Understanding the difference between artificial intelligence, general intelligence, and consciousness, and what the next decade might actually look like.
Artificial Intelligence has become part of everyday life extremely quickly. Whether we are using ChatGPT to write content, Microsoft Copilot to help with work, or AI-powered tools to generate images, videos, and even software, the technology is advancing at a pace that would have seemed almost impossible just a few years ago.
However, as AI becomes increasingly capable, two terms are starting to appear more frequently in conversations about its future: Artificial General Intelligence (AGI) and Sentient AI.
These terms are often used interchangeably, particularly in the media and on social platforms. The reality is that they describe two very different concepts, and understanding the distinction is important when considering where AI is heading.
AGI is about intelligence, reasoning, learning, and the ability to perform tasks across different areas. Sentient AI, on the other hand, raises a much deeper question about whether a machine could actually experience anything, rather than simply process information and produce an answer.
I find both concepts fascinating, especially when considering how quickly technology has evolved during my career. We have moved from computers that required explicit instructions for almost everything to systems that can generate complex content, reason through problems, and interact with us in remarkably natural ways.
But does being intelligent mean being conscious? Could a machine eventually understand that it exists? And perhaps most importantly, how close are we to either of these becoming reality?
Understanding Artificial General Intelligence
To understand AGI, we first need to look at the artificial intelligence we are using today.
Most current AI systems are examples of what is commonly described as Artificial Narrow Intelligence (ANI). These systems are designed or trained to perform particular categories of tasks, even when those categories are extremely broad.
For example, AI can identify objects in photographs, translate languages, generate software code, analyze medical images, and answer complicated questions. Some systems can perform several of these tasks simultaneously, which is why modern AI feels considerably more capable than earlier generations.
However, impressive capabilities do not necessarily mean that a system possesses general intelligence.
A model might solve a complicated mathematical problem yet struggle with a simple situation requiring everyday common sense. It might generate thousands of lines of working code but misunderstand an instruction that a human developer would immediately recognize as ambiguous.
Artificial General Intelligence represents the next proposed stage of AI development.
AGI generally refers to an artificial system capable of learning, reasoning, adapting, and performing a wide variety of intellectual tasks at a level comparable to humans. Rather than being limited to particular applications, it would be able to apply knowledge and skills across unfamiliar situations.
Imagine an AI system that could independently learn a new programming language, understand a business problem, develop a solution, test it, recognize mistakes, and improve its approach without requiring detailed human instructions at every stage.
Now imagine that same system moving into an entirely different discipline, such as engineering, scientific research, or financial analysis, and demonstrating comparable capabilities.
That ability to transfer knowledge and adapt across different domains is central to the idea of AGI.
There is an important qualification, though. Researchers do not universally agree on exactly what qualifies as AGI. Some definitions emphasize human-level performance across cognitive tasks, while others place greater importance on adaptability, autonomy, or the breadth of problems a system can solve.
This matters because we might not wake up one morning to discover that AGI has suddenly arrived. Instead, we may progressively develop systems that satisfy different definitions of general intelligence.
Google DeepMind has proposed a framework that measures AGI development across multiple levels of performance and generality, recognizing that intelligence is not necessarily a single milestone.
Why Today’s AI Is Not Necessarily AGI
One of the remarkable things about modern AI is how convincingly it can appear to understand what we are asking.
When interacting with a large language model, we can have detailed conversations, explore complicated technical problems, and receive explanations that feel remarkably similar to speaking with another person. However, there are still significant limitations.
Today’s systems can produce incorrect information with considerable confidence. They can struggle with unfamiliar situations, make reasoning errors, and require human verification when completing complex tasks.
Even systems designed to operate autonomously can encounter problems when working through extended sequences of decisions.
This becomes particularly important in professional environments. An AI system might be able to generate a PowerShell script, explain a Microsoft 365 security configuration, or recommend changes to a network architecture. But that does not mean it fully understands every operational consequence of implementing those changes.
