Technologies in 2024 and Predictions for 2025. Part 1
The Evolution of AI Models and Emerging Trends
In the first part of our new podcast series with Dr. Pierre Taner Kirisci, a blockchain and AI expert, we explore the rise of diverse AI models and the future of AI technologies. Dr. Kirisci revisits predictions made a year ago, examining how AI has evolved and providing insights into what to expect in 2025. Discover how specialized AI models, AI agents, and the integration of blockchain are shaping the future of business and technology.
A year ago, we explored the evolving landscape of AI technologies, spanning blockchain, AI, and the fintech ecosystem. I came up with some predictions and now, 12 months later, it’s time to assess the outcomes and share the insights on where the journey is actually going.
The Rise of Diverse AI Models
A prominent forecast was the emergence of smaller, more specialized AI models. This prediction has been fulfilled beyond expectations. While large-scale models like ChatGPT continue to dominate discussions and usage, the past year has witnessed the introduction of various alternatives, such as BERT, Llama, T5, Mistral, Gemini, XAI, and Claude by Anthropic. Each of these models comes with distinct strengths and weaknesses, offering unique capabilities tailored to specific applications.
Although these models are often referred to as “smaller,” they remain large language models (LLMs) trained on vast datasets. Their defining characteristic lies in their generic usability, enabling a broad range of applications. This evolution highlights the necessity for an understanding of the taxonomy of AI models to better grasp their roles and future trajectories.
Categorizing AI Models: AI Taxonomy
To comprehend the current state and potential developments of AI models, it’s crucial to establish a clear taxonomy:
Commodity Models
These models, including popular ones like GPT, Claude, and Gemini, are trained on publicly accessible data encompassing text, audio, and images. While they offer versatility, their generic nature raises concerns about trustworthiness. Users must often validate the outputs due to the phenomenon of AI hallucinations, where the models generate plausible-sounding but inaccurate information.
Specialized Complexity Models
The next tier involves models trained with domain-specific and expert-level data. These models, while still leveraging some public data, incorporate specialized knowledge, enhancing their reliability and value for businesses. As a result, they provide higher value for the business customers..
Proprietary Models
These models are trained on proprietary or internal data, such as confidential benchmarks or subscription-based datasets. Tailored for specific corporate ecosystems, they deliver significant benefits by leveraging unique, in-house data. Examples include domain-specific foundation models, such as those designed for economic or healthcare applications.
Generic AI Models: From Tools to Companions
Generic LLMs, widely accessible and customizable, are increasingly being used as alternatives to traditional web browsers. This trend is expected to grow, with models becoming integral to everyday tasks by 2025-2026. My personal opinion on these generic LLM models is that they will become just like what web browsers and search engines became for us in the 90s.
The Emergence of AI Agents
A notable development unforeseen last year is the rise of AI agents. Unlike traditional LLMs, these agents specialize in specific tasks and interact in multiple modes through text, voice, image, and audio. They function as companions, simplifying tasks such as booking flights or finding the best deals, thus, represent a proactive evolution of AI capabilities.
AI agents are gaining traction as the next big innovation, with the potential to revolutionize application interaction by providing context-aware and time-sensitive assistance.
For instance to find the cheapest flight within the next two weeks. It takes effort to do this yourself. It’s always a burden to do that, and an AI agent can do that for you. There are many other examples where AI agents can come in, and I’m quite convinced that AI agents will become the next big thing
Unlike LLMs, which predict likely outputs, future AI agents are expected to analyze and adapt to user contexts, enabling more natural and meaningful interactions.
Leveraging AI for Corporate Success
Corporations today are dealing with transaction and business data, making AI essential for extracting meaningful insights. AI, especially through AI agents, provides an ideal solution for navigating and analyzing this data. While traditional user interfaces in the crypto and blockchain space remain complex and cumbersome, these environments are perfectly suited for AI agents. They can efficiently interact with these abstract interfaces and derive value from the ecosystem.
In the next one to two years, businesses that capitalize on their corporate data using AI and blockchain infrastructure will have a significant advantage. Companies that develop efficient and user-friendly AI-driven solutions to analyze financial, transactional, or proprietary data will likely lead in their respective fields. The ability to integrate high-quality internal data into AI models is a crucial determinant of success.
I think there will be a lot of things happening here in the next one to two years. Corporations will be recognizing that they have to do something to leverage their corporate data to gain benefits out of this. If you could provide an efficient way to analyze the data with the help of AI agents using blockchain technology as a back-end infrastructure, you will be the winner of this game.