Can AI really dream at all? And will the dream be a revelation or a hallucination?
The dystopian novel of Philip K. Dick (and subsequent Blade Runner cult films) was projecting human-like emotions and sentiment on the Nexus androids, an anthropomorphic representation of the AI. In the novel, the android antagonists are indeed more human than the human protagonist, intentionally. They are a mirror held up to human action, contrasted with a culture losing its own humanity.1
If the AI is to be another sentient species, one could argue that their evolutionary stage is, at best, Amoeba.2
AI is already, amoeba or not, poised to disrupt the world, as much as, if not more than, any and all previous big tech steps, from inventing the wheel to internet.
Most probably, the technological leap and the consequent impact would be of the magnitude of electricity versus the steam, or the gun powder versus the bow and arrow: a major difference in knowledge and technology prowess, changing the very basis of humanity development, for generations to come.
But what is AI?
It’s a catch-all expression encompassing the capability of computational systems to perform tasks typically associated with human intelligence, such as learning, reasoning, problem-solving, perception, and decision-making.
But this is, most of the time only an ‘aspirational’ definition, as the current stage of development of such systems is rather on the basic side. For now, the most spectacular AI, the Generative AI, is actually a vast collection of statistical models and algorithms definitions that allows the machines to (more or less) correctly associate communication items and provide ‘answers’ to user prompts – or, rather, statistically-proved succession of words that are most probable to have been previously written in the vast collection of texts used to train the AI.
So, highly simplifying, it’s a glorified VLOOKUP excell function, coupled with exhaustive tables to compare and a smart CONCATENATE option.
And how does this buzz word translates in the day-to-day life? AI’s range of applicability varies depending on their degree of complexity and sophistication, with some being around for a long time now, while others are still only theoretical concepts:
- Reactive Machines – systems that act based on input, do not store and learn from previous experience, but are able to compute large databases and make sophisticated choice of best options, in a limited time-frame – such as the Deep Blue system from IBM that was able to beat at chess on of the greatest chess masters.3
- Machine Learning – a collection of algorithms and databases that allow the system to learn without/outside of the initial programming; an ability of the system to use additional experimental information as events happen, to collect and then use the information for further deductions. This is already manifesting in several day-to-day applications, such as recommendation systems (Netflix, Amazon, etc.), virtual assistant (Alexa), facial recognition systems, or Spam filters.
- Deep Learning – basically more advanced machine learning systems, using neuronal-modeled links, able to recognize complex patterns and make sophisticated choices of options, in a repetitive way
- Natural Language Processing (NLP) systems - computational systems with the ability to understand, interpret and generate human-like language. All the generative AIs, such as the Chat GPT, or the now-infamous Grok, would fill this category, together with their extension towards computer vision – the ability to see and interpret images&videos, with immediate applications in various industrial fields, in self-driving cars and more recently even in battlefield assessment and target acquiring.
- Affective Computing systems – with the ability to recognize, interpret, process, or simulate human feeling, emotion, and mood – an evolution of some of the virtual assistants and/or the generative AIs, so far with poor or limited results.
And then you also have the Super AI, the Theory of Mind, or even the Self-Aware AI, all at the limit between hypothetical theory and straight forward Sci-Fi.
Beside the training-choosing-learning algorithms, there is now under development the next step – the Agentic AI: ‘automation/robotization’ of the AI, with software engines developed to take action in an autonomous or even independent way, based on the AI algorithms and the choices made, be it about an industrial environment or on a personal preference. And this might be the area of most impact for the future, as humans do tend to delegate more and more agentic action towards tools that make their life simpler.
But if asking Alexa to book a ticket to a movie theater based on your Netflix preferences seems harmless enough, when more important decision have to be made, humanity should be forced to keep control and engagement, the man-in-the-loop with supervision and go/no-go decision power. As Francis Fukuyama was saying, the relationship between Principal and Agent is based on trust: that agent has the knowledge to do its job and the loyalty to observe the principles. But how much can we trust AI? Especially for the loyalty to principles? As we, humans, understand less and less about their programming, or even have been totally excluded, how will we know how/what are the principles they will observe?4
It’s undeniable that AI will bring tremendous opportunities and development rate for the humanity. As with other giant scientific and technological advances, this seemingly incremental step will allow us to exponentially accelerate our knowledge development and our scientific discoveries.