A human architect considers business requirements, organizational culture, potential failures, security implications, and experiences gained from previous projects. AI can increasingly assist with these considerations, but consistently demonstrating comparable judgment across unfamiliar environments remains a challenge.
It is worth emphasizing that AGI does not necessarily need to think exactly like a human. A machine could achieve broadly human-level intellectual capabilities using entirely different internal processes.
The important measure is what it can reliably accomplish, rather than whether its internal mechanisms resemble a biological brain.
What Would AGI Actually Change?
If genuine AGI becomes possible, its impact could extend considerably beyond the productivity improvements we are experiencing today.
Instead of simply helping people perform tasks more efficiently, AGI could potentially undertake complex projects independently, moving between different areas of knowledge as required.
Consider scientific research. An AGI system might analyze existing research, develop hypotheses, design experiments, interpret results, and propose new directions for investigation.
In business, it could potentially understand an organization’s operations, identify inefficiencies, design improvements, and coordinate their implementation.
Healthcare, engineering, education, cybersecurity, and software development could all experience significant changes.
However, increased intelligence would also introduce new challenges. Systems capable of making complicated decisions independently would require appropriate governance, accountability, and safeguards.
We would need to think carefully about what decisions these systems should be allowed to make, how their actions are monitored, and who remains responsible when something goes wrong.
The transition from AI assisting humans to AI independently making consequential decisions may prove just as significant as the development of AGI itself.
Moving Beyond Intelligence: What Is Sentient AI?
This is where the discussion becomes particularly interesting.
While AGI focuses on what a machine can accomplish, sentience focuses on whether a machine can actually experience anything.
Sentience is generally understood as the capacity to have subjective experiences, particularly experiences that can feel pleasant, unpleasant, or otherwise meaningful to the entity experiencing them.
For humans, these experiences include things like pain, pleasure, fear, happiness, and sadness.
Consciousness is closely related but is a broader and heavily debated concept. It generally concerns subjective awareness or experience, while self-awareness refers more specifically to the ability to recognize oneself as an individual.
These distinctions are important because sentience, consciousness, and self-awareness are not necessarily identical.
Consider what happens when you accidentally touch something extremely hot.
Your nervous system detects the heat, your brain processes the information, and you instinctively move your hand away. But there is also the experience of pain. It actually feels like something.
A machine could be programmed to detect excessive heat and immediately withdraw a robotic arm. It might even announce that the temperature is painful or dangerous.
However, that does not establish that the machine experiences pain.
The difference is between processing information about an event and having a subjective experience of that event.
This is the fundamental question surrounding Sentient AI.
Could we eventually develop artificial systems that do not simply recognize emotions, describe experiences, or simulate human responses, but actually experience something internally?
At present, we do not have a scientifically accepted method for demonstrating that an AI system possesses subjective experience.
The Difference Between AGI and Sentient AI
One of the biggest misconceptions is that achieving AGI would automatically create a conscious or sentient machine. That is not necessarily true.
An artificial system could potentially possess extraordinary intellectual capabilities without experiencing emotions, feelings, or any form of subjective awareness.
Conversely, a system might theoretically possess some form of subjective experience without demonstrating human-level general intelligence.
The distinction can be summarized quite simply:
- Artificial Intelligence (AI): Systems capable of performing tasks associated with human intelligence.
- Artificial General Intelligence (AGI): Systems capable of demonstrating broad, adaptable intellectual abilities across many different disciplines.
- Sentient AI: Artificial systems that genuinely possess subjective experiences or the capacity to feel.
- Artificial Superintelligence (ASI): Hypothetical AI that substantially exceeds human intellectual capabilities across a broad range of domains.
Notice that these concepts are not necessarily sequential stages of development.
AGI does not automatically lead to sentience, and superintelligence does not automatically imply consciousness.
It is entirely possible, at least conceptually, that we could develop systems vastly more intelligent than humans that experience absolutely nothing.
Intelligence is about what a machine can do. Sentience is about whether there is anything it actually feels like to be that machine.
For me, this is perhaps the most important distinction in the entire conversation.
Can AI Become Conscious?
