AI Singularity: 3-Month Prediction?

AI Singularity Predictions

Since large language models (LLMs) have become widespread, experts have shared varied opinions on AI's future. Some researchers believe the singularity, where machines surpass human intelligence, could happen within decades. However, the CEO of Anthropic thinks it could be as close as three months away.

Analyzing AI Predictions

A new analysis from AIMultiple examined 8,590 predictions from scientists, entrepreneurs, and others to understand the different AI forecasts. The study tracked how these predictions have changed. While it considered different AI thresholds like artificial general intelligence (AGI) and AI superintelligence, industry leaders were generally more optimistic. Most believed AGI would likely occur within fifty years. The arrival of LLMs has shifted timelines for AGI and the singularity.

The report notes that current surveys predict AGI around 2040, whereas pre-LLM surveys predicted it around 2060. Entrepreneurs forecast it even sooner, around 2030.

The Inevitability of AGI

The analysis suggests reasons why many believe AGI is inevitable. Unlike human intelligence, machine intelligence may not have any limits. Computing power doubles roughly every 18 months—Moore's Law—so LLMs could soon match human intelligence in calculations per second. If computing faces limitations, quantum computing might compensate.

The report states that while most experts believe Moore’s law is ending, quantum computing could efficiently train neural networks and help AI algorithms achieve singularity.

Is AGI a Certainty?

However, not all experts are convinced AGI is a certainty. Some argue that human intelligence is more complex than current AGI definitions suggest. They believe human intelligence includes multiple facets. Deep learning pioneer Yann LeCun suggests rebranding AGI as “advanced machine intelligence,” arguing that human intelligence is too specialized to replicate.

The report also implies that AI is a useful tool for discovery, but cannot make discoveries independently. AI can improve experiment design and management, but even the best AI might not be able to cure cancer.

Although predictions vary, the consensus is that algorithms will cause incredible change. Whether these changes are beneficial is up to us.