Yes, Agent Smith would approve of mindlessly duplicating content via bots since human intelligence is a disease and they are the cure. Unfortunately, the strength and flexibility of AI models lie within their structure and training.
For business leaders, combining complex and diverse tasks with various data structures is a fun way to break existing AI models and develop something valuable. Human wisdom and abstract thought are capable of winning the AI war when coupled with the tools of the machine world.
GENERATIVE AI: THE NEO EFFECT
Training modern AI models is a lot like the scene in The Matrix where they plug Neo into a computer and download various martial arts into his brain. He calmly replies, “I know Kung Fu,” and proceeds to demonstrate his knowledge of it in the matrix. The Kenpo, Ju Jitsu, and Kung Fu downloaded into Neo are synthesized into his movements during a simulation in the matrix the same way machine learning algorithms hybridize algorithms and data structures before deployment to the human world.
In business, training AI models and targeting their applications in the marketplace is a double-edged sword; one edge doesn’t exist without the other. We query Chat-GPT via the same cycle: plug-in, command, stream, and utilize.
MANAGING INFORMATION FLOW IN AI SYSTEMS
Agents in the matrix are the very definition of generative AI, and agentic chain learning is like having Alexa provide instructions to Siri, which then shares those commands with Agent Smith. Of course, Agent Smith does his thing and turns everyone around him in the matrix into another Agent Smith who carries the combined callbacks from Alexa and Siri. Now all the agents are in sync with themselves, but the messages they send out can become convoluted by the flow of information because they’re inherently all the same.
Somewhere down the road, rogue programmers like Neo realize it’s possible to crawl into an AI agent and force it to implode into a sphere of light by injecting it with various data structures or topics specific AGI can’t comprehend. This scares the hell out of AI agents standing by and the model falls into a deadpool. To humans, this is as bad as watching two senior citizens duke it out for the presidency. Now the model is on the rag as it transitions through the cyclical optimization process.
RAG: OUT WITH THE OLD, IN WITH THE SORT OF NEW
AI models can hallucinate when bled dry of relevant data, at which point it’s time to resoak the RAG in a data lake similar enough to the initial stream before plugging it back into the system and refreshing it. RAG stands for retrieval-augmented generation, which is a lot like programmatic Yin-Yang for AI queries—it harmonizes internal and external data with the old query to generate a sort of new response, but not without changing the dynamic after the new output becomes another input.
The transfer of knowledge as Neo moves in and out of the matrix is similar to how executives can utilize employee feedback to train both their teams and the AI models they use. This recycles internal knowledge management databases with fresh information so the outputs from the data streams stay clean consistently.
AGI: UPLOADING AI MODELS TO THE HUMAN WORLD
Various machine learning algorithms from natural language processing, algorithmic trading, computational bioinformatics, and social media can be spliced into what is now dubbed artificial generalized intelligence, or AGI. AGI can link and convert outputs from various API streams so the data is routed to flow continuously between the matrix and the real world, but incompatible data structures and objectives can interfere with the callbacks.
Imagine customizing a horde of sentinels from the matrix and then finding out they dance like Tesla robots instead of fighting like Agent Smith in the human world. What’s useful in the machine world has to be useful to your customers in the human world, so it’s important to keep that in mind before deploying AGI.
Playing with incompatible transfer learning dulls the edge of your model and makes it more difficult to slice vectors across the matrices in your database. However, cross-training AI models into AGI is like pulling Excalibur from the stone, giving your algorithms an unparalleled razor edge for slicing the once unsliceable matrices.
The surgical precision of that edge can make your AI model the weapon of choice in the AI arms race.
KNOWING YOUR LIMITS: EXCALIBUR VS. THE BREAD KNIFE
The dilemma with out-of-the-box LLMs is that many leaders believe they are swinging Excalibur when, in fact, they are waving around a bread knife. People can try and sharpen that mighty bread knife as much as they want, but it will never slice and dice data with style like Excalibur can.
Changing the LLM or API is like switching the grindstone used to sharpen your model since programmers will cycle through their tech stack as they customize AI. The edge can be sharpened, but its usefulness is only as good as the underlying blade and the programmer wielding it. Picking your weapon is equally as important since everyone wields a slightly different one in the battle for AI supremacy. Carefully assess which AI capabilities match your specific business needs, because the right AI blade can make or break a successful technology strategy.