The Future of Artificial Intelligence: Power, Control, and the Responsibility of Those Who Build It

By Laveda Jones

Artificial Intelligence has become one of the most transformative technologies in human history. It has the potential to revolutionize medicine, education, science, engineering, entertainment, business, and countless other fields. At the same time, like every powerful technology before it, AI introduces new questions about responsibility, control, and unintended consequences.

I personally do not subscribe to predictions of an inevitable AI singularity occurring naturally, where artificial intelligence suddenly becomes self-aware and rapidly surpasses human intelligence without human involvement. Many predictions surrounding the singularity are speculative, and there is significant debate among scientists, engineers, and futurists about whether such an event is even possible.  Right now, A.I. infrastructure is in the hands of billion-dollar corporations and companies (some privately owned).  Take one bad actor that has a political agenda (looking at you Elon Musk) and he or she could literally be as dangerous and a nuclear threat.  Neural networks, the engine behind A.I. is autonomous.  It is an algorithm that should not allow bad biases, not to say that these could be tweaked for malicious purposes.  What could turn an A.I. to “go rogue in a destructive manner” is the training data you provide it.  Feed it politically skewed training data, hateful content, destructive based content, or anything that is a “purposeful” subset of unbiased sampling of public data, and the A.I. could do bad things.  Public data contains biased information by default, but because that data is spread out over unbiased opinion, it gets balanced out.

However, I do believe we should take seriously the possibility that humans could create increasingly powerful AI systems that have significant influence over society. The greatest concern may not be an AI system spontaneously becoming destructive on its own, but rather what happens when extremely powerful technology is placed in the hands of individuals, organizations, or governments with harmful intentions.

Right now, A.I. infrastructure is extremely expensive, resource hungry beasts, needing billions of GPU chips, servers, high speed communication channels, and of course the land to build the data centers, many megawatts of electricity, the water to cool the hardware, etc.  What is surprising about this is that the majority of those resources are applied to the “training process”.  Trillions of data inputs, the processing to obtain those inputs, human interaction, etc. End the end, although it’s ever adapting, you end up with a data model which is essentially a polynomial parametric equation.  Of course, this equation probably fits on billions of pages of printed paper.  A.I. is based on regression analysis, in itself, has resulted in fascinating discoveries. Neural Networks are nothing more than these formulas at an enormous auto computed and scale, in storable format. Regardless, the model created is comparatively simple compared to the massive training requirements to produce them. Consider these models as intellectual property of the big companies that produce them, meaning the typical workflow is something like this:

Prompt and attachments -> post to A.I. server -> run through model -> download result outcome.

As you can see in the previous example, there is a long journey from user to outcome.  And that journey takes your input to a A.I. company trained model, meaning that it typically includes “globally available data training inputs”, not your organization’s own data.  There are many use cases where you would not want outside data mixed in with your own training data.  Considering the model has a smaller footprint than everything fed into it, this means, and this is already happening,  that companies are allowing a controlled instance of the model (formulas) to travel as a unit to the organization’s own data centers, or even a single user computer, and allowing for organizational slice input of the training data.  This is huge, because it gives organizations better control.  This is the very paradigm of agentic A.I. which reduces the round trips and performance of input output.

The Concentration of AI Power

Today, much of the world's most advanced AI infrastructure is controlled by extremely large technology companies. Developing state-of-the-art AI models requires enormous financial resources, specialized hardware, massive data centers, highly skilled engineers, and access to vast amounts of computational power.

This creates a unique situation in technology history: a small number of organizations have access to capabilities that can influence communication, information discovery, business operations, and even public opinion.

The concern is not simply that these companies exist. Many technological advances throughout history have required large investments and specialized expertise. The concern is that when powerful technology becomes concentrated among a limited number of actors, society must ask important questions:

• Who controls the technology?
• What safeguards exist?
• How transparent are the systems?
• Who is accountable when something goes wrong?

A powerful AI system controlled by a malicious individual, organization, or government could potentially create serious risks. While AI is not directly comparable to nuclear weapons, the comparison highlights an important point: some technologies become strategically important because of their ability to affect society at scale.

Potential risks include:

• automated cyberattacks,
• large-scale misinformation campaigns,
• financial manipulation,
• privacy violations,
• autonomous weapons systems,
• exploitation of vulnerabilities in critical infrastructure.

The danger does not necessarily come from AI having its own intentions. The danger comes from humans using powerful tools without sufficient safeguards.

AI Does Not Think Like Humans

A common misconception is that artificial intelligence works like a human brain. It does not.

Modern AI systems are based largely on neural networks, which are mathematical systems designed to identify patterns in enormous amounts of data. These systems contain billions or even trillions of numerical parameters that are adjusted during training.

A neural network does not have beliefs, desires, emotions, or personal goals. It does not "want" anything. It does not independently decide what is good or bad.

