AGI, general artificial intelligence could surpass us as early as 2026

The year 2029 may be remembered as the moment humanity met its digital double: artificial general intelligence (AGI). With the advancement of technologies like GPT-4, we are approaching a point where AI could match or even surpass human cognitive abilities. A scenario that raises a fundamental question: how will our lives change?

General artificial intelligence and technological singularity: we are talking about an epochal change

I'll put it simple: artificial general intelligence (AGI) it is a type of artificial intelligence that can learn, understand and apply its knowledge in a wide range of tasks, just like a human does. Basically, it's like having a computer or robot that can think and solve problems in many different areas, not just one specific one. The Technological Singularity it is a concept that conjures images of a future in which artificial general intelligence not only assists humanity, but surpasses it in ways we can only imagine today. This event represents a possible turning point in the history of humanity. The Singularity is not only a technological achievement, but also a moment of deep reflection on our role in a rapidly evolving world.

GPT-4 and the approach to AGI

GPT-4, the latest iteration of OpenAI's generative language models, marked a significant step towards AGI. This technology, which already shows surprisingly advanced language understanding and generation capabilities, is just the tip of the iceberg. His arrival accelerated the path towards a moment which, according to previous forecasts and expectations, was later in the years. challenging previous predictions and expectations. Historically, AI researchers have predicted that the Singularity will occur on average around 2050. Future Studies experts, who periodically review their estimates, are divided into this range:

Vernor Vinge, who coined the term “Technological Singularity” and refined its concepts in the 80s and 90s, predicted this inflection point by 2050, the middle of this century. Ben Goertzel, another leading researcher in the field of AI, between 2000 and 2010 predicted general artificial intelligence (and the technological Singularity) by the end of the 30s of this century. Ray Kurzweil, in 2005, predicted the arrival of the “great change” by 2029. He actually wrote a book about it, “The Singularity is Near”.

And then there's Paul Pallaghy

the AI ​​engineer (and futurologist) Paul Pallaghy from the University of Melbourne has perhaps detailed more than anyone the predictions on the advent of this key moment in the history of our species. In 2012 you drew up a sort of small roadmap that would bring us closer to general artificial intelligence, with all the consequences of the case. He started from an intuition, which later turned out to be correct: quantum computers would not be needed to develop natural language processing (NLU) models, as several colleagues predicted. We could have done it with the computational capabilities of "traditional" computing. It was a prediction that, with hindsight, we can judge as authoritative: Pallaghy was working on those very models at the time.

To be concise, in 2012 these were his predictions:

1. Adequate natural language understanding (NLU) by 2020.
2. Human-like natural language understanding (NLU) by 2022.
3. Artificial General Intelligence (AGI) by 2024.
4. Androids by 2026.
5. The Artificial Intelligence Singularity potentially as early as 2025, but almost certainly by 2029.

He reiterated them several times, until 2021 in which he condensed them in a post on his blog. Many of her colleagues didn't remember them well, for the simple fact that in 2012 they probably thought they were crazy. Today's reality, however, is strikingly similar to the first two points. They are just a year out of phase. Natural NLU? 2021. “Humanoid” NLU (GPT-4)? 2023. Now what?

Pallaghy's revised estimates

• 2023: Humanoid NLU (achieved with GPT-4).
• 2024: Mature NLU and advanced disembodied AGI.
• 2025: Thousands of androids available for less than €100.000.
• 2026: The first potential for a disembodied superintelligent AI Singularity.
• 2027: Millions of androids available for less than €50.000.
• 2028: The first potential for an android-based AI Singularity.
• 2029: Tens of millions of near-superintelligent androids available for around €25.000.

In summary, his central prediction is a Singularity in 2028, which however could occur as early as 2026. It is essential to understand one thing, if you want to grasp the meaning of Pallaghy's predictions. Something that many do not understand, and undermines their ability to evaluate the timing of these phenomena. The general artificial intelligence that could trigger the Singularity, i.e. the hypothetical autonomous acceleration of the evolution of AI, does not require self-awareness. It has no need to be “conscious” to act. It doesn't have to be an "artificial life". It just needs to be a little more advanced than GPT-4 (or perhaps the next iteration, GPT-5), with more long-term and short-term memory and a little more training to achieve appropriate goals.

General artificial intelligence: what consequences?

As we move towards AGI, crucial ethical and social questions emerge. How will we manage an intelligence that will surpass ours in many aspects? What safety measures and ethical guidelines should we implement to ensure its rise is beneficial to humanity? These are not simple theoretical questions: they require concrete answers and immediate actions.

It is essential that we prepare adequately. This means not only advancing technology, but also thinking about the broader implications of AGI, on the positive implications like on the negative ones. We must consider how AGI will affect work, education, ethics, and society as a whole. It is a task that requires the collaboration of scientists, philosophers, legislators, sociologists (especially them!) and citizens.

An uncertain, promising, dangerous future

The road to the Technological Singularity is full of uncertainties, but also immense possibilities. Artificial general intelligence could represent an existential threat or an unprecedented opportunity for humanity. The key is to navigate these scenarios with wisdom, prudence and an ever-keep eye on the broader implications of our technological choices. We could come close to 2026, and in 2029 we could impact an exorbitant future.