What Is Artificial General Intelligence and Why It Matters to Business
Artificial general intelligence (AGI) represents what many consider the Holy Grail of AI: machines that can understand, learn and apply knowledge across different situations and tasks, at or exceeding the human level.
Today’s AI systems consist of what’s called “narrow AI,” which excels in specific tasks. Deviate from the task and the AI system stops working. This is why Google DeepMind’s AlphaGo can defeat a human world champion in the game Go but cannot do simple tasks outside the game. Human beings, on the other hand, can do many different things: Talk, walk, eat, draw, sing, write, cook and more without breaking a sweat.
In business, narrow AI performs tasks that run the gamut from fraud detection in financial transactions, an AI assistant that composes written content, or an image generation AI model that can create illustrations for ads. However, these AI models will not suddenly be able to do unrelated tasks like following up on sales calls or discerning which prospect is more likely to buy the product.
AGI would change the game completely. It would possess human-like general problem-solving abilities and cognitive flexibility. Just as a human who learns to cook can apply those organizational and timing skills to project management, an AGI system could take lessons from one domain and apply them to completely different challenges. This adaptability is what makes AGI such a transformative concept for business.
“AGI is a theoretical pursuit to develop AI systems that possess autonomous self-control, a reasonable degree of self-understanding, and the ability to learn new skills. It can solve complex problems in settings and contexts that were not taught to it at the time of its creation. AGI with human abilities remains a theoretical concept and research goal,” according to AWS. It notes that AGI is also called “strong AI” and narrow AI is referred to as “weak AI.”
Everyone Is Aiming for AGI
Most of the major tech players in AI have this goal of developing AGI: OpenAI, Google, Meta, Anthropic and others. OpenAI CEO Sam Altman fanned the flames on Jan. 5 by writing in his personal blog that the company is “now confident we know how to build AGI as we have traditionally understood it.”
In an interview with Bloomberg, Altman said its ‘o3’ AI model passed ARC-AGI, a key AGI threshold, meaning it matched humans in doing unknown tasks. However, o3 was trained on the ARC-AGI public dataset.
“OpenAI’s new o3 model represents a significant leap forward in AI’s ability to adapt to novel tasks,” according to Arc Prize, a public competition to beat the ARC-AGI test that is co-founded by the creator of ARC-AGI, Francois Chollet. “This is not merely incremental improvement, but a genuine breakthrough, marking a qualitative shift in AI capabilities compared to the prior limitations of LLMs [large language models]. O3 is a system capable of adapting to tasks it has never encountered before” and approaching human-level performance.
But Arc Prize cautioned that its ARC-AGI test is “not an acid test for AGI” and passing it “does not equate to achieving AGI.” That means o3 has not achieved AGI because it “still fails on some very easy tasks, indicating fundamental differences with human intelligence.”
Meta’s Chief AI Scientist Yann LeCun, one of the so-called “godfathers of AI,” recently argued something similar: AI systems are not even smarter than a cat or dog. While Meta CEO Mark Zuckerberg has stated that the social media giant is gunning for AGI, LeCun said no one has achieved it yet.
Achieving AGI would have big implications for business. For example, one AI system can analyze market trends while simultaneously redesigning the supply chain to adapt to those changes. It can handle customer service while using those interactions to inform product development. It can manage operations while developing innovative solutions to efficiency problems. It can make strategic decisions by synthesizing information across multiple industries and domains, which requires high-level general reasoning.
For now, AI systems with “such generality” in capabilities come at a “steep cost,” according to ARC Prize. It said that AGI costs around $17 to $20 per task compared with $5 to pay a human being to do the same thing while consuming “mere cents” in energy. However, it believes the costs of AGI should fall and become competitive with human work within a “fairly short” time.
The Levels of AGI
One of the sticklers for gauging whether or not AGI has been achieved is its fluid definition. That’s why Google DeepMind made a stab at defining AGI by creating the six levels of AGI against which AI systems can be measured.
This framework aims to organize and ground in metrics what has been a fluid definition of AGI. The levels are based on the performance and generality of the AI model’s capabilities. They range from Level 0 (no AI) to Level 5 (superhuman), which it also calls artificial superintelligence.
OpenAI has its own five levels of AGI, which Altman talked about during an interview with PodiumVC. It was first reported by Bloomberg.
Level 1: Chatbots, AI with conversational language
Level 2: Reasoners, human-level problem-solving
Level 3: Agents, systems that can take actions
Level 4: Innovators, AI that can aid in invention
Level 5: Organizations, AI that can do the work of an organization
Notably, the development of AGI relies on several key technological components: Machine learning architectures need to evolve beyond pattern recognition to true understanding; systems need to develop causal reasoning — understanding not just that things happen, but why they happen. They also need common sense reasoning and the ability to transfer knowledge between different contexts. Most importantly, they need to develop something akin to human consciousness and self-awareness, although the topic of machine consciousness is controversial and has even led to the firing of a Google engineer who claimed an AI model was sentient.
According to IBM, AI needs to improve on the following capabilities in order to reach AGI:
- Visual perception — Current computer vision capabilities remain no match for human capabilities in object detection and facial recognition. AI systems struggle with partially hidden objects, for example.
- Audio perception — AI systems still struggle with understanding accents, sarcasm and other emotions.
- Fine motor skills — Robotic AI systems still struggle to master fragile objects, use tools in real world settings and adapt to new tasks.
- Problem-solving — AI systems need both reasoning and critical thinking skills to solve problems like a human being.
- Navigation — Autonomous vehicles still face challenges in understanding the outside world, something a teenage driver would have no problems navigating.
- Creativity — AI systems copy but are not originally creative.
- Social and emotional engagement — AI systems today still have trouble recognizing emotions, intepreting facial expressions, tone of voice and body language.
Solving AGI can be a game-changer for business since it can revolutionize every aspect of operations. For example, it could optimize processes in ways humans can’t imagine, identify market opportunities human leaders can’t see, and solve complex problems with unprecedented speed and accuracy.
However, the challenges are significant as well. The development of AGI would necessitate major economic and social changes. Industries could be transformed overnight. Jobs would be created and eliminated at unprecedented rates. Companies would need to completely rethink their organizational structures and business models.
On a societal level, the race to develop AGI has even led to fears by AI pioneers that AI could go rogue — and even wipe out humanity. Such alarm bells were rung by no other than Nobel laureate Geoffrey Hinton, one of the godfathers of AI. He quit his job at Google to warn society about his fears — appearing in many national news programs and speaking to the MIT Technology Review.
In 2023, AI luminaries including Yoshua Bengio (another godfather of AI), Stuart Russell, Elon Musk, Apple co-founder Steve Wozniak and others signed an open letter calling for a pause in AI development.
On the other side of the debate are people like LeCun and Andrew Ng, the founder of Google Brain, who believe that such fears are overblown and AI will represent a renaissance of society and business. Meanwhile, the race towards AGI continues. While no one knows when AGI will be achieved, its wide impact makes it a technology that cannot be ignored.
The post What Is Artificial General Intelligence and Why It Matters to Business appeared first on PYMNTS.com.