‘AI won’t destroy humanity - but it could change us forever,’ says Dr. Uzy Hadad of the Jerusalem College of Technology and urges caution over apocalyptic AI predictions, warning that the greater danger may be humans becoming overly dependent on machines - while powerful economic interests shape the debate over the technology's future.
Artificial intelligence is often portrayed as an existential threat: Machines could escape human control, millions of jobs could disappear, and humanity itself could eventually be placed in danger.
But Dr. Hadad, head of the Artificial Intelligence program at the Jerusalem College of Technology, says the public should approach such predictions with caution - and take a closer look at both the technology itself and the powerful economic interests surrounding it.
"There are a lot of very strange questions here in this whole story," Hadad says in an interview, referring to increasingly dire warnings from researchers and industry figures.
He points in particular to a widely circulated post by a researcher who previously worked at OpenAI and Anthropic and warned of extreme AI-related dangers. While the post attracted enormous attention, Hadad says the intensity of the reaction should itself prompt questions.
He also notes that the technology industry has repeatedly produced dramatic predictions about the imminent disappearance of entire professions - predictions that have not always materialized.
According to Hadad, the debate over AI safety cannot be separated from a major economic struggle over control of a market worth enormous sums.
On one side are the major companies developing proprietary, closed AI models. On the other is the rapidly expanding open-source ecosystem, which allows companies to operate and develop their own models at significantly lower costs.
Dr. Hadad cites the example of a large Israeli company that previously paid enormous sums for access to commercial AI models. After switching to open-source solutions, he says, the company was able to reduce its costs by approximately 90% while also adapting the technology more closely to its needs.
"These big companies are suddenly under pressure," he says.
That dynamic, Hadad argues, should also be considered when examining growing calls for AI regulation.
Regulation may be necessary to protect the public, he says, but it could also have the unintended effect of strengthening the position of existing technology giants and making it harder for smaller companies and open-source competitors to enter the market.
Hadad does not dismiss concerns about artificial intelligence. One of the most immediate issues, he says, is privacy.
Users need to understand what information they provide to AI systems and how that information can be used. At the same time, he notes, the collection of personal information is hardly a phenomenon that began with ChatGPT.
Smartphones, social networks and online services have been collecting vast amounts of personal data for years.
For Hadad, the key to using AI safely is knowing when a human being is still capable of evaluating its output.
"If you ask the AI in a way that treats it as some kind of oracle that will give you the answer, and then you immediately approve it, you can get yourself into trouble," he explains.
When users understand the relevant field and are capable of checking an AI-generated answer and identifying mistakes, however, the technology can become a powerful tool - helping develop ideas, write and edit material, check code and save considerable amounts of time.
Will AI turn against its creators? What about the nightmare scenario in which machines eventually turn against humanity?
Hadad believes much of the public's fear is influenced by popular culture and decades of science-fiction stories.
"Doomsday" scenarios naturally attract attention, he says, making them a powerful tool not only for entertainment but also for generating publicity.
From a technological perspective, he argues, it is important to understand what a language model actually is.
At its core, a language model attempts to predict the next word based on enormous quantities of information on which it was trained. Its increasingly sophisticated capabilities come from systems that connect the model to additional tools and allow it to perform increasingly complex tasks.
The principal danger, therefore, is not necessarily that AI will suddenly become conscious and decide to destroy humanity.
Rather, Hadad says, the greater concern is what humans may choose to do with increasingly powerful technology.
"Like any technology, there is a danger that people will use it maliciously," he says.
Jobs will change - but may not disappear
Hadad also rejects simplistic predictions of mass unemployment.
AI is already transforming professions, he says, but transformation does not necessarily mean elimination.
In software development, for example, programmers increasingly use AI to generate code. Their role is consequently shifting toward reviewing, correcting and directing the systems.
A similar process could occur in law, accounting and other professions. AI may perform an increasing amount of technical work, while humans remain responsible for professional judgment and accountability.
But this transformation carries its own danger; The danger we should really fear.
Hadad believes society should be much more concerned about what happens when people become dependent on AI for tasks they once performed themselves.
If people become accustomed to allowing machines to read for them, write for them, think for them and solve every problem for them, they could gradually lose the very skills that make them capable professionals and independent thinkers.
He cites computer scientist Cal Newport, who has warned that while social media damaged people's ability to read and concentrate, AI could potentially damage their ability to write.
Hadad says senior software developers have already described this phenomenon to him. Their work has become faster and easier, but sometimes also less intellectually demanding.
Instead of spending hours wrestling with a problem, searching for a solution and thinking deeply about the issue, the machine increasingly performs much of the process.
For Hadad, that is the real warning sign - keeping humans in the loop.
The greatest challenge of the AI era, he says, is therefore not figuring out how to escape the machine, but ensuring that humans remain capable of doing what machines cannot replace: thinking critically, questioning answers, creating and taking responsibility.
That challenge extends to education. Schools and universities, Hadad argues, must teach students how to use the new tools while ensuring they do not graduate with little more than the ability to write an effective prompt.
At the Jerusalem College of Technology, he says, new AI tools and infrastructure are already being incorporated into academic studies. At the same time, the institution is reducing the emphasis placed on homework assignments and moving toward more examinations and in-person presentations.
The goal is to ensure that behind a product that an AI system can generate is a student who actually understands it.
The objective, Hadad says, is for graduates entering the technology industry to possess deep, fundamental knowledge - rather than being someone who "just played around with some kind of tool."
Ultimately, Hadad's outlook is considerably more optimistic than the apocalyptic predictions dominating some AI debates.
Artificial intelligence may not take away people's jobs, and he does not believe it is on the verge of destroying humanity.
But it does present humanity with a new challenge.
As machines become capable of doing more and more, people will have to prove that they still know how to think for themselves.
Hebrew interview: