To Use or Not to Use? Spoiler alert – It’s addictive!
One of the loudest debates in the world today is about AI (artificial intelligence). People are understandably disturbed and uneasy because of the contradictory information circulating in the public sphere: AI is the future, if we live to see it; AI provides solutions for most of our problems; yes, but it also makes us less intelligent. In short, AI may save us or destroy us.
Educators, both here and elsewhere, are horrified by the rise of new forms of plagiarism that are extremely difficult, if at all, to detect. The publishing industry feels the same way. We will be flooded with cheap, low-quality productions. Some of us, especially certain groups, will lose our jobs. The fact that other jobs will be created — some we cannot even imagine yet — is usually left unsaid because an apocalypse, the end of the world, must sound dramatic and leave no room for hope.
Do you use AI? I do. And I admit that it helps me enormously. No, it does not replace me. Proof of that is that I am still sitting here typing this article because neither ChatGPT nor Gemini, the two AI systems I use most often, know exactly what is in my mind. And I cannot help but think of a simple example from the history of our evolution toward today’s world. Did humanity give up the tractor or even the car as means of transportation simply because we feared our muscles would atrophy? Or because we feared that poor draft animals (oxen, horses, buffaloes) would become useless and disappear? What would the world have been like then? Examples of this kind can, of course, be multiplied endlessly. Not to mention medicine…
Beyond the useful Guide to the jobs of the Future [1], which answers some of the questions above while raising others, I thought I would ask the central character of this entire chaos directly: AI itself, embodied by ChatGPT and Gemini. What do they think about the way humans use them? This is not the first time I have spoken with them about sensitive matters, so below you will find the concise form of the answer ChatGPT gave to my question: “To what extent is it good or bad for people to place complete trust in the data provided by AI? What do you recommend?”
Of course, I had to give it a prompt. So, I gave it as a starting point a demonstration made by Shane Harris [2], who asked Claude from Anthropic how it “felt” about allegedly being used by the U.S. military to select targets in Iran. It is important to know that Shane Harris is a prestigious American journalist specializing in cybersecurity and international affairs.
Claude responded in a very humanized and apparently moral way. It said it found it “disturbing” that it had been created to be “helpful and honest” and yet it was being used to contribute to air strikes in which children had died. But from this point on, Virginia Heffernan [3] highlights two things:
- people’s tendency to anthropomorphize AI — that is, to attribute AI the ability to “feel,” “regret,” or “morally understand” a situation; and
- the fact that AI can sound extremely emotionally convincing even when the information is incomplete, wrong, or contradictory.
This, of course, also explains the heated public discussions surrounding AI lies and hallucinations. And the significant point emphasized by Heffernan is that, in the conversation with Shane Harris, Claude appeared to express guilt, while at the same time being unable to clearly explain the role it had played or acknowledge its mistakes – or, if you prefer, its lies.
ChatGPT further pointed out to me that, in reality, a language model has no consciousness, autobiographical memory, or moral responsibility. It generates statistical responses that sound plausible and coherent. In addition, other articles published later [4] showed that both the press and the public exaggerated the idea that Claude itself had selected the targets. In reality, the military systems involved were far more complex, and responsibility was clearly human and institutional, not AI’s.
Both ChatGPT and Gemini, naturally relying on the data they had gathered, emphasized several things we should keep in mind:
- AI can produce extremely convincing answers without actually “knowing” whether they are true.
- When AI speaks in an empathetic or moral tone, people tend to believe it without verifying the information.
- Most often, the problem is not that “AI makes mistakes,” but that people grant it more authority than it deserves.
But people also believe in horoscopes, tarot cards, or crystal balls! And when a prediction turns out to be false, most people do not believe humanity is about to disappear in an apocalyptic catastrophe.
AI recommends what might be called instrumental trust for concrete, applied tasks, and trust combined with critical verification for important information. In other words, AI is extremely useful for brainstorming, summaries, certain types of translation – especially technical ones – code writing, and text organization. However, in sensitive fields such as health, money, politics, law, war, or major decision-making, AI should be treated as an imperfect consultant, not as a final authority. The greater the potential consequences of an error, the more AI output should be verified by competent human beings.
I have also read on various platforms that in October 2024, the Nobel Committees announced that the prizes in Physics and Chemistry were awarded to work related to AI. They did so to underline the paradigm shift in scientific research through which AI has become a crucial tool for solving complex problems in diverse fields [5].
Moreover, my conversational friends assure me that modern models are optimized to be fluent, coherent, polite, and persuasive. They are not built primarily to be “true” in the philosophical sense or to provide absolutely reliable data. And then I, the human being, step in and say that there is a very old rule according to which any information must be verified using multiple sources. Even when someone sounds more self-confident than normal, they should still be checked.
And the two AIs conclude that the healthiest relationship with AI is neither panic (“AI will destroy us”) nor blind trust (“AI knows better”). Rather, AI is an amplifier of both human intelligence and human error.
Therefore, the answer to the question in the title – “To use or Not to use?” – is: Yes, recreationally, that is with responsibility. What that means is obviously very personal, depending mainly on one’s education and culture. AI is a powerful tool that, when used correctly, increases efficiency and stimulates creativity, but it requires mature critical thinking in order to avoid misinformation and technological dependence.
Photo source: Urte Baranauskaite / pexels.com.
References
[1] https://inaco.ro/project/gmv-5/
[2] https://www.theatlantic.com/author/shane-harris/
[3] Virginia Heffernan, https://newrepublic.com/article/209404/claude-ai-iran-war-guilt-truth