The apparent intelligence comes from its ability to recognize patterns and generate responses based on those learned patterns.

A useful analogy is that an AI model is not like a human mind; it is more like an extremely sophisticated prediction engine. It analyzes relationships between information and generates the most statistically likely response based on its training and instructions.

The ability of AI systems to appear intelligent comes from the incredible scale of these mathematical operations.

The Importance of Training Data

One of the most important aspects of artificial intelligence is the information used to train it.

AI models learn from massive datasets that may include:

• books,
• articles,
• websites,
• scientific papers,
• images,
• code,
• conversations,
• and other publicly available information.
• However, public information is not automatically objective or unbiased.
• Human knowledge contains:
• cultural assumptions,
• historical biases,
• misinformation,
• propaganda,
• stereotypes,
• incomplete information,
• and conflicting viewpoints.

Because AI systems learn patterns from human-generated information, they can unintentionally inherit some of these problems.

This is why AI developers spend significant effort on:

• filtering training data,
• improving data quality,
• human feedback,
• safety testing,
• bias evaluation,
• and alignment techniques.

The quality of an AI system depends not only on the amount of information it receives but also on the quality and diversity of that information.

A model trained on intentionally manipulated or heavily biased data could produce outputs that reflect those distortions.

This creates an important possibility: AI does not need to become "evil" to produce harmful results. A poorly designed objective, poor-quality data, or malicious human input may be enough to create significant problems.

The Massive Infrastructure Behind AI

Another aspect of AI that is often overlooked is the enormous physical infrastructure required to create these systems.

Modern AI requires:

• thousands or millions of specialized GPU processors,
• massive data centers,
• high-speed networking,
• enormous amounts of electricity,
• advanced cooling systems,
• specialized engineering teams,
• and substantial financial investment.

Training a large AI model can require months of computation across thousands of machines.

One of the fascinating aspects of AI is that the training process is enormously larger than the final model itself.

During training, AI systems process enormous quantities of information. However, the final model is essentially a compressed mathematical representation of patterns discovered during that process.

The model does not contain a complete copy of every document, image, or piece of information it was trained on. Instead, it contains numerical relationships that allow it to generate useful outputs.

In simplified terms:

Massive amounts of data + enormous computing power + training algorithms = AI model

The result is a relatively compact mathematical structure compared to the enormous amount of information used to create it.

The Future of AI: From Public Models to Personal and Organizational Intelligence

The current AI model most people interact with follows a simple workflow:

User prompt and attachments → AI company's servers → AI model processes request → response returned

This approach is extremely powerful, but it creates questions for organizations that handle sensitive information.

Businesses, governments, universities, and research organizations may not want their private information sent to an external AI service.

Examples include:

• confidential business plans,
• medical information,
• intellectual property,
• legal documents,
• financial data,
• proprietary software code.

This is leading to a major shift in AI development.

Instead of relying only on massive public AI models, organizations are increasingly exploring:

• private AI deployments,
• customized AI assistants,
• local AI systems,
• retrieval-augmented generation (RAG),
• and AI agents connected to internal business systems.

The future may not simply be one giant AI model answering everyone's questions.

Instead, we may see a world where every organization has its own AI systems trained or connected to its own knowledge base.

The Rise of Agentic AI

One of the most significant emerging trends is what is called agentic AI.

Traditional AI responds to requests:

"Answer this question."

Agentic AI attempts to complete larger goals:

"Analyze this problem, gather information, create a plan, use available tools, and complete the task."

An AI agent may be able to:

• access databases,
• interact with software,
• manage workflows,
• monitor systems,
• analyze information,
• and perform multi-step operations.
• This represents a major evolution in computing.

For decades, humans have interacted with computers by giving commands through keyboards, menus, and applications. Agentic AI introduces a new possibility: humans describing objectives while AI systems handle much of the execution.

However, this also increases the importance of security, oversight, and accountability.

A system capable of taking actions is fundamentally different from a system that only provides information.

The Real Question Is Not Whether AI Will Replace Humanity

Much of the public conversation about AI focuses on whether AI will become smarter than humans.

But perhaps the more important question is:

How will humans choose to use AI?

Every major technology has created both opportunities and risks.

The printing press accelerated education but also misinformation.

The internet connected the world but also created new forms of manipulation.

Social media gave people a voice but also amplified harmful content.

AI will likely follow a similar pattern.

The future of artificial intelligence will not be determined only by algorithms and computers. It will be determined by human decisions:

• Who controls AI?
• Who benefits from AI?
• Who is protected from AI's risks?
• What ethical standards guide its development?
• Artificial intelligence is not inherently good or evil. It is a tool created by humans.
• The greatest challenge of AI may not be creating intelligence.
• The greatest challenge may be creating the wisdom and responsibility to use it.