As technology becomes increasingly sophisticated, it's helpful to understand the different stages of AI development.
Stage 1: Rule-Based AI
The earliest and most basic form of AI is rule-based systems. These systems operate based on predefined rules and algorithms programmed by developers. A simple example is a chess game AI that operates by knowing all possible moves based on coded rules. However, rule-based systems cannot learn or adapt beyond their programming. They excel at tasks with clear rules like diagnosing mechanical issues or processing forms. But they lack flexibility and cannot handle new scenarios.
Stage 2: Context Awareness and Retention
The next evolution of AI involves understanding and retaining context from previous interactions. Smartphone assistants like Siri demonstrate this by remembering conversational context. If you ask about a sports team and then ask when their next game is, the assistant understands "their" refers to the team. Systems at this stage become more versatile through retaining information.
Stage 3: Domain-Specific Mastery
Stage 3 systems move beyond just retaining context - they become highly skilled experts in specific domains. IBM's Watson and Google’s DeepMind are good examples. Watson excels at answering questions on Jeopardy while DeepMind has mastered complex games like Go. These systems demonstrate sophisticated abilities unmatched by humans in their specialized areas. However, their intelligence remains narrow and domain specific.
Stage 4: Thinking and Reasoning AI
Stage 4 systems attempt to mimic human reasoning. Using techniques like machine learning, these AIs can understand complex concepts, solve unfamiliar problems, generate creative ideas, and make inferences about vague information. Their reasoning capabilities resemble an enhanced human mind specialized for thinking tasks.
Stage 5: Artificial General Intelligence (AGI)
AGI, or strong AI, refers to machines with general intelligence rivaling humans across a broad range of capabilities. AGI would possess qualities like self-awareness, consciousness, and adaptability. It could perform any intellectual task a human can, from learning languages to understanding emotions.
Stage 6: Artificial Superintelligence (ASI)
ASI surpasses human intelligence in most if not all domains. An ASI could potentially innovate, problem solve, and gain insights beyond human capability. However, ASI also raises major concerns around misuse, unintended consequences, and existential risk. The ethics and control of such powerful AI is a source of much debate.
Stage 7: The Singularity
The Singularity refers to a hypothetical point where ASI recursively self-improves beyond human control. Futurists propose this could dramatically accelerate technological growth, leading to unforeseeable societal change. However, the Singularity remains speculative. The advent of ASI and its implications are still unknown.
Current Stage of AI Development
We are currently somewhere between Stage 4 and Stage 5, with a great deal of debate as to when we will actually reach AGI.
The two projections below are from data provided by Dr. Alan Thomson, an AI expert and consultant, advising Fortune 500s and governments on post-2020 large language models. The data is fitted using polynomial regression. In December 2023, using the current milestones and percentages, this method says there’s an 80% probability of hitting AGI by 26/Jan/2025.
Sources:
Comments