Can Artificial Intelligence Attain Human-Level Intelligence? Insights from 5 Experts.
The artificial intelligence landscape has changed in recent years, and what began in the public eye as a thriving field with promising applications has grown into a $100 billion industry.
The result was ever larger and more complex language models that were often released in a hurry and without proper testing and control.
These models can do a lot of what humans can do, and in many cases they do it even better. They can beat us in advanced strategy games, create amazing art, diagnose cancer, and compose music. And there is no doubt that AI systems seem somewhat “intelligent”. But can he be as smart as humans?
There is a term for this: artificial general intelligence (AGI). While this is a broad concept, for simplicity you can think of AI as the point at which AI acquires general human-like cognitive abilities. In other words, this is the point at which AI can take on any intellectual task that a human can perform. Agha is not there yet; Existing AI models are hampered by the lack of certain human qualities, such as true creativity and emotional awareness.
The researchers asked five experts if they thought AI could match the general intelligence of AI, and five out of five said yes. But there were slight differences in how they approached the issue. More questions arise from their answers. When can we achieve artificial general intelligence? Can he surpass humans? And what exactly is “intelligence”?
Here are their detailed answers:
Paul Formosa (artificial intelligence and philosophy of technology)
Artificial intelligence has already reached and surpassed human intelligence in many tasks. He can beat us on strategy games like Go and StarCraft, beat us on many language tests, and write a college essay that is acceptable to students.
Of course, he can also make up or “hallucinate” and be wrong, but humans can too (although not in this way).
Given a fairly long time scale, it seems likely that AI will achieve artificial general intelligence, or “human-level intelligence.” That is, he has mastered enough of the interconnected areas of intelligence that humans possess. However, some may worry that despite the advances in AI to date, the AI will not be truly “intelligent” because it doesn’t (or can’t) understand what it’s doing because it’s unconscious.
However, the advent of artificial intelligence indicates that we can have intelligence without awareness, because intelligence can be understood from a functional point of view. And an intelligent being can do intelligent things like learning, reasoning, writing articles, or using tools.
Our AI may never have consciousness, but it is increasingly capable of doing intelligent things. In some cases, he is already doing it at a higher level than we are, and this trend is likely to continue.
Christina Maher (computational neuroscience and biomedical engineering)
Artificial intelligence will reach human-level intelligence, but perhaps not anytime soon. Human-level intelligence allows us to think, solve problems, and make decisions. This requires many cognitive abilities, including adaptability, social intelligence, and learning by doing. And AI already meets many of these requirements. What remains is for AI models to learn inherent human traits, such as critical thinking, and understand what an emotion is and what events can trigger it.
As humans, we learn and experience these traits from the moment we are born. Our first experience of “happiness” is too early to even remember. We also learn critical thinking and emotional regulation in childhood by developing a sense of our “emotions” as we interact with and experience the world around us. Most importantly, it would take the human brain many years to develop such intelligence.
Artificial intelligence has not yet acquired these capabilities. But if humans can learn these traits, it’s likely that AI can too, perhaps faster. We are still figuring out how AI models should be created, trained and interacted with in order to develop such traits in them. In fact, the big question is not whether AI will achieve human-level intelligence, but when and how.
Mirglili Sedals (Artificial Intelligence and Swarm Intelligence)
I believe that artificial intelligence will surpass human. Why? The past provides insight that we cannot ignore. Many people think that tasks like playing computer games, image recognition, and content creation (among other things) can only be done by humans, but advances in technology have proven otherwise.
Today, the rapid development and adoption of artificial intelligence algorithms, as well as the abundance of data and computing resources, have led to previously unimaginable levels of intelligence and automation. If we follow the same path, more generalized AI will no longer be a possibility, but a certainty of the future. It is the matter of time. AI has made significant progress, but not yet in tasks that require, for example, intuition, empathy and creativity. But breakthroughs in algorithms will make it possible.
Moreover, once AI systems achieve human-like cognitive abilities, there will be a snowball effect, and AI systems will be able to improve themselves with minimal human intervention. This kind of “intelligence automation” will profoundly change the world.
Artificial General Intelligence remains a major challenge, and there are ethical and social implications that need to be addressed very carefully as we continue to move forward.
Dana Razazdejan (artificial intelligence and data science)
Yes, AI will become as smart as humans in many ways, but how smart it will be will largely be determined by advances in quantum computing. And human intelligence is not as simple as knowing the facts. It has several aspects, such as creativity, emotional intelligence, and intuition, that modern AI models can mimic but cannot match. However, artificial intelligence has advanced significantly, and this trend will continue. Existing models are limited by relatively small and biased training datasets, as well as limited computing power.
The advent of quantum computing will change the possibilities of artificial intelligence. And with improved quantitative AI, we will be able to populate multiple AI models with massive datasets comparable to the natural multimodal data collection that humans do and achieve through interaction with the world. These models will be able to support fast and accurate analysis.
An improved version of continuous learning should lead to the development of very complex artificial intelligence systems that, after a certain point, will be able to improve themselves without human intervention.
So AI algorithms running on stable quantum computers have a high chance of achieving something like generalized human intelligence, even if they don’t necessarily match all aspects of human intelligence as we know it.
Marcel Schart (machine learning and AI alignment)
I think that artificial general intelligence will one day become a reality, although the timeline remains very uncertain. If artificial general intelligence is developed, then exceeding the level of human intelligence seems inevitable. And humans themselves are proof that a very flexible and adaptable intelligence is allowed by the laws of physics.
There is no fundamental reason to believe that machines are fundamentally incapable of performing the computations necessary to achieve human problem-solving capabilities.
Moreover, AI has clear advantages over humans, such as better speed and memory capacity, fewer physical constraints, and the potential for greater rationality and iterative self-improvement. As computing power grows, AI systems will eventually surpass the processing power of the human brain.
Thus, our main task is to better understand intelligence itself and learn how to create AI. Existing AI systems have many limitations and are far from mastering the various areas that characterize AI.
The path to artificial general intelligence is likely to require unpredictable breakthroughs and innovations. And the average expected date for AGI on Metaculus, a good prediction platform, is 2032.
A survey of experts conducted in 2022 showed that by 2059, the probability of achieving human-level artificial intelligence is 50%.
The report was prepared by Noor Gilani, Technology Editor of The Conversation.
Source: Science Alert
You must log in to post a comment.