artificial intelligence becomes foundational to nearly every sector—from healthcare to logistics to creativity— the question is shifting from “What can AI do?” to “What should we do with it next?”

At the Creative Destruction Lab (CDL-Paris) Artificial Intelligence Stream, that future is being built from the ground up. The program is designed specifically for deep tech ventures developing commercial opportunities based on the application of machine learning and artificial intelligence.

In its most recent session, held in late December 2024 at Arboretum—Europe’s largest solid wood campus with a minimal carbon footprint—we spoke with:

• Nicole Pereira, CDL-Paris AI Stream Lead
• Professor Thomas Åstebro, Academic Director of CDL-Paris and Executive Director of the ION Management Science Lab at HEC Paris

Together, they shared insights on emerging AI trends and how CDL’s mentorship model is helping shape the next generation of machine learning ventures.

From Broad Buzz to Deep Focus

The rapid evolution of AI has generated massive hype, and rightly so. Breakthroughs in large language models, generative AI, and multimodal systems have captured public imagination and venture capital alike. But beneath the surface of high-profile applications lies a quieter, more strategic shift: AI is becoming deeply specialized.

“Artificial intelligence is a very broad concept,” said Thomas Åstebro. “We all know about the large language models, but there’s a broad spectrum of focused applications where AI is just beginning to deliver real value.”

That “focus” is where CDL-Paris is placing its bets. Rather than building the next chatbot or text-to-image engine, many of the most promising ventures are now prioritizing foundational tools and technologies—the unseen layers that power everything from enterprise automation to responsible AI governance. Instead of chasing trends, the program is nurturing ventures solving specific, high-value problems, especially in areas like:

• Synthetic data generation
• Computer vision
• Explainable AI
• Infrastructure and model optimization

“We really try to focus more on the infrastructure level,” said Nicole Pereira, “supporting companies that are building the backbone of trustworthy and scalable AI.”
 

The Rise of Small, Specialized Models

As the initial wave of AI excitement crests, the conversation is shifting from sheer model size to strategic application. While 2023 was dominated by massive, general-purpose large language models (LLMs), the next chapter in AI may be defined not by scale—but by specialization. The horizon may belong to smaller, more specific systems - models that are lighter, faster, and purpose-built for specific use cases.

“These large language models in the future might be called small,” Thomas Åstebro predicted. “Not as big, not as energy consuming, but very focused—and much more reliable for a particular application.”

This shift speaks to a wider trend in AI commercialization: efficiency over scale, precision over generalization. Success won't come from building the biggest models—but from building the right ones.

CDL’s Role in Scaling with Focus

With AI capable of serving dozens of industries, many startups face the same early challenge: too many potential applications, and not enough clarity on where to begin.

“That tends to be the challenge,” said Nicole Pereira. “Companies come in with vast potential, and we work with them to answer: What market do we enter first? Where do we build traction now?”

CDL’s methodology is specifically built to help startups sharpen objectives, prioritize time and resources, and engage with experts on the frontier of research.

Each venture:

• Defines clear 8-week goals for each session
• Receives targeted mentorship from exited entrepreneurs, AI scientists, and active investors
• Engages in four intensive sessions designed to drive commercial clarity and technical focus

And uniquely, CDL is a non-profit organization. There are no fees to participate, and CDL takes no equity in the ventures it supports. It’s all shaped for massively scalable, seed-stage, science- and technology-based companies.

A Global Platform for AI Deep Tech

The CDL-Paris Artificial Intelligence Stream, operated by the HEC Paris Innovation & Entrepreneurship Institute and its Deep Tech Center, is part of a global network of CDL sites that support breakthrough science-based ventures across themes such as Climate, Space, Health, Blockchain, and more.

Focused specifically on the commercialization of AI-driven technologies, the AI Stream brings together a global community of experienced entrepreneurs, leading scientists, active investors, and CDL Fellows to help founders scale ventures built on the advancement and application of artificial intelligence.

CDL’s unique ecosystem also includes HEC Paris business students, who work hands-on with participating ventures—developing financial models, assessing market opportunities, and refining strategic roadmaps for growth.

This collaborative model is particularly powerful in the field of AI, where many founders come from academic or technical backgrounds. At CDL, they receive not just access to capital, but structured mentorship to build go-to-market strategies, navigate ethical and regulatory complexities, and make critical decisions that accelerate their path from lab to market.
 

Watch the interview with Gabriele Mancini, AI Act Lead Author, during one of the session of the AI Stream:

What’s Next: Marseille

As the CDL-Paris AI stream prepares for its final session, the ventures that remain have passed through rigorous rounds of mentorship and evaluation. The concluding event will take place in April 2025 at TANGRAM in Marseille—CMA CGM’s new innovation hub dedicated to tackling global challenges through technology.

There, CDL will bring together ventures and mentors from four streams: Artificial Intelligence, Climate, Space, and Next Generation Computing. Each stream shares a single mission: to help founders build what’s next—with purpose and precision.

Stay tuned for insights from the final session in Marseille, where CDL’s AI, Climate, Space, and Next Gen Computing streams will converge.

Watch the teaser for the Marseille finale: