AGI Progress could be more rapid than the larger society expects. Society is rapidly accelerating its adaptation to Artificial Intelligence advancements. Based on this, Sam Altman, CEO of OpenAI, has made a provocative forecast.
The emergence of Artificial General Intelligence (AGI) may be merely a few years away—potentially as early as 2025. This revised timeline diverges sharply from previous expert estimates. As a result, experts are reevaluating the proximity to a groundbreaking AI milestone. One where machines could rival, or even outperform, human cognitive abilities across virtually all spheres.
What is Artificial General Intelligence (AGI)?
Artificial General Intelligence (AGI) has always been the most cherished goal of AI. The level of achievement par excellence, encompassing cognitive skills that match, or even excel, humans’ across almost all domains. 
Altman had historically forecast AGI progress to remain elusive for at least another decade. But he has revised his outlook: AGI is no longer a distant aspiration. It is an imminent engineering problem with a solution he thinks may come much sooner than some may think.
Researchers, including those at OpenAI, largely understand the foundational steps and technologies required to achieve AGI. Despite the clarity on the route, the exact timeline, challenges, and the final form of AGI might still be subject to significant uncertainty.
The progress in AI research, particularly at OpenAI, is outpacing previous predictions or benchmarks. Despite advancements, there’s ongoing debate or uncertainty among experts. They seem to disagree about the comprehensive definition, capabilities, or requirements of a truly “general” intelligence.
Importance of Achieving AGI
Achieving Artificial General Intelligence (AGI) could be the most transformative milestone in human history. This could hold immense importance for society and the economy.
AGI has the potential to surpass human intelligence in nearly all domains. This could revolutionize healthcare by discovering cures for previously incurable diseases and enhance global food security through optimized agricultural practices. It could help to solve complex environmental challenges such as climate change.
Economically, AGI Progress could automate mundane and dangerous jobs. This alone would lead to increased productivity, saving hours. Humans would compress months into hours and days of productivity. On a personal level, AGI could assist in personalized education and vastly improve accessibility for people with disabilities. It could also extend the human lifespan through advanced biotechnological interventions.
However, it also raises critical questions about job displacement and ethical decision-making. It also highlights the need for stringent regulatory frameworks to ensure that the benefits of AGI are equitably distributed and its risks mitigated.
The Path to AGI Progress
Altman states that the route to AGI Progress is “mostly clear,” with OpenAI advancing faster than expected. However, what AGI fully entails is still uncertain. Some experts say that AGI should mimic human knowledge and learn, adapt, and handle tasks beyond its original programming and training data. This means AI could create new solutions and insights rather than being limited to existing information.
While Altman is optimistic about AGI’s arrival, recent benchmarks show that AI systems still struggle with basic reasoning tasks. One of these benchmarks is FrontierMath. In tests measuring how well models handle new problems, even advanced models like GPT-4 and Google’s Gemini 1.5 Pro solved less than 2% of the issues presented. This indicates that, by some definitions, current AI models are far from achieving AGI Progress, especially in autonomous reasoning and generating solutions beyond their initial datasets.
However, insiders at OpenAI suggest that the upcoming version of GPT-4 (codenamed o1) shows a significant improvement in reasoning capabilities. There are also rumors of enhancements in the next generation of Google’s Gemini models, hinting that we may soon see AI systems better equipped for complex problem-solving tasks.
Steps of AGI Development
OpenAI’s approach to AGI is structured into a series of levels, with AGI itself at the final stage. The company outlines five stages of AI development leading to AGI, each marking a significant advancement in capabilities:
Level 1: Basic chatbots that mimic human conversation through simple text generation, similar to models from recent years.
Level 2: “Reasoners” that can handle more complex tasks, like OpenAI’s upcoming o1 model.
Level 3: “Agents” — AI systems that can act autonomously without human input. One such is Google’s rumored “Jarvis” or Anthropic’s AI systems like Claude.
Level 4: “Innovators” — These are AI models that not only perform tasks. They also generate new ideas and creative solutions, pushing beyond human capabilities.
Level 5: AGI — In this category are AI that can reason, learn, and operate at the level of an entire organization, solving problems, innovating, and adapting independently.
Altman believes AGI could be achieved in the next few years and briefly mentions Artificial Superintelligence (ASI)—AI that exceeds human intelligence in all areas. However, he clarified that ASI is still “thousands of days away,” stressing that AGI is a major milestone on the path to more advanced AI.
The distinction between AGI and ASI may blur as AI advances, but their capabilities remain different. AGI is expected to perform at a human level across various tasks, while ASI would surpass that, achieving higher forms of cognition and insight.
The Role of Partnerships in AGI Progress
A major breakthrough accomplishment of this magnitude cannot be achieved alone. Altman’s optimistic timeline for AGI Progress might be influenced by OpenAI’s partnership with Microsoft, which includes a deal that concludes once AGI is reached. This could require Microsoft to renegotiate and potentially increase payments to access OpenAI’s models, raising the stakes for both companies. Reportedly, this agreement is causing strain in their relationship, adding pressure to the race toward AGI.
While the leap to AGI may seem close, the transition will likely be gradual. Just as generative AI has slowly integrated into many industries, AGI will probably emerge incrementally, improving over time until it becomes commonplace. AGI will develop steadily rather than a sudden revolution, incorporating new abilities as they become available.
AGI’s integration could be more subtle than expected, with society slowly adapting to the increasing presence of intelligent systems. At the current pace, we may not even notice when we’ve crossed into the AGI era, as it will likely become part of the infrastructure supporting everything we do.
Wrapping Up
Artificial General Intelligence (AGI) is around the corner and is only the beginning of all things AI. From the insights at our disposal, it will be gradual and nuanced, steeply associated with advancing technology and industry. Therefore, as AI grows, businesses must continue to get themselves up to speed and adjust to the culture in which Artificial Intelligence is integrated into every aspect of life and work.
