Exploring the Enigmatic World of AI

Attempting to Understand an Uncertain Future

In today’s technologically-driven world, numerous software startups continue to flood the market with Artificial Intelligence for services or goods, (AI for X). These ventures are prevalent across multiple fields, from AI innovations for corporate technology, healthcare and even dating.

While the influx of such companies may indicate that AI has reached an acceptable level of maturity and readiness, it is essential to probe deeper into reality. Considering AI’s potential dangers and ethical implications, moral guidance is necessary to dissect the current AI landscape.

Man must control technology; technology must not control man. It is time for us to ask a crucial question: Does AI aid or inhibit man’s goal of sanctity and eternal life?

The AI Revolution: On the Brink of Unknown Territory

Despite the substantial progress in artificial intelligence, a comprehensive understanding of these technologies remains elusive. As we strive to analyze AI’s current potential and the direction in which it is heading, it is vital to embark on an intellectual expedition.

Understanding AI’s intricate workings involves acquainting ourselves with critical concepts, such as “generative AI” and “foundation models.” The breakthroughs in AI that have captured global attention involve generative AIs, which have pushed its capabilities to unprecedented heights.

Generative AIs are based on complex foundation models, allowing them to perform “human-like” tasks considered unattainable a few years ago. These cutting-edge models are often licensed and utilized by startups and large corporations seeking to harness their capabilities.

Progress and Precautions: Navigating the AI Realm

While AI technology’s speedy advance may be awe-inspiring, it’s equally crucial to be aware of the potential dangers that come with it. The unpredictability of artificial intelligence lies in its rapid pace of learning and improvement, making it challenging to determine the future consequences of this technology.

Despite the hype surrounding AI’s capabilities and its emerging omnipresence, there is still an undeniable need for caution. As the AI landscape rapidly transforms, the consequences and impact remain uncertain.

Generative AI systems, which create eerily human-like responses to textual prompts, realistic images and artificial voices, continue to develop rapidly. To truly grasp the possible directions AI development could take, it might be helpful to draw similarities to other revolutionary technologies in their nascent stages.

Consider the steam engine, for instance. In the early 1700s, no one could have predicted that the rudimentary steam-powered pumps invented for mine water extraction would eventually evolve into highly efficient steam turbines, a key component in electrical power generation. This could not have been anticipated, especially when the concept of electricity had not yet been discovered. Early steam engines were conceived through instinct and experimentation rather than a deep understanding of thermodynamic principles.

According to George Musser, a renowned author working on a book about exploring human and machine intelligence, the first steam engines were developed long before the scientific principles underlying them were understood.¹ Thermodynamics eventually emerged as a response to the development of the steam engine, becoming one of the most universally applicable branches of physics.

Today, we witness a strikingly similar sequence of events with AI. Advanced AI systems are being developed through intuition and trial and error, while our comprehensive understanding of these machines remains vague.

As we embark on this journey to uncover the secrets of AI’s “thought” processes, we must remember that our estimations will likely differ significantly from the eventual outcome. Regardless of AI development’s path, our collective responsibility is to remain vigilant and informed about the dangers and challenges innate to this rapidly evolving frontier.

Exploring the Facets of Machine “Reasoning”

A growing number of researchers firmly believe that a particular type of foundation model has achieved what can be considered reasoning, serving as a significant step forward in AI development.

To better comprehend this concept, it is essential to become familiar with the term “large language model.” A large language model is a form of generative AI trained exclusively on textual data, representing foundation models. Some notable examples include cutting-edge chatbot technologies employed by leading tech companies.

The ongoing debate revolves around whether large language models have transitioned from simply memorizing and regurgitating known information to synthesizing it in new ways, which implies reasoning abilities. Can AI make a judgment in the same way the human mind reasons? If so, has AI achieved the same operational faculty of the soul equal to human intelligence?

Google researcher Blaise Agüera y Arcas points to the capacity of current large language models to tackle complex tasks as evidence of their reasoning capabilities. He argues that it is possible to obtain correct responses to basic mathematical queries from a large language model when provided with sufficient context. This is true even in scenarios where the exact product of two multi-digit numbers is not in its training data.

The development of foundation model AIs requires a comprehensive analysis by moral theologians to assess whether it should be continued. This is not unlike the Chinese experiment where doctors surgically inserted a uterus in male mice to determine if they could give birth. One must ask, for what purpose? Does this have a legitimate place in the plan of God for society? How could this new technology be misused morally?

Navigating the Perils of AI: Ethics and Misuse

As the understanding of AI’s capabilities grows, it’s of utmost importance to ensure that advancements in machine intelligence are approached responsibly and morally, preserving the greater good of man for the glory of God. By keeping a vigilant eye on potential dangers and engaging in informed discussions, we can strive towards a future where AI serves as a valuable tool for the betterment of society, not its detriment.

