Artificial General Intelligence: The Next Frontier In Technology For an AI to be classified as AGI, it needs to have certain characteristics such as common sense, background knowledge, transfer learning, abstraction, and causality. Vision from the Marvel Cinematic Universe is the aptest example of what AGI can be like
Opinions expressed by Entrepreneur contributors are their own.
You're reading Entrepreneur India, an international franchise of Entrepreneur Media.
In 1950, British computer scientist Alan Turing proposed an imitation game, the Turing Test, to determine if a computer possessed the capabilities to think or mimic a human. Seventy-three years later, humans are seeking answers and having a heart-to-heart conversation with an artificial intelligence named ChatGPT. Why? Because it possesses the ability to rebuff a user's improper requests, accept its mistakes, answer follow-up questions, and carry out a full-fledged conversation. The creators of ChatGPT in haste, have also launched GPT-4, with Sam Altman, CEO, OpenAI, calling it "our most capable and aligned model yet; it is still flawed, still limited." So, what next?
What is artificial general intelligence?
A hypothetical-theoretical form of AI, artificial general intelligence refers to machines with intelligence which can mimic that of humans and understand and perform intellectual tasks and carry out a wide range of activities. For an AI to be classified as AGI, it needs to have certain characteristics such as common sense, background knowledge, transfer learning, abstraction, and causality. Vision from the Marvel Cinematic Universe is the aptest example of what AGI can be like.
The current stage
Artificial intelligence can be categorized into three- artificial narrow intelligence, artificial general intelligence and artificial superintelligence. Currently, we are operating at ANI level with Rodney Brooks, an MIT roboticist and co-founder of iRobot, believing AGI will not take shape at least before 2300. AGI may take years, but today's AI advance pace is a good one, with generative AI being the most popular one. "Generative AI, with its ability to create and automate content and tasks, will generate greater inclusivity and access when it comes to business, education, job opportunities and healthcare, creating a world where we can graduate to a post-scarcity global economic regime," shared Sunil Gopinath, CEO, Rakuten India. Generative AIs include the likes of Dall-E 2, Midjourney, Deep Dream Generator, and Big Sleep.
The opportunity and market players
According to industry reports, the global AGI market is expected to be valued at approximately USD 144.2 billion by 2026, with a CAGR of 41.6%. AGI means a machine which is at par with humans to carry out tasks and activities. Do acknowledge that as no AGI system or technology doesn't exist, all the possibilities of it, such as helping perform mundane activities of humans, remain a piece of hypothesis.
Players working in the AGI space include OpenAI, AGI Innovations Inc, Apprente, DeepMind, and New Sapience. Building an advanced AI calls for heavy investment and high technology infrastructure, and hence you'll find few names across the media being attached to the developments. So, is the AI space just for big players? "In my opinion, big and small players will co-exist and play an equitable role in the AI space— you could call it division of labor. In the value chain, startups will continue to focus on idea generation and prototype building, while incumbents will be involved in startup acceleration, integration of ideas and solutions, marketing, and building economies of scale," shares Arpit Sharma, Senior Manager, Technology Research and Advisory, Aranca.
However, Sachin Dev Duggal, Co-founder & CEO, Builder.ai. feels otherwise, "the power of large voices in AI can drown out small ones, potentially leading to a homogenized global view that ignores local perspectives and cultures. The control of AI technology by a small number of companies and locations raises questions about national sovereignty and the potential for misuse of this technology."
The future towards AGI
The biggest challenge or concern in developing an AGI is its value and ethical understanding. "Think of the current stage of AI development as a new-born child without a clear value system, with the potential for rapid and significant progress. However, there are significant challenges that need to be addressed. This includes the lack of human fabric in machine-human conversations, which can lead to biased algorithms and lost context across different demographics," shares Duggal.
One of the notable critiques was philosopher Hubert Dreyfus who argued machines could never acquire intelligence since they had no body, no childhood and no cultural practice. Human knowledge cannot be incorporated into a computer. Researchers argue that AGI can be developed through large-scale models, datasets and computing power. As AI keeps improving, AGI can be achieved in the near future. But it also poses an existential threat to the human race as it can surpass human intelligence and evolve into what is known as a superintelligence (ASI).
McKinsey and Company, in its 2020 report titled "An executive primer on artificial general intelligence" shared four ways to measure the progress of an AI becoming an AGI. This included 1. the object-recognition capabilities of a two-year-old, 2. the language-understanding capabilities of a four-year-old, 3. the manual dexterity of a six-year-old, and 4. the social understanding of an eight-year-old. This method is said to be a replacement for the Turing Test.
ChatGPT, Google's BARD and its echoing platforms and tools might feel like an AGI due to its advanced capabilities, but it is a language model built over a large dataset. In March, a petition was launched 'Pause Giant AI Experiments: An Open Letter' which has garnered over 27,565 signatures called for at least a six-month ban of AI systems stronger than GPT-4. Recently, US President Joe Biden called for a meeting with key players of the AI industry including Microsoft, Google and OpenAI to talk about product safety and potential risks.
GPT-4 is taking us one step towards achieving AGI, and it will arrive much before Brooks' expected timeline. It can be 2050, 2030 or even 2027.