What Is Artificial General Intelligence (AGI)?

Artificial general intelligence (AGI) is a hypothetical type of AI that would match the type of cognitive abilities that humans have, which would allow it to learn and apply its knowledge across many different tasks, rather than being focused on a narrow and specific area.

AGI is both exciting and a little frightening. Step into a future shaped by the power of AI and you’ll find yourself in a world populated with the darkest sci-fi nightmares‌ — ‌think Terminator, HAL and a host of others. It’s a future that often evokes a deep-seated fear.

Indeed, this apprehension is not unfounded. Legitimate concerns abound about safety, security and trustworthiness of AI, including the erosion of privacy and control, the proliferation of hyper-realistic fake information on an unprecedented scale, new legal and psychological challenges, the spectre of job displacement and the possibility of runaway AIs unleashing apocalyptic scenarios upon us all.

The article answers these questions and more:

• What is Artificial General Intelligence (AGI)?
• What’s needed in order to reach true AGI?
• How will AGI affect the way we live?
• How will AGI affect the workplace?
• What guardrails are needed to ensure safe AGI?

To get a sense of AGI, imagine the cognitive capabilities of a human — ‌with the ability to learn, converse, plan and perform tasks across a range of domains, from the scientific to the creative to the philosophical to the personal — combined with speed, memory, reliability and scalability of high-performance computing.

This is what separates AGI from today’s AI, often called artificial narrow intelligence (ANI). Current AI systems excel at specific tasks, such as playing chess or recognising faces, but AGI would be able to perform any intellectual task a human can, with the same flexibility and adaptability.

Today’s large language models (LLMs) powering AI models provide a glimpse of AGI, with the ability to carry on open-ended conversations with natural language, coupled with their access to vast shops of knowledge in just about every domain.

While AGI isn’t yet a reality, it’s worth discussing how we use AI today, as those decisions will shape how we manage technology in the future.

Recent Salesforce Research finds that as brands increasingly adopt AI to increase efficiency and meet increasing customer expectations, nearly three quarters of their customers are concerned about unethical use of the technology.

We believe that a much better future is possible with AI if we make the right choices in the near future, especially as AI grows more powerful and even incomprehensible to humans. We can already see the extraordinary potential of artificial general intelligence (AGI), which will be like nothing that came before it.

Already, they are beating human experts on a range of knowledge and problem-solving tasks across a range of subjects, from math and physics to law and medicine.

In the best-case scenario, with the appropriate guardrails, AGI promises to turn everything we value about intelligence — the ability to understand context, to solve novel problems and to plan for the future — into “intelligence on-demand”, a resource anyone can use at any time.

Any routine human cognitive activity will eventually be aided by or replaced by an AI. All routine business operations will be orchestrated by AIs and individuals will integrate personal AGIs so seamlessly into their daily lives that it will become unimaginable to exist without them.

However, even the most advanced LLMs today remain a significant distance from the promise of true AGI. Human learning sets a high bar. Humans with a lifetime of knowledge can permanently extend their understanding of the world with as little as a single experience or conversation, from simple facts to entire physical tasks. LLMs are still largely limited to the contents of their training data and require costly and time-consuming retraining sessions when that content needs to be updated or corrected.

The ability to learn in real time will be a key feature of AGI, allowing such systems to continuously augment and refine the data on which they’re originally trained as they explore the world. AGI will be capable of extracting meaningful data from virtually all of its experiences, including exchanges with humans, sensory input from its surroundings and, of course, information found online and via other networked sources.

Additionally and equally importantly, AGI will be defined by its ability to judge what new information to seek out in the first place and when. Today’s LLMs are entirely dependent on human developers to learn, bombarded with information like a student in a classroom at the mercy of the teacher.

AGI will be more like an ambitious adult, reacting to mistakes, confusion, or even some innately cultivated sense of curiosity with a continual desire to expand its knowledge base. Such AGI won’t just be more resilient and resourceful, but will truly grow in its capabilities and value over time, driven largely by its own appreciation for the value of knowledge.

A key aspect of human relationships and communication is our ability to anticipate, understand and even model the nature of our fellow humans’ minds. Whether it’s our ability to guess the next move of the driver in an adjacent lane, the needs of a customer, coworker or friend or the strategy of a competitor in a game, much of our time each day is spent considering the contents of minds other than our own.