Because AI is a paradigm shifting tech when it comes to data use, for the power to parse, the statistic association and large numbers predictions; and if we have something in abundance, it is, certainly, numbers, data, parameters, to play with. For now AI might be constrained by our computational and energy limitations (AI requires huge compute power and that in turn is using incredible amount of energy to function and water to cool), but couple AI models with quantum computing, and AI capabilities will skyrocket, both in terms of ‘exhaustiveness’ and in terms of speed.
How could this benefit humanity? In all sorts of ways, from tackling the complexities of our environment to new drugs discovery, from faster and more accurate medical diagnose to complex modeling and economic predictions, from automating our day-to-day mundane tasks to really understanding the human nature and societal fabric. And of course, from supervising in real-time the battlefield, to autonomously acquire and eliminate targets, in war.
And this is where the challenges lay.
The AI is relying on training from vast quantity of data, on ingesting, analyzing and categorizing tremendous levels of information. But the GIGO principle is a relentless truth: garbage in - garbage out, and the thought of AI models training on Facebook personal posts, or on X comments and content, is vastly disturbing. There are already numerous AI hallucination examples, which for now can be timely observed, controlled and corrected; but what if an agentic AI is taking action based on a hallucination? What if there are chains of Agentic AIs, that would depend on one-another, and at least one of them would encounter a hallucination? This is where better might become the enemy of good: in the words of Yuval Harari, “We build machines to be better than us - an excavator would dig more&faster than a human spade. This is both the great promise and the great threat of AI”5
At the same time, the AI is vastly based on algorithms, on software written to pursue a certain goal. The main question here is ‘who’s goal’?
Who is the entity that gets to decide the purpose and therefore the course of action for the AI?
Is it a human, maybe owner of some vast tech empire, including some social media platforms? Does that person have moral values, aims truly for the good, or is it only profit-motivated? Is he maybe poised to invest into and own the political world too?
Or, even more alarming, will it be the an artificial entity that is setting up the purpose, an AI that is programming another AI, in ways that even the creators of AIs cannot understand6? Will we end-up in the famous ‘Genie problem’ that simply chooses the most effective way of terminating world suffering – by terminating life?
Like it or not, enthusiast or frighten, the AI technology is here to stay and it’s ours to make or break. The potential is great and AI could help humanity become truly the master of our universe. But the perils are also frightening and the current trust-less world, the current divisive society, the post-factual reality we live in and the 19th century-like resurgence of political and imperialistic world views are not good omens for safely mastering this technology.
Same Harari was highlighting the arms-race of building AIs, with a potential to develop in a new ‘silicon curtain’ that would further divide not only the societies, but also the AI domains, to the point of totally different AI races.7
And in a world without trust between people, an ‘Alien Intelligence’8 would definitely require the proverbial 3 laws of Robotics9:
1) A robot may not injure a human being or, through inaction, allow a human being to come to harm.
2) A robot must obey orders given it by human beings except where such orders would conflict with the First Law.
3) A robot must protect its own existence as long as such protection does not conflict with the First or Second Law.
If only we would have Mr. Asimov to legislate.
Written by Doru Orăselu, Doru is a Chartered Marketer and has succesfully finalized the Profesisonal Diploma in Marketing with Oxford College of Marketing and the Romanian Institute of Marketing.
1 "Entering the Posthuman Collective in Philip K. Dick's Do Androids Dream of Electric Sheep?" Jill Galvan , Science Fiction Studies,#73 = Volume 24, Part 3 = November 1997
2 Yuval Noah Harari: “ChatGPT is the ‘amoeba of AI evolution’”, YouTube BigThink channel, Nov 2024
3 “Game Over: Kasparov and the Machine”, Vikram Jayanti, 2003
4 “Agentic AI and the Delegation problem”, Francis Fukuyama, YouTube Frankly Fukuyama channel, Jul 2025
5 “AI: how can we control an alien intelligence?” Energy Tehc Summit, Yuval Harari and Stephen Fry, YouTube Yuval Noah Harari channel, Jul 2025
6 “OpenAI, Google, and Meta Researchers Warn We May Lose the Ability to Track AI Misbehavior”, Rhett Jones, Gizmodo.com, Jul 2025
7 “Nexus: A Brief History of Information Networks from the Stone Age to AI”, Yuval Noah Harari, Random House, Sep 2024
8 “AI: how can we control an alien intelligence?” Energy Tehc Summit, Yuval Harari and Stephen Fry, YouTube Yuval Noah Harari channel, Jul 2025
9“I, Robot” Isaac Asimov, Del Rey Apr, 2008
