Extensive Research #2: Building Moats with Vertical AI
What is the winning strategy for building your AI product?
Recent launches from tech giants like OpenAI and Google have pushed the boundaries of what these models can achieve, igniting a frenzy of excitement and anticipation across industries. However, amidst this excitement, a sobering reality looms: as Naveen shares, the current transformer and diffusion models underpinning these advancements will eventually hit scaling limitations, much like Moore's Law. While throwing more computational resources and data at these models may yield incremental gains, a real fundamental breakthrough will be required to overcome their inherent constraints.
The Rise of Vertical AI
It is within this context that the concept of vertical AI has emerged as a promising solution to address the industry-specific challenges that traditional horizontal AI solutions have struggled to tackle effectively. Vertical AI refers to specialized artificial intelligence systems tailored to specific industries or verticals, offering industry-specific solutions designed to optimize operations and unlock growth opportunities.
Unlike horizontal AI, which focuses on general business functions, vertical AI is engineered to address the unique challenges and requirements of a particular industry. By incorporating subject matter expertise and leveraging domain-specific data, vertical AI models can generate more personalized and reliable results, tapping into the wellspring of expert knowledge that has traditionally been siloed within each industry.
As the world grapples with the complexities of an ever-evolving technological landscape, the ability to build defensible competitive positions, or "moats," has become paramount. Vertical AI promises to be a potent tool in this endeavor, enabling companies to leverage the power of artificial intelligence while tailoring it to their specific industry needs, thereby creating a differentiated and defensible competitive advantage.
Short-term and Long-term Strategies
As vertical AI gains traction across industries, the strategies employed by players in this space will play a crucial role in determining their long-term success and competitive positioning. In the short term, most vertical AI players are focusing on developing a distinct vertical advantage by creating proprietary fine-tuned models and delivering targeted user experiences tailored to their specific industry.
This approach leverages the power of domain-specific data and subject matter expertise to train smaller, more focused AI models that can outperform larger, more generalized models in terms of latency, accuracy, and cost-effectiveness. By specializing in a particular vertical, these players can create a reinforcing feedback loop, where the more they develop their vertical advantage, the more competitive they become across all parameters.
To deliver the most compelling vertical solution over time, these players recognize the need for deep integration between their AI models and user experiences. As the models become increasingly fine-tuned, future user interface changes must be closely coupled with the underlying models to ensure that the benefits of this specialized training are seamlessly delivered to end-users through intuitive and efficient workflows.
Agentic Automation and the Future of Vertical AI
As the capabilities of artificial intelligence continue to advance, the concept of agentic automation has emerged as a powerful paradigm that promises to revolutionize the way we interact with and leverage AI systems. At the heart of this paradigm lies the notion of intelligent agents – software entities that can perceive their environment, reason about their goals, and take actions to achieve those goals autonomously or in collaboration with human users.
In the context of vertical AI, the combination of agentic automation frameworks and industry-specific models holds immense potential for transforming workflows, streamlining operations, and unlocking new frontiers of efficiency and productivity. By seamlessly integrating the power of large language models (LLMs) with domain-specific knowledge, data sources, and user interfaces, vertical AI solutions can evolve into intelligent agents capable of executing complex, industry-tailored workflows with minimal human intervention.
The concept of "reimagining applications and automating intricate, industry-specific tasks" has captivated the tech world, and it is here that the idea of a "Mixture of Experts" architecture comes to the fore. This approach combines task-specific agents, pre-defined code and workflows, and external tools and data sources to create a seamless and cohesive "scaffolding" for intelligent agents to navigate complex domain-specific challenges. Vertical AI solutions, tailored to specific industries, have the potential to revolutionize how we interact with technology and automate intricate tasks.
A vertical AI solution tailored for healthcare could integrate AlphaFold 3, a model that predicts protein, DNA, RNA, and small molecule interactions with at least 50% better accuracy than existing methods. By combining AlphaFold 3 with agents for medical diagnosis, treatment planning, and patient data analysis, the solution could deliver cutting-edge healthcare experiences.
While the current focus is on refining task-specific agents and industry-tailored workflows, the ultimate ambition of agentic automation is the development of general-purpose agents. These agents would possess the capability to tackle any task or challenge simply by understanding a human-provided description or prompt. This long-term goal drives ongoing research and advancements in areas like reasoning, planning, and context awareness, which remain challenging for current AI models.
Why Vertical AI is Important Now
The rise of vertical AI solutions is well-timed, driven by technological advancements and changing industry dynamics. Two key factors highlight the importance of vertical AI. Firstly, the ability of AI to now harness unstructured data, including contracts, records, and multimedia files, has been revolutionary. This has brought industries that heavily rely on unstructured information into the modern era, as they were previously underserved by technology. Secondly, an estimated 80% of the world's data falls into the unstructured category, emphasizing the significance of this development.
Predictions on the Evolution of Automation
As generative AI continues to mature, and vertical AI solutions gain traction, the automation landscape is poised for a seismic shift that will redefine what is possible across industries in four key areas:
Omni AI Assistants: Intelligent agents will act as omnipresent copilots, augmenting human capabilities and transcending traditional application boundaries.
Human-in-the-Loop Frameworks: Recognizing the limitations of LLMs, human-in-the-loop approaches will be crucial for effective AI deployment, focusing on advisory and assistant-oriented workflows.
Gradual AI Adoption in Automation: The integration of AI in automation platforms will follow a pragmatic "crawl, walk, run" approach, starting with simple tasks and gradually increasing complexity.
Code Generation as a Foundation: Code generation will be a fundamental element in developing genAI-based applications, harnessing the strengths of LLMs in generating code as text with well-defined performance measures.
These predictions envision a future where AI is seamlessly integrated into our daily lives and workflows, acting as intelligent assistants that enhance our capabilities while adhering to human-centric principles. The gradual adoption of AI in automation, coupled with the power of code generation, holds the promise of unlocking unprecedented levels of efficiency, productivity, and innovation across a wide range of industries.
References and further reading
[1] Andrew Ng, AI Agentic Workflows The Batch, Issue 241-245
[2] Evaluating AI systems – Anthropic
[3] AI Agents are disrupting automation: Current approaches, market solutions and recommendations
[4] Reflexion: Language Agents with Verbal Reinforcement Learning - arXiv:2303.11366
Hey there!
I'm Kevin Wang, a product manager by day and a passionate builder of tulsk.io, an AI Agent Tool, in my free time. I'm on a mission to simplify complex concepts in web3 and AI, so they're easier to grasp and apply to your own learning and growth.
Shape Your Future
Are you curious about the forces shaping the future of AI and Web3?
At Curiosity Insights, we dive deep into the AI and Web3 realm, providing you with the insights you need to navigate the complex world of emerging technologies. Our mission is to empower you to make informed investment decisions, drive innovation, and stay ahead of the competition.
Upgrade your understanding with the Premium Curiosity Insights newsletter. Here's what you'll unlock:
In-Depth Research: Stay ahead of the curve with our weekly coverage of the latest AI and Web3 developments. Our expert curation and analysis provide you with a comprehensive understanding of these dynamic fields.
Exclusive Access to Playbook: Get insider access to industry experts and their insights.
Full Archive Access: Dive into our extensive archive covering artificial intelligence and Web3. This treasure trove of knowledge will empower you to explore and discover the trends and technologies that matter.
Join us on this fascinating journey as we unravel the mysteries of AI and Web3. Sign up now and shape your future with Curiosity Insights!