This question takes us beyond computer science and into neuroscience, philosophy, psychology, and our understanding of human consciousness.
The difficulty is that we still do not fully understand how subjective consciousness arises in humans.
We understand a great deal about neurons, brain activity, sensory processing, and the relationship between different areas of the brain. However, explaining precisely why physical processes produce subjective experiences remains an unresolved scientific and philosophical problem.
If we cannot fully explain how consciousness emerges in biological systems, determining whether something entirely different could become conscious presents an enormous challenge.
Some researchers believe consciousness may depend on particular kinds of information processing or functional organization. Under these theories, artificial consciousness may be possible if a system develops the necessary characteristics.
Others believe biological processes, embodiment, or mechanisms not yet understood may be essential.
Neither position has been conclusively established.
Research published in Trends in Cognitive Sciences in 2026 proposes evaluating potential indicators of consciousness based on established neuroscientific theories. This is an important direction because it attempts to move the conversation beyond simply asking AI systems whether they are conscious.
After all, an AI telling us that it feels happy does not prove that happiness is being experienced.
Likewise, a system stating that it is not conscious does not independently settle the scientific question.
We need more than convincing language to determine whether subjective experiences genuinely exist.
Why AI Can Seem Human Without Being Sentient
Anyone who has spent considerable time interacting with modern AI systems will recognize how natural these conversations can become.
AI can express sympathy, recognize emotional language, remember conversational context when memory features are available, and adapt its responses to different situations.
At times, it can genuinely feel like we are speaking with another person.
However, natural language is not reliable evidence of subjective experience.
Modern language models learn complex patterns and relationships from large quantities of information. Their responses can demonstrate sophisticated reasoning and context sensitivity, but this does not establish that they possess feelings or an internal conscious perspective.
The distinction matters because humans naturally attribute intentions and emotions to things that behave in familiar ways.
We already do this with animals, fictional characters, and sometimes even ordinary computers when they frustrate us.
As AI becomes more convincing, distinguishing between simulated emotional behavior and actual subjective experience will become increasingly difficult.
This is not necessarily evidence that artificial consciousness is impossible. It simply means that convincing behavior alone is insufficient to establish its existence.
How Close Are We to AGI?
Predicting the arrival of AGI is difficult, partly because experts disagree on its definition and partly because technological progress is rarely predictable.
Some researchers and technology leaders believe broadly capable AGI systems could emerge within the next few years. Others believe significant scientific breakthroughs are still required.
What we can establish is that AI capabilities have advanced considerably.
Stanford University’s 2026 AI Index Report documents continued improvements across scientific reasoning, mathematics, coding, and other technical benchmarks. Some frontier systems now match or exceed human baselines on selected challenging tasks.
However, these achievements should not be confused with reliable general intelligence.
A system can outperform humans on a demanding benchmark while still struggling with tasks that require adaptability, common sense, or sustained independent decision-making.
The September 2026 AI Impacts survey report, based on responses from 1,580 AI researchers collected in December 2024, provides an interesting perspective.
Its aggregate forecast placed a 50% probability on high-level machine intelligence arriving by approximately 2042, compared with 2047 in the previous survey. The survey used a particularly demanding definition involving machines performing essentially all tasks more effectively and economically than humans.
These predictions are not guarantees, and they do not establish when a universally accepted form of AGI will emerge.
There is also substantial skepticism about whether simply building larger versions of existing systems will be sufficient. A 2025 Association for the Advancement of Artificial Intelligence (AAAI) report found that 76% of respondents considered scaling current AI approaches unlikely or very unlikely to produce AGI.
My view is that AGI is a realistic possibility within the next 5 to 10 years, perhaps sooner under some definitions, but we should be cautious about treating any particular date as inevitable.
We might experience increasingly capable systems that automate substantial amounts of intellectual work long before researchers agree that AGI has actually been achieved.
How Close Are We to Sentient AI?
Sentient AI is considerably harder to predict.
There is currently no scientifically confirmed example of a sentient artificial intelligence system, and we do not have a universally accepted test that could establish when one has emerged.
