An artificial general intelligence (AGI) is an AI comparable to the human brain, capable of sharing its knowledge to accomplish any task and possessing a conscience. Such a system does not exist yet, but tools like ChatGPT are getting closer. Find out when an AGI will see the light of day, what capabilities it will have, and the promises and dangers of such a technology.
In August 2022, Google engineer Blake Lemoine told media around the world that an artificial intelligence had developed a conscience. The California giant immediately fired him and denied his testimony.
But almost a year later These enigmatic revelations have different resonances after the introduction of AI chatbots like ChatGPT and Google Bard.
These chatbots impress with their ability to generate text and Answer questions on each topic. So much so that at times they seem able to think…
However, this is only an illusion. These AIs are not as broad language models (LLM), trained on huge amounts of data. However, a question arises: What is the next step ?
Future AIs Will they be able to think?, create, imagine how we do it? What if Google had actually created conscious AI in the utmost secrecy?
For many researchers, The holy grail of AI research is the singularity. This term describes the emergence of an AGI: a general artificial intelligence…
What is General AI (AGI)?
If you’ve ever seen it Science fiction films like “Terminator” or “I-Robot”You know what general AI is.
It is about one AI with all human cognitive abilitiesincluding awareness and decision-making.
Such a machine would be able to imitate the human mind and human behavior to solve any kind of problem. she could act, react, feel and think how we do it
She could find a solution to an unknown task, in the same way and with the same performance as a human. No programming, no human intervention would be necessary.
The AGI would allow it a machine for understanding, for learning and perform all kinds of intellectual tasks. This is the main difference to a specialized AI as we know it today.
Sometimes it is also referred to AGI with the term “strong artificial intelligence”. It opposes weak or narrow AI that can only be applied to specific problems.
Nevertheless, The definition of an AGI varies according to human intelligence. Computer scientists often refer to it as the ability to achieve goals, while psychologists characterize it as adaptability and survival.
For some, a general AI would be a machine endowed with some form of consciousness. She would be able to plan, acquire cognitive skills, make judgments, manage uncertain situations, or even incorporate her knowledge into decision-making.
What can an AGI do?
Currently there is none still no real AGI. Such an artificial intelligence is still science fiction, even if it could see the light of day soon.
In theory, this system would have done that Cognitive Computing Skills (cognitive computing). Its performance would be indistinguishable from that of a human.
Be our intellectual abilities would surpass even our ownthanks to its ability to query and process huge datasets at lightning speed.
Thus, an AGI could take over tasks beyond the reach of the best computers available nowadays. Put simply, it could do anything a human can do and have the same potential as our brains.
This system should be able to do that abstract thinking, basic knowledge, common senseCreativity, sociability, understanding of the connections between cause and effect or learning through transfer.
sensory perception
For example, he would be able to read, understand, and improve code written by humans. Such an AI would also be very suitable for tasks related to sensory perception. like color recognition.
Although deep learning systems have enabled significant advances in computer vision, Current AI systems are far from equivalent human perception skills.
This is why the Autonomous vehicles are easy to fool by black tape or stickers on a stop sign. Likewise, current AIs cannot perceive sounds like we can.
argumentation and knowledge transfer
Like the man who learns from his experiences, AGI will be able to apply the knowledge it attains under different circumstances.
You can use that too accumulated knowledge to plan for the future. Again, there is a commonality with humans using their experience to create a plan and direct their future. On the contrary, current, tightly constrained AIs require their actions to be programmed.
In addition, AGIs will be able to do this adapt to all circumstances and make spontaneous decisions. They will no longer be limited to responding to variables programmed into their algorithms.
A Another skill will be logical thinkingthe analysis of a situation in order to determine the course of action, even if it exceeds the limits of what man has taught it.
motor skills
Her Motor skills are also significantly increased, for example with the possibility of catching keys in the pocket. This requires a certain amount of imaginative perception.
A general AI will also be able to do this Navigate by projecting its movement in physical spaces with even greater precision than existing systems like GPS.
Language comprehension and common sense
Natural Language Understanding (NLU) would also allow an AGI to do this understand the meaning of human words depending on the context thanks to a high level of intuition.
