The AI industry is evolving and introducing new and unique research and models on a daily basis. Whether we’re talking about healthcare, education, retail, marketing, or business, AI and machine learning practices are beginning to change the way industries operate. All organizations are embracing AI to include its potential in people’s daily lives. With automation and AI’s excellent capabilities to learn, reason, and execute decision making, the field of AI is advancing rapidly.
The well-known big language models, which have gained a lot of popularity recently, are the best example of AI conquest of the world. The famous ChatGPT, which uses the GPT transformer architecture to generate content, is currently the talk of the town and the go-to chatbot for most people who know it. Recently, a Twitter user, Jay Hack, discussed an intriguing trend in AI known as AI model stacking in his tweet thread. Referring to the concept as “complete models,” Jay mentioned how AI models use other similar models to perform tasks and make decisions.
Stacking is basically having an AI model that can invoke other models to solve a complex task, resulting in emergent intelligence. The main idea behind the approach is to have AI models use other models as tools or means to perform one or multiple subtasks. Some of the examples cited are: GPT generating its own copies to solve subtasks, GPT using a vision model to draw beautiful portraits, etc.
Jay has discussed the self-referential nature of stacking which can help develop models that have Artificial General Intelligence (AGI). He has mentioned how by stacking multiple AI models on top of each other, each model can make use of the capabilities of the models below it, resulting in a system with higher overall intelligence. This approach is considered the frontier in building systems that can perform tasks previously considered outside the scope of AI.
Two of the recent examples of such LLMs that have used this concept to great purpose are BabyAGI and AutoGPT. Both LLMs recursively call themselves. On the one hand, where BabyAGI trains and tests various AI agents in a simulated environment and tests their ability to learn and perform difficult tasks. On the other hand, AutoGPT uses GPT-4 and GPT-3.5 via API to create entire projects iterating on your own directions. AutoGPT even built a website using React and Tailwind CSS in less than three minutes.
Other domains where stacking is becoming popular are ViperGPT, which provides GPT access to the Python REPL (Read-Eval-Print Loop) and a high-level API for manipulating Computer Vision models. SayCan is also emerging in robotics, where an LLM is used as the backbone for robotic reasoning. Another recent project called ‘toolkit.club’ uses LLM to build and implement tools for other AIs. This uses a loop where the agent requests a tool, LLM creates and deploys the tool, and thus the agent uses the tool.
Consequently, AI stacking is advancing rapidly and opening the doors to new capabilities. It can solve complex tasks that a single LLM query might not be able to solve. With correct use and overcoming the limitations related to AI security, this approach can work wonders in the future for future developments.
This article is based on this tweet thread which deals with the previous topic. All credit for this research goes to the researchers of this project. Also, don’t forget to join our 18k+ ML SubReddit, discord channeland electronic newsletterwhere we share the latest AI research news, exciting AI projects, and more.
🚀 Check out 100 AI tools at AI Tools Club
Tanya Malhotra is a final year student at the University of Petroleum and Power Studies, Dehradun, studying BTech in Computer Engineering with a specialization in Artificial Intelligence and Machine Learning.
She is a data science enthusiast with good analytical and critical thinking, along with a keen interest in acquiring new skills, leading groups, and managing work in an organized manner.