Today’s LLMs are still only accidentally empathetic. Although LLMs can be said to demonstrate faint flickers of this level of comprehension as an embedded feature of their training data, it’s either explicit or robust in its definition. No serious attempt has yet been undertaken to endow LLMs with the kind of empathy and understanding that defines humans at their most perceptive.

AGI will be expected to connect with humans at a deeper, more meaningful level by leveraging an understanding of the needs and psychology of humans as both individuals and groups, the social conventions that unite us and, of course, the simple fact that we are physical, mortal beings — and all that entails — rather than interchangeable digital entities.

Another boundary that AGI will need to cross is embodiment. In the simplest forms, embodiment refers to engagement with a physical or virtual world in a way that mirrors a human’s experiential understanding of that world.

Our senses deliver a kind of “data” loosely akin to text, images, or sounds, and we also integrate them from one moment to the next, suffusing each with meaning that goes far beyond any particular detail.

When we learn a new task, for instance, such as a better way to cut a vegetable on a cutting board, we integrate the tactile experience of the vegetable’s skin against our fingers, the feeling of the knife handle in our hand, the concern about the blade as it gets closer to our skin and visuals of everything we see and even the sounds and scents of the slices piling up.

Additionally, we integrate these moments with concepts that are rarely directly captured by today’s datasets — an understanding of causality (how each step in the process must be completed before the next can happen), an understanding of physics and materials (how much force it takes to break the skin of the vegetable with a knife or the friction the blade faces once it hits the cutting board), and even human psychology (why cooking is a worthwhile pursuit).

Today’s AI is only dimly aware of these concepts. Embodiment represents a major chasm that must be crossed to achieve AGI.

When one imagines AGI ability applied throughout our lives, the applications are, quite possibly, limitless. AGI will perform many tasks that humans already do with unparalleled efficiency and those far beyond human capabilities.

Here are some ways that AGI might be used:

• Acting as an everyday copilot, providing guidance and assistance in various tasks from shopping to party planning
• Offering expert-level understanding of diet, fitness, medication and disease mechanisms, leading to effective diagnosis and therapies
• Assisting healthcare workers by handling administrative and logistical tasks, reducing their workload and minimising medical errors
• Entertaining with not only recommendations for films, music and books, but also creating entirely new works based on preferences
• Serving as a career collaborator, mentor and problem-solver and even contributing to the development of new products and industries

True AGI would not only be able to productively contribute to any such conversation‌ — ‌and perhaps even consistently solve our problems given enough time and information — but make the reasoning behind its “thinking” clear, even tailoring the nature of that conversation to the expectations of the user.

Imagine a kind of “everyday AGI copilot,” ready to provide whatever input or guidance its user might need and take some of the usual headaches out of life. They will take care of shopping, plan a party, make a financial decision, or mediate a disagreement with a friend.

They will also be present in the home, work, school, in transit, in every environment in which you operate. AI will engage with vast networks of AIs from those environments to enable local services and capabilities.

AGI promises to transform our relationship with our health as well, bringing expert-level understanding of diet, fitness, medication, and general life practices to everyone capable of accessing it.

Potentially, AGI could understand the mechanisms of disease so comprehensively that it can offer effective diagnosis and therapies to treat and cure them. Imagine real-time health monitoring and detection, such as sensing the earliest cancer cells dividing and recommending genome-specific therapies to eliminate the disease.

For doctors and other healthcare workers, all of the administrative and logistical work will be done by AIs, freeing at least a day a week for every worker. Medical errors are common and AIs are likely to be one of the best tools for helping medical workers avoid them, like assessing test data, drug dosage and continuously updating the workers knowledge.

Imagine an intensive care unit where humans and AIs collaborate in full-time patient monitoring and testing or an operating theatre where humans and AIs collaborate, often carrying out surgical tasks that require extreme levels of control as in brain surgery.

For entertainment, AGI won’t just be able to suggest films, music and books we’re likely to enjoy‍ — ‍something today’s AI models often do quite well — but would be able to literally create such works entirely from scratch, based on the precise preferences and desires of the user.

Advances in both text and video generation are already unfolding at an impressive pace, which raises an obvious question‍ — ‍will future AI models be able to invent entire storylines, translate them into detailed shooting scripts with characters, dialogue, transitions and the like, then generate feature-length, ultra-HD video footage to bring it to life?

