JERUSALEM, Israel — Artificial intelligence is entering a new phase in which AI systems are being developed to create and improve other AI systems, according to Israeli MedTech expert and innovator Dr. Ronen Tal-Botzer during a technology briefing attended by members of a Philippine media delegation, including representatives from NET25.
Tal-Botzer is a technology entrepreneur and biomedical informatics specialist whose work spans artificial intelligence, machine learning, genomics and digital health.
With more than two decades of academic and industry experience, Tal-Botzer said artificial intelligence has evolved significantly since the rise of generative