Despite the revolutionary advancements in AI, some experts argue that today’s large language models possess little to no reasoning capabilities. Proponents of this view suggest that what people perceive as reasoning could merely be an extensive database of memorized information. As these models expand and acquire new capabilities, it may not necessarily mean that teaching them a language, leads to an acquired ability to reason.

Predicting the future of AI is as challenging as forecasting the impact of railroads, planes and rockets for the pioneers of steam engines. A fundamental cause of this uncertainty lies in the limited understanding of the pretraining data used in AI models. This lack of knowledge stems from two sources.

• Firstly, numerous AI companies have opted against disclosing the contents of their pretraining data.
• Secondly, the size of these datasets, comprising almost all openly accessible web text complicates determining whether an AI’s response stems from pre-existing knowledge within the dataset or is reasoned out.

Despite these limitations, growing evidence suggests that large language models possess some form of rudimentary reasoning skills, albeit far below human standards.

Unlike other forms of communication, such as images or sound, language offers a powerful tool for describing and understanding everything in our world. It allows us to create abstract models of our environment, even without using other senses like sight or hearing, as understood by philosophers of antiquity. This quality is the human ability to differentiate between essence and accident, and it is questionable whether a computer will ever achieve this.

A vast language database enables AI language models to eloquently discuss topics like the relationship between two colors despite never having seen them because they use data derived from human senses and defined by language, all a consequence of the human intellect.

Moreover, language serves as the primary interface for many other online systems initially designed for human use but could be repurposed by generative AIs, such as search engines. When considering the evolution of these large language models, it becomes apparent that we might soon see AI-powered personal assistants tailored to our individual data.

Tech companies are already experimenting in the early stages of this notion by allowing AI algorithms to access and synthesize information from our emails, calendar appointments and documents within their systems. However, these initial attempts are still relatively rudimentary and prone to mistakes and abuse.

The rapid development of AI technology, while undeniably fascinating, also raises numerous concerns surrounding its potential dangers through the increased risk of privacy invasion, targeted manipulation and even biased decision-making.

Agüera y Arcas suggests that the emergence of such hyper-personalized AI assistants appears inevitable based on the current trajectory of artificial intelligence technology. Though specifics about future products from major tech companies remain undisclosed, the existence of such assistants is an undeniable implication.

Another key inference, is that future AI systems will attain new capabilities resembling human skill acquisition by providing them with access to cloud-based software developed for human users, which offers specific services.

The most basic example of this process involves granting chat-based AI access to search engines like Google. However, the internet is home to many other search engines besides Google, including repositories for coding, legal decisions, academic research and more.

Generative AI systems are increasingly integrated with services initially designed for humans through “plugins” used in phone, tablet and computer software.

For instance, travel search services, such as those owned by prominent travel companies, and online shopping services can be accessed through advanced AI chatbots via plugins. This integration showcases just one aspect of AI’s potential for transforming the way we interact with technology, causing man to be inextricably linked to technology.

The necessity of plugins for large language models stems from the fact that, despite being trained on enormous amounts of information, they lack access to specific knowledge that is not publicly available on the internet. Moreover, their knowledge base is only as current as the data they were last trained on. Furthermore, even with a wealth of information, these models will continue to have difficulties with certain types of reasoning and decision-making.

As we provide access to the same resources the human intellect already uses, the true potential of cutting-edge AI for X services and startups becomes increasingly evident. Rather than simply offering a licensed and repackaged version of an existing foundational model, these innovative companies can begin incorporating a wide array of data and services. For example, an “AI for legal counsel” service could utilize databases filled with legal rulings, while an “AI for medical diagnostics” system could access medical literature repositories. By harnessing the rudimentary reasoning capabilities of large language models, these systems can provide individuals with far more reliable and accurate answers than these models are currently capable of generating.

The gap between the steam engine’s invention and the locomotive launch spanned over a century. New scientific disciplines emerged along the way, catalyzing the countless innovations that ushered in the Industrial Revolution. Similarly, the evolution of generative AI may follow this same trajectory, with the near future showcasing groundbreaking achievements alongside years of fine-tuning and fervent attempts to exploit and benefit from these novel technologies.

Predicting the future with such advanced cognitive aids is as challenging as forecasting the impact of railroads, automobiles, jet planes and rockets from the perspective of the early steam engine inventors. We must be vigilant about the potential dangers and misuse of AI technology and seriously consider the moral implications associated with its development and application.

How does AI “think,” and are we truly ready to bear the consequences of this technology’s evolution?

While AI advancements show great promise in practical matters, it is essential to recognize the potential dangers and moral implications of its misuse. As technology evolves, so too must our understanding of its ethical implications. Man’s reason for existence is to sanctify himself by corresponding with God’s grace in preparation for his judgment and to be with God in Heaven for eternity. How will AI help mankind achieve his purpose in life? Without an answer to this question, the most significant concern with AI is not addressed.

Reference:

• 1 https://www.scientificamerican...