The AI industry is evolving and coming up with recent and unique research and models every day. Whether we discuss healthcare, education, retail, marketing, or business, Artificial Intelligence and Machine Learning practices are starting to shift how industries operate. Every organization is adopting AI to incorporate its potential in people’s on a regular basis lives. With automation and AI’s excellent capabilities to learn, reason, and execute decision-making, the sphere of AI is rapidly advancing.

The well-known Large Language Models, which have recently gained numerous popularity, are the most effective example of AI’s takeover 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 a lot of the individuals aware of it. Recently, a Twitter user, Jay Hack, discussed an intriguing trend in AI referred to as the stacking of AI models in his tweet thread. Referring to the concept as “models all the best way down,” Jay has mentioned how AI models use other similar models to perform tasks and perform decision-making.

Stacking is largely having an AI model that may invoke other models for solving a posh task, thereby leading to emergent intelligence. The essential idea behind the approach is to have AI models use other models as tools or mediums to perform a subtask or multiple subtasks. A few of the quoted examples are – GPT generating its own copies for solving subtasks, GPT using a vision model for drawing beautiful portraits, etc.

Jay has discussed the self-referential nature of stacking that may help develop models having Artificial General Intelligence (AGI). He has mentioned how by stacking multiple AI models on top of one another, each model could make use of the capabilities of the models below it, leading to a system with greater overall intelligence. This approach is seen because the frontier in constructing systems that may perform tasks that were previously considered out of reach for AI.

Two of the recent examples of such LLMs which have utilized this idea for nice purposes are BabyAGI and AutoGPT. Each of those LLMs recursively call themselves. On the one hand, where BabyAGI trains and evaluates various AI agents in a simulated environment and tests their ability to learn and perform difficult tasks. Alternatively, AutoGPT uses GPT-4 and GPT-3.5 via API to create full projects by iterating by itself prompts. AutoGPT even created a web site using React and Tailwind CSS in under three minutes.

Other domains where stacking is popularizing are ViperGPT, which provides GPT access to a Python REPL (Read-Eval-Print Loop) and a high-level API for manipulating Computer Vision models. SayCan can also be emerging in robotics, where an LLM is used because the backbone for robotic reasoning. One other recent project called ‘toolkit.club’ uses LLMs to construct and deploy tools for other AIs. This uses a loop where the agent asks for a tool, the tool is made and deployed by LLM, and the agent thus uses the tool.

Consequently, stacking AI is rapidly advancing and opening doors to recent capabilities. It could solve complex tasks that a single LLM query may not give you the option to unravel. With correct usage and overcoming the constraints regarding AI safety, this approach can work wonders in the longer term for further developments.

This text is predicated on this tweet thread that discusses the above topic. All Credit For This Research Goes To the Researchers on This Project. Also, don’t forget to affix our 18k+ ML SubReddit, Discord Channel, and Email Newsletter, where we share the most recent AI research news, cool AI projects, and more.

🚀 Check Out 100’s AI Tools in AI Tools Club