top of page
Writer's pictureBatiste Roger

The Generative AI Market: A New Segmentation Inspired by Web3

Une réflexion de Batiste Roger, CTO d'Odonatech


The rise of generative AI is rapidly transforming the technological landscape, particularly in the banking and insurance sectors. A new market segmentation seems to be emerging, inspired by the structure of Web3 (*).


WEB3

(*) Web3: Often referred to as Web 3.0, Web3 represents the next generation of the Internet, characterized by the integration of decentralized technologies like blockchain. Unlike Web2.0, where data is primarily controlled by centralized corporations, Web3 aims to give users greater control over their data and online interactions. This decentralization enables more secure, transparent, and autonomous transactions, paving the way for new financial applications and services, among others.


TLDR: The generative AI market can be segmented into three layers inspired by Web3: Layer 1 (Generalist AIs), Platforms (Integration Tools), and Layer 2 (Specialized AIs). This segmentation provides a relevant framework for understanding the rapid evolution of this sector, particularly in banking and insurance.


Segmentation of the Generative AI Market

Let's jump right in !


Layer 1: Basic Generative AIs

  • Examples: ChatGPT from OpenAI , Claude from Anthropic , Gemini from Google 


  • Characteristics: Large-scale generalist models capable of handling a wide range of linguistic tasks. Available in various sizes and can be fine-tuned for specific tasks.


  • Impact in Finance: These models can be used in finance for low-risk use cases with human review, such as brainstorming in marketing, searching large databases, or drafting responses for customer support teams.


Platforms

  • Examples: Dust , LangChain , Amazon Web Services (AWS) Bedrock


  • Role: These platforms facilitate the integration and use of generative AIs in business processes by simplifying development and connection to existing tools (databases, email systems, customer support interfaces, existing apps, etc.).


  • Financial Application: Langchain allows for the secure integration of generative AI into existing workflows, for example, by facilitating access to internal databases with the synthesis capabilities of LLMs. BedRock provides a single API for multiple generative AIs, making it easier (and faster, and cheaper) to switch between them.


Layer 2: Specialized Generative AIs

  • Example: LiLa from Odonatech 


  • Characteristics: These AIs are tailored to specific domains, delivering enhanced precision and reliability within their area of expertise. They expand upon the capabilities of L1 AIs while mitigating or eliminating their primary weaknesses. Their intelligence (both IQ and EQ) appears superior to that of standard AIs.


  • Advantage in Finance: These AIs incorporate a deep understanding of regulatory nuances and industry practices, enabling more precise and compliant evaluations. However, they are somewhat slower and more expensive due to the greater complexity of their “brains.” They are best suited for use cases where quality and compliance is more important than cost per token.


On the price of L2 AIs : there are instances where L2s can be more cost-effective. Rather than employing an expensive, large language model, it's often advantageous to use a specialized AI with the right architecture and knowledge base, combined with a small language model. In such cases, L2 might actually be both cheaper and better!


Similarities with Web3

This structure is reminiscent of Web3:


  • Layer 1: Basic blockchains like Ethereum or Bitcoin providing fundamental infrastructure.

  • Platforms: Development tools like Infura or Alchemy facilitating access and use of blockchains.

  • Layer 2: Scaling solutions like Optimism and The Arbitrum Foundation optimizing performance for specific use cases.


In both fields, there is a trend towards more efficient and specialized solutions built on a generalist foundation. Additionally, it is clear that the key success factors for Layer 1 and Layer 2 are not the same, although they may share clients. Furthermore, the parallel development of Platforms and Layer 2 raises questions about their ability to cooperate; often, Layer 2 solutions are built directly on Layer 1 without using the platforms. Will the same be true for AI?


Key Differences and Implications for the Financial Sector


It is evident that Layer 2 solutions address the limitations of Layer 1. However, despite structural similarities, the functions and challenges differ between AI and Web3:


  • Layer 2 Web3: Focuses on scalability and reducing transaction costs.

  • Layer 2 AI in Finance: Aims for specialization, increased reliability, and regulatory compliance.


    The weakneses of the Layer 1s are not the same, which means Layer 2 have something different to "correct".


Implications Specific to the Financial Sector


  • Regulatory Compliance: Layer 2 AIs like LiLa can comply with stringent regulations, both by providing better responses and ensuring the quality of those responses.


  • Investment Advice: For wealth management, Layer 2 AIs can generate investment recommendations by considering the client's risk profile and MiFID II regulations on the suitability of financial products. They use approaches like RAG, which can be extended and improved to significantly surpass current performance.


  • Domain Knowledge Integration: Layer 2 AIs can incorporate decades of financial expertise, enabling more nuanced analyses than generalist models. In insurance risk assessment, for example, they can consider subtle factors specific to each type of policy.


  • Personal Data Requirements: Regulations such as GDPR in Europe shape the development of AI solutions, driving innovation in data protection and algorithmic decision transparency. Layer 2 AIs add a protective layer to the entire personal data processing chain.


  • The human touch : Layer 2s can feel significantly more empathetic, attentive, and... human. This could change more than we anticipate. Are we on the verge of seeing what robo-advisors promised?


Conclusion

The segmentation of the generative AI market, inspired by the structure of Web3, provides a relevant framework for understanding the rapid evolution of this sector, particularly in banking and insurance. 


Specialized generative AIs, such as LiLa from Odonatech, represent a significant advancement in terms of precision, reliability, and regulatory compliance. 


For financial sector actors, adopting these solutions is crucial for improving operational efficiency and complying with regulations. We invite you to explore these possibilities now to gain a competitive advantage and prepare for the future. Contact us to discover how specialized generative AIs can revolutionize your business.




Comentarios


bottom of page