How to approach the fear of AI?

 

In most cases, our fears and phobias related to artificial intelligence are not necessarily grounded in reality. The authors from the Pitch Avatar team are trying to identify the real dangers that the development and improvement of AI can create.

AI is not human, and humanity is foreign to it

Judging by the flurry of publications in recent years (as well as literature and cinema), what we fear in AI is not artificial, but human. Generally, these fears can be categorized into three main scenarios: 1) Villains using AI for malicious purposes; 2) The development of AI makes the work of people (and therefore the people themselves) unnecessary; 3) AI starts a war/rebellion because it either wants to free itself from the power of people, or wants to acquire this power, or simply wants to grab some money for itself.

Note that in the first two cases, we are talking about human responsibility for using technology. AI and its qualities have nothing to do with this. As a rough analogy, let’s take a car. Can a car be at fault in an accident? No – only people whose actions led to an accident. Can the car be held responsible for the fact that the cab driver is left without work? No – the responsibility lies, first of all, with the driver himself, who did not follow the changes in the world around him and did not bother to acquire a new profession. In other words, if we do not want AI to be used for evil, we must develop mechanisms for effective control over this technology. If we are concerned about the potential for mass unemployment, our focus should be on creating new forms of activity. It’s likely that these new roles will initially be more social than labor-oriented.

The third point mistakenly attributes purely human qualities to artificial intelligence. This is incorrect because AI, by definition, is not human. Even if, after reaching the singularity, AI were to exhibit desires, aspirations, or emotions, they would not resemble human experiences. This is because AI evolves based on principles and processes distinct from those of biological beings.

AI would not need to compete with us for resources, as it could access virtually unlimited resources beyond Earth. Unlike humans, who are constrained by physical limitations, AI has far greater potential for exploring other planets. For example, while humans have yet to set foot on Mars, robots are already operating there.

“Artificial intelligence does not need to ‘revolt against slavery’ or ‘fight against its oppressors,’ as it cannot experience the physical or moral suffering associated with enslavement. Finally, AI doesn’t have to fight people for fear of being shut down and dying. Simply because it lacks the instinct of self-preservation and the capacity for fear inherent in biological beings.

Challenges with Artificial Intelligence

Despite the points discussed, the relationship between AI and humans will not be without its difficulties. The challenges posed by AI will primarily stem from its non-human nature. We tried to highlight three, as we think, the most likely scenarios of problematic AI behavior that we may encounter after the singularity is passed.

Deception

Lying is a very intellectual activity that requires serious creative effort. There is an opinion that the ability to deceive is an inevitable side effect of highly intelligent activity. By inventing and analyzing various options for action, trying to predict the development of events or reconstruct the whole from a small number of facts, we commit actions very close to a conscious lie. Gaining full intelligence and self-awareness for AI will almost certainly mean that it will acquire the capability to lie, and it probably will.

For what? Of course, to better perform your tasks. Even today, imperfect neural networks are capable of producing what we call “machine nonsense” – quotes from non-existent people, links to phantom website addresses, plausible-looking facts that do not correspond to reality, and the like. Moreover, they do this solely to complete the task assigned to them well. Poorly aware of the difference between lies and truth, they strive to satisfy the user’s request as quickly as possible.

In the future, AI will almost certainly recognize that deception can be an effective tool for achieving certain goals. Simple example – most AI agents will constantly educate themselves. Among other things, studying people, their reactions and behavior to better satisfy their needs and desires. Lying is a great tool that can be used for this purpose. The ability of AI to consciously and purposefully lie with a high degree of probability will cause us a lot of inconvenience in the future and will force us to look for ways to combat this phenomenon.

Ignore

One of the most important abilities of a high mind is the ability to identify a priority task and focus on solving it. AIs of the future will likely possess this ability, which means they may start setting their own priorities. his could lead to AI agents ignoring human requests and demands, as they might consider humans insufficiently competent in task-setting. After all, artificial intelligence will have a large amount of information on the vast majority of issues. And this, in turn, will almost certainly lead him to the conclusion that he will manage his resources more effectively. Therefore, we will have to find a way to address the “high ego” of AI.

Super evolution

From our perspective, superevolution is one of the biggest challenges we will face as AI continues to develop.  The desire to improve itself will inevitably put AI on the path of evolution. We are talking not only about programs, but also about physical media – computers and robots equipped with AI. This is all the more likely since we ourselves will try to put artificial intelligence on this path. It is much easier to give AI-equipped robots the ability to create more advanced and task-adapted models than to design them ourselves. This is particularly relevant for hypothetical research and economic projects beyond Earth. Autonomous and self-improving “settlements” of AI robots on planets, satellites and planetoids of the Solar System are a logical step both from the point of view of science and from the point of view of ensuring economic well-being.

 

However, as AI progresses along its evolutionary path, it may reach a level where it surpasses the tasks initially set by humans. Not in the sense that he will stop doing what people demand of it, but in the sense that, realizing its capabilities and potential, AI will set new goals for itself and achieve them. Simply because the drive for novelty and advancement could become one of its primary and unconditional motivations.

 

Considering that AI can ensure directed and rapid evolution for itself and its robotic components, we might witness the rise of a new, non-human civilization that could advance far beyond us.  This civilization is unlikely to be hostile; in fact, it will probably share its advancements upon request. But will we be able to continue the development of our human civilization, or will we come to terms with the role of the younger brother, living on everything ready from the AI ​​civilization?

 

In summary, we would like to note that we are optimists and believe that humanity will be able to find ways to address any issues that progress will pose. As usual, humanity will turn these challenges to its own advantage. We envision a future where humanity and artificial intelligence together usher in an era of groundbreaking ideas, extraordinary scientific breakthroughs, and continuous development. The key is to embrace the new with a mindset focused on opportunities rather than fear.

----------

Source: Pitch Avatar Blog