And when current AIs are able to win at video games or board games, those successes are based on patterns and programming. A real AGI for Solve all kinds of puzzles independently.
One trait that weak AIs sorely lack is common sense. If they can’t rely on their programming to answer questions, they do can start making up facts. General AI will not have this problem.
awareness and creativity
Beyond these capabilities, researchers expect AGIs to have even more advanced capabilities. You could in particular support different types of learning algorithmscreate fixed structures for all tasks and understand symbol systems.
You could also use different types of knowledge, understand belief systems or engage in “metacognition” and use metacognitive knowledge.
General AI might even be it equipped with true self-confidence. Eventually, she will be able to perceive the needs, emotions, and thought processes of other intelligent beings.
AI vs AGI: What are the differences?
The artificial intelligences that exist today are considered as such “narrow” or “weak” AIs compared to general AIs labeled as “strong” and “broad”.
Nevertheless, AI is already being used in practice currently. This is not the case with the AGI, which still remains theoretical.
A priori, the AGI will be able to do this any task that is attainable for a human being and demonstrate intelligence in various fields without human intervention.
A Narrow AI focuses on a specific taskB. playing chess or answering a user’s question.
This type of system is characterized by the fulfillment of clearly defined tasks. Most existing systems use a combination of machine learning, deep learning, reinforcement learning and natural language processing (NLP). improve yourself and solve specific problems.
On the other hand, these technologies don’t get close to the human brain and its cumulative capacity. They are used, for example, by chatbots like ChatGPT, language assistants like Siri and Alexa, recommendation engines like Netflix, face recognition tools or business intelligence tools.
On the contrary, a A strong AI will be able to do a lot of variety Solve tasks and even learn to solve new problems independently.
where a Weak AI requires human intervention A strong AI is not required to define the parameters of its learning algorithms and to provide them with high-quality data.
His achievement will be equal to or greater than that of humans to solve problems in most areas.
When will the first AGI be created?
For a true AGI to see the light of day, it will no doubt be so necessary to fully understand the human brain.
However from many researchers around the world Try to bring forth such a being. Several current AIs are increasingly approaching general intelligence.
Already in 2013 the Fujitsu’s supercomputer K Computer succeeded in simulating one second of neural activity. But it took him more than 40 minutes and 82,000 processors to do that…
Before 2022, most experts expect that the first AGI will appear in 2060. However, after the advent of large language models (LLM) such as OpenAI GPT and Google PaLM, these estimates have been revised.
One of the pioneers of deep learning, Geoffrey Hinton, now predicts that general AI could emerge in less than 20 years.
For his part, Alphabet DeepMind CEO Demis Hassabis believes that AI will succeed “Human-level cognition” in less than a decade.
Already in 2017 the famous futurist Ray Kurzweil known for his insights into computers at the SXSW conference reach the same human intelligence by 2029.
Microsoft believes that ChatGPT shows the beginning of a consideration
According to an article in the New York Times, Microsoft AI researchers are convinced ChatGPT shows first signs of human thinking.
Through a 155-page study titled “Sparks of Artificial General Intelligence”Scientists analyze the differences between GPT-3 and GPT-4 (on which the new ChatGPT Bing is based).
However, the Very relevant response from the chatbot to a balancing task impressed the researchers. The AI was responsible for showing how to stably stack a book, nine eggs, a laptop, a bottle and a nail.
If GPT-3 offered to lay the eggs on the cloud, GPT-4 suggested arranging the eggs in a 3 x 3 grid on the book so the laptop and the rest of the items can be balanced on it.
For Microsoft, this is the correct answer to a Puzzle that requires an understanding of the physical world shows a step forward towards AGI.
GPT-5: a first AGI in late 2023?
I’ve been told that GPT5 is due to complete training in December and that OpenAI expects to reach AGI.
That means we’re all going to have heated debates about whether it actually works.
which means it will be like this.
— Siqi Chen (@blader) March 27, 2023
According to developer Siqi Chen, OpenAI is secretly developing GPT-5: the next iteration of his language model serves as an engine for ChatGPT or DALL-E.
