All technological developments lead to discussion and conjecture. Particularly in the case of artificial intelligence, this is even more obvious (AI). For decades, this technology has been the subject of conjecture and science fiction, solidifying its status as a menace to society and a technology of the future.
A contemporary kid AI is a model of an artificial intelligence system that was created by a parent AI under human direction. This would suggest that AI has developed the ability to “procreate” by writing code and figuring out the variables required for the new algorithm to function.
AI may eventually be able to make judgements on its own without human input at the highest degree conceivable. What is a child AI? The short answer is that it can actually learn how to develop new Artificial Intelligence.
This understanding raises a number of additional concerns regarding the effects of child AI. These include the boundaries, potential, and even morality of artificial intelligence. We will address these issues in this essay in an effort to advance the conversation on AI.
The key issue at hand is whether an AI can independently learn how to create another AI. Artificial intelligence has thus far relied on human direction to complete tasks. In essence, it was developed with the intention of creating a new AI. We previously created the first AI offspring in this way.
However, scientists are working to achieve unsupervised learning, which will do away with the requirement for human guidance. Despite the fact that unsupervised learning poses a number of novel difficulties for academics to overcome, every advancement in artificial intelligence gets us one step closer to this objective.
Examples of modern child AIs
In 2017, when Google’s Brain Team unveiled their discovery, it was clear that it would have a significant impact on society. This initiative was led by none other than AutoML, an AI that developed a new Google AI Child by pushing the boundaries of automated machine learning.
The reinforcement learning strategy was used by the researchers on this team. With this, they are able to suggest a young AI that is later trained and assessed. Since the new AI outperformed the more experienced ones, the findings were more than impressive.
A few months later, the first AI offspring, known as NASNet, was produced via AutoML (Automated machine learning). This AI was created to classify objects in images, and it was tested on the COCO and ImageNet datasets for object detection. The outcomes revealed that NASNet outperformed other algorithms in both datasets.
The utilisation of hyper networks is another situation that merits attention. A hyper-network created by researchers can forecast the parameters of another untrained neural network.
One million algorithms were used in the training of GHN-2, another AI that creates AI. It can then forecast the parameters of another system after that. With excellent success, the researchers have tested the untrained systems using the parameters that GHN-2 has predicted.
What Are the Child AIs’ Limitations Right Now?
Artificial intelligence algorithms can produce other AIs with astounding results, as we have already seen. It is crucial to emphasise that there are some restrictions.
First of all, human supervision is required for the systems in use today. This implies that the parent AI must be developed in order to produce another AI system. Of fact, researchers are always pushing the limits of contemporary technology, and they have already demonstrated in theory that an AI can independently generate another AI.
We also need to be aware of the fact that machine learning demands a lot of computer power, which is both expensive and unsustainable. One possibility that could change all of that is the development of quantum computers, but that is still in the future.
The intricacy of the system is another problem that emerges in hyper networks. We might not be able to determine where the error occurred if it occurs when we use an AI algorithm to forecast the parameters of another AI. Blackbox AI is another name for it.
Please read Is Black Box AI Dangerous in our article. if you want to find out more information on the topic.
Is it possible for an AI to develop another AI without the help of a human?
Whether an AI can produce a kid AI without human intervention is probably one of the most often asked questions in today’s world. This is the point at which technology truly begins to resemble science fiction, and it could not be that far off.
Is it currently possible for an AI to create another AI without human intervention? is the succinct response. In our essay “Can an AI Create Another AI,” we go into great detail on this topic.
While there is currently an AI that can write code, GHN-2 can predict parameters that we may then employ in an untrained system. With the use of one of 12 coding languages, OpenAI’s Codex can translate commands given in natural language into fully functional websites.
Of course, Codex has limits at the moment, which eliminate the possibility of an AI developing independently. It yet demonstrates that it has the capacity to evolve into something larger.
Paired Open-Ended Trailblazer, on the other hand, is a technological advancement that would enable the idea of an AI producing a child AI without human interaction (POET). Virtual bots must navigate a variety of environmental obstacles and challenges that the algorithm creates. The algorithm generates both the layout of the barriers and the evaluation of the bot’s skills.
Both Codex and POET are unique technologies with various uses. To develop a new AI algorithm, it might be possible to combine these two methods. In this method, the AI would be able to evaluate the surroundings and develop the code necessary to create a child AI without the assistance of humans.
Artificial general intelligence: A futuristic concept?
Artificial intelligence currently appears to have advanced to a level that we have only read about in science fiction. Every new technological development raises the question of whether artificial general intelligence (AGI) is attainable. An AI that can create a child AI would be potentially moving in that direction if such a thing is possible.
But what does Artifical General Intelligence actually mean? The phrase describes a system that is capable of comprehending, learning, and carrying out every task that humans are capable of. This indicates that the algorithm is cognitively capable of analysing any task, even one with unknown parameters.
All of the AI systems created to produce new AIs make reference to this issue. Both POET and the OpenAI Codex are already capable of understanding natural language and independently analysing and designing environments. However, we are still a long way from the point where these AI algorithms are comparable to the human-level AI that the AGI envisions.
When will AI outperform humans in particular situations is the subject of a highly intriguing paper with the title “When Will AI Exceed Human Performance?”. This study found that there is a 50% possibility that human-level machine intelligence (HLMI) will be attained in about 45 years.
It is impossible to deny that technology will continue to advance. Artificial intelligence will keep finding new uses in our daily lives, which will increase its importance.
Advances in computer vision, natural language processing, tables for tabular organised data, and many other applications will be made possible by algorithms like AutoML. Child AIs’ near-term goals are to provide fixes for certain problems. AIs will be used by researchers to create new algorithms that outperform the ones now in use.
Unsupervised learning and open-ended artificial intelligence are both subjects of growing research at the same time. It won’t be long before an AI is developed that can give birth to a kid AI without any guidance or assistance from humans.
