Artificial intelligence has had a tremendous impact on our daily lives. The blessings of it are clear to students and teachers, lawyers and doctors, even plumbers and A/C technicians.
But is AI also dangerous? There is a feeling of unease, and even some dread, among the general population. Some scientists want to ban AI development. Does it really pose a danger to humankind?
The answer is yes. Many things can destroy the world: an unchecked nuclear war, an errant virus that escapes a lab—as COVID did in 2019—or a giant meteor hitting Earth. Harmful code, spreading through the internet, has the potential to cause harm. That is not to say that the meteor, the virus or the bomb plans to destroy the world. It is just an unchecked phenomenon that can cause damage.
AI, in and of itself, is a tool that cannot “plot” to kill humans. However, like all powerful tools, we need to make sure we handle AI safely and responsibly. The main target of a safety program should not be AI per se. Even before the emergence of Large Language Models (LLMs), sites that were not secure could be hacked, and harm could be inflicted. Our main challenge is to secure the internet from harm, be it human or artificial.
But there is a much more imminent and insidious danger.
Widespread use of AI can cause both technological and scientific stagnation, as well as a warped view of reality. The scientific stagnation can arise if new scientists are not educated. It is the scientists who will raise the questions of tomorrow. If no new questions are raised, there will be no need for new answers. It is only a person who has a deep grasp of a field that can envision the next step.
If there are no such people, technological advances will halt. There is already a sharp decline in computer science enrollment in colleges. It has caused alarm in the high-tech industry, which is afraid it will not have enough manpower for future development.
A much subtler danger is that AI will induce a warped view of reality. In 2024, Google introduced the Gemini AI Image generation system. Users realized that when asked to generate images of historically white figures such as Vikings, the American founding fathers and even Nazi SS troops, the images were invariably of black, Asian or Native American people.
Google immediately pulled the system to address the problem. It discovered that the issue arose because the system was built with diversity in mind. This type of error was so egregious as to be easily discernible. But less flagrant mistakes may not be flagged by the user and could promulgate false ideas and facts that will be taken at face value. Even when a worldview or philosophy of the software developer is not specifically coded into the AI algorithm, mass prejudices may be introduced as facts.
Take, for example, the case of Israel and Zionism. AI may follow the prevalent majority of opinions on the web and inform users that the Jews are a settler-colonial society with no connection to Judea, Israel is committing genocide and Zionism is the root of all evil in the world.
LLMs, being programmed to use the available data to construct their sentences, will report this as the truth, since the vast majority of the internet, the United Nations, the International Court of Justice and human rights groups say so.
When asked specifically, AI itself often points out that all these organizations have an anti-Jewish bias, but the topic is not answered on its own merit, rather by quoting what is said about it. The LLM algorithm is not a critical-thought algorithm. It is a statistical algorithm that uses the data it is provided. Generally speaking, that data is the entire content of the web.
It is difficult even for a human being to view facts objectively; it is all but impossible for a statistical program to do so.
The Bible tells us of the Tower of Babel. Mankind wanted to build a tower all the way to Heaven. God is said to have come down and introduced different languages, halting the building of the structure.
The danger of the Tower of Babel was that everyone spoke the same language, so everyone had the same ideas about everything. To avoid stagnation, God had to introduce different languages, different ideas and different ways of seeing the world. The danger of AI is that it could create a Tower of Babel. People could treat it as an omniscient oracle, the one and only final arbiter of truth. That is indeed an existential threat to humanity.
AI is a powerful tool, but it needs to be used for the tasks it is most suited for. It should never be taken as an arbiter of values, and the facts it states should always be verified.