Imagine an expertly-crafted sequel to your favourite film, book or even video game, allowing you to return to a beloved fictional world as often as you want to find new adventures waiting for you. And even this, as ambitious as it sounds, is just the beginning of the ways AI might entertain us in ways currently unimaginable. When such capabilities are combined with virtual and augmented reality, the vision grows even broader.

Although it’s hard to conceive today, AGI agents would make for such great conversation partners and creative collaborators that, over time, we’d come to view them as peers and maybe even friends.

As AGI develops, its place in our world may soon be as social as it is practical. Considering that humans have developed unexpectedly deep bonds with machines as far back as the ELIZA chatbot in the 1960s‍ — ‍about as simple and limited a conversation agent as can be imagined — we shouldn’t be surprised by the possibility of superhuman intelligence becoming a captivating companion.

AGI will surely deliver similarly transformative capabilities to reshape the way we work and redefine productivity. Its reasoning abilities, access to data, and conversational skills will make it a natural collaborator and even mentor for almost any kind of career.

Whether it’s simply gut-checking a decision or asking for a complete strategy to accomplish a task, AGI may eventually become the ideal partner in getting anything done. Of course, augmenting human workers is only the beginning. AGI could extend to the development of entirely new products, such as the next great basketball shoe, smartphone, or refrigerator, and entirely new product categories.

Just as AGI will likely help us all remain healthier as individuals, it promises a similar boost in the health and stability of organisations.

Imagine bringing expert-level intelligence‍ — ‍and, likely, far beyond — to every moment of every day of a business’s life, ensuring that problems are anticipated as far in advance as possible and addressed in the very first moment action can be taken.

AGI will be able to actively manage large, complex networks, such as urban traffic flow, global supply chains, and energy grids, adapting in real time to the real world. In the chemical and biological sciences, AGI will create AI alchemists with the ability to control vast numbers of molecules and design and produce novel materials such as proteins for biotechnology and medical research and crystal structures that lead to entirely new alloys and materials.

AGI will also accelerate technology to create more powerful, intelligent robots. Whether that takes the form of Hollywood-style bipedal robots or something else entirely, it’s all but certain that AGI wouldn’t just change knowledge work, but labour as we know it.

Highly-efficient machines at warehouses, factories, farms and other facilities will take the baton from human operators and unleash productivity and throughput never before possible, all while reducing accidents, mistakes and defects. At some point AGI might be able to encompass all of the mammalian senses — sight, sound, smell, taste and touch — creating an entirely new robotic species.

In addition, connecting the human brain to AI systems that can read signals directly from the brain has enormous potential for a variety of tasks. Neural prosthetics will improve the functions of the brain, such as memory loss or the damage caused by a stroke, and AI-enabled limbs would bridge humans and robotics. VR and AR experiences could be more immersive and intuitive by directly tapping into the user’s neural response. And, eventually, the human brain and AGIs could form a more symbiotic relationship that becomes a new species.

Salesforce ProGen — an AI language model trained on the largest protein database available — shows how generative AI can lead to potential solutions for addressing challenges in human disease and the environment.

Throughout history, technological advancements have consistently yielded both beneficial and adverse outcomes. We’ve shared what we foresee as the benefits of AGI for humanity. We know that bad actors will try to exploit the technology for harm, with autonomous weapons, invasive surveillance, economic manipulation and cyberattacks and disinformation campaigns on an unprecedented scale. AGI will reflect the best and worst of us.

The future of AGI depends on building safeguards to make sure that AI systems do not make wrong decisions, are competent at their assigned task, and are more immune to cyberattacks, where bad actors can take control or hijack the AI systems and perform hostile actions.

It’s also coming up with legal frameworks to account for mistakes (similar to autonomous driving — if the car is in an accident, determining who is liable — the owner of the car or the company that owns the AI).

The way forward with AGI is to make trust the number one priority, which is a significant challenge given AGI will perform tasks with less human supervision (ideally none). This is a pivotal moment, the time that historians will look back on to assess whether we leveraged AI to make progress on issues like trust, equality, health and climate change or allowed technology to deepen our crises.

We will need to take steps — through human oversight, ethical frameworks and global cooperation — ‍to ensure this revolutionary technology is aligned with human values and that the benefits serve all of humanity.