This creates a fundamental problem with predicting timelines.
For AGI, we can at least evaluate performance through tasks, reasoning capabilities, adaptability, and reliability.
For sentience, we are attempting to measure something subjective, potentially without knowing exactly which mechanisms produce it.
We could conceivably develop AGI in the 2030s or 2040s without creating anything that experiences consciousness.
Alternatively, future research might reveal that certain artificial architectures already possess characteristics associated with consciousness, forcing us to reconsider our assumptions.
I would be extremely cautious about anyone claiming that sentient AI will arrive within five, ten, or even twenty years.
It may take decades. It may require fundamentally different computing architectures. It may prove impossible using existing approaches.
We simply do not know.
What concerns me more is the possibility that increasingly intelligent systems will appear conscious long before we develop reliable ways to evaluate that claim.
That uncertainty could create significant ethical and social challenges.
What Happens If AI Actually Becomes Sentient?
If artificial sentience were ever established, it would introduce questions far beyond technology and productivity.
We would need to reconsider how we define consciousness, intelligence, and potentially even moral responsibility.
If a machine could genuinely experience suffering, would we have an obligation to prevent it? If it developed preferences or self-awareness, would switching it off raise ethical concerns?
Would sentient systems deserve some form of protection, and who would decide what those protections should look like?
These are not necessarily questions we need to resolve for today’s AI systems, but they are worth exploring before the technology potentially forces us to confront them.
Importantly, intelligence alone should not automatically imply moral status. A machine capable of solving complicated scientific problems is not necessarily a being capable of experiencing harm.
Likewise, the ability to simulate emotional responses does not establish that a system deserves the same considerations as a conscious individual.
Any future decisions would need to be grounded in serious scientific research, rather than emotional attachment to convincing AI personalities.
What Should We Actually Be Preparing For?
Much of the public discussion about AI focuses on dramatic scenarios involving machines becoming conscious, surpassing humanity, or deciding that humans are no longer necessary.
While these questions make interesting discussions, I believe there are more immediate issues deserving our attention.
AI systems are already transforming software development, cybersecurity, education, business operations, and the way people interact with information.
Organizations need to understand how these technologies are being used, what information they can access, and what decisions they should be permitted to make.
Governance, privacy, security, reliability, and accountability are not problems that can simply be postponed until AGI arrives.
The challenge is making sure that increasingly capable systems remain useful, understandable, appropriately controlled, and aligned with the responsibilities we place upon them.
For people working in technology, that means continuing to develop our own understanding rather than assuming AI will eventually replace the need for technical knowledge.
In many ways, the ability to question AI-generated information, identify mistakes, and understand the consequences of automated decisions will become even more important.
Intelligence Is Not the Same as Being Alive
Artificial Intelligence has already changed how we work, learn, and create. I suspect the next decade will bring developments that make today’s systems seem relatively primitive.
We are moving toward AI that can reason more effectively, work autonomously for longer periods, and perform tasks that previously required substantial human expertise.
Whether these developments eventually satisfy the definition of AGI remains uncertain, but I believe we are much closer to broadly capable artificial intelligence than we were even five years ago.
Sentient AI is an entirely different question.
Creating something that can think, reason, and solve problems does not necessarily mean creating something that can feel, experience, or understand its own existence.
We may eventually discover that intelligence and consciousness are more closely connected than we currently understand. Equally, we may discover that machines can achieve extraordinary intelligence without ever possessing anything resembling human experience.
For now, the most important thing is to understand the distinction and avoid confusing impressive technological capabilities with evidence of consciousness.
I am excited about what lies ahead, particularly the opportunities AI presents for scientific discovery, education, and solving complex problems. However, I also believe we need to approach these developments with curiosity, responsibility, and a healthy degree of skepticism.
Because the question is no longer simply whether machines can become intelligent.
The much bigger question is whether intelligence alone is enough to make something conscious, and whether we would even recognize consciousness if we managed to create it.
That, to me, is where the future of artificial intelligence becomes truly fascinating.