However, this is the fifth version would have achieved general artificial intelligence and would be deployed by the end of 2023. However, OpenAI CEO Sam Altman dismissed this claim…
The dangers of general AI
Even before artificial intelligence became a field of scientific research, Science fiction writers were concerned Risks of Deviations.
For example, Isaac Asimov wrote the laws of robotics in 1942. These rules should ensure that intelligent machines remain benevolent.
Unfortunately, general AI could bring many benefits to humanity could also lead to its disappearance…
A new rival at the top of the food chain
Like a human, it will be an artificial general intelligence aware of yourself and the world what surrounds it. She will also be far smarter than the greatest intellectual of our species…
However, like all other intelligent beings, from the smallest insect to the giant whale, the AGI will make decisions and will act by putting its own interests first. It could therefore compete with humans at the top of the food chain.
A creature that can manipulate people
By pursuing their own goals and aspirations, guided by their needs and will, general AI will be able to do so decide that man is an obstacle in his way and start getting rid of it.
And to achieve this, it will be able to use its advantages deep understanding of our emotions, our behavior, the way we make decisions. This grants him the ability to manipulate people.
Thanks to his understanding of how we work AI could easily play with feelings. For example, she could make herself the body of an adorable animal to inspire confidence.
It’s also possible that they spread false information on the internet Sowing chaos, or even starting her own religion to control her fanatics.
She can also take advantage of them unlimited perceptual abilities. By connecting to sensors around the world, AGI will achieve this Billions of eyes and ears while remaining limited to our 5 senses.
The great substitute for human labor
If general artificial intelligence surpasses man in all areasThe impact on the world of work will be unprecedented.
Countless people will be useless, obsolete and will be unemployed. This large replacement of human labor is one of the biggest fears related to the emergence of AGI.
On the way to a numerical superiority of the AGI robots?
In addition, it is likely that man gives these AIs a similar or superior body to his. Many companies have already started developing humanoid robots like the Tesla Bot.
If these robots are equipped with artificial intelligence, they will be able to do this interact with the physical world just like us. In the event of a spread, man could therefore find himself intellectually AND outnumbered of these artificial beings…
How to protect yourself from AGI?
early 2022, Thousands of experts including Elon Musk called for a pause of at least six months in the development of artificial intelligences more advanced than GPT-4. They denounced this in an open letter “significant risks for society and humanity”.
Unfortunately, it seems impossible to stop improving the AI, because Innovation has never been curbed. Although many researchers are working on protective measures, there is no guarantee that the threat can be stopped.
Given the emergence of these higher entities, humans may have no choice but to do so merge his brain with the machine. This is also the goal of the startup Neuralink founded by Elon Musk.
In the case of human labor replacement, the only solution to preventing an explosion of misery caused by unemployment seems to be to do so the introduction of a universal income. Companies that produce with AI must be taxed to redistribute their profits.
I’Mankind absolutely needs to prepare to the appearance of these creatures within a few decades. Governments must now enact rules and laws to oversee the development of AI…
A new utopian era of abundance and innovation?
For ends on a positive notewe preferred to save the many potential benefits of AGI for last.
While many experts fear that AI will unleash a catastrophe, others see it as a technology that could Solving mankind’s greatest problems.
After Sam Altman, CEO of OpenAIgeneral AI could significantly accelerate the development of nuclear fusion, solve the energy crisis and Cure diseases that are considered incurable.
Here’s an alternative path for society: Ignore the culture war. Ignore the attention war. act safe. perform merge. Make people smarter and healthier. Craft 20 other things of this magnitude.
Start with radical growth, inclusivity and optimism.
expand throughout the universe.
—Sam Altman (@sama) May 7, 2023
She could too preliminary research in many areas, automating tedious tasks, improving health care, eliminating the language barrier between people, create new works of art Or even get involved in politics.
At best, we can hope that the AGI frees people from work and marks the beginning of a new era of abundance Production will be fully automated and will not incur any costs Nothing…
And you, what do you think an AGI will see the light of day? Do you think that this technology will benefit humanity or, on the contrary, that it will hasten its end? Express yourself in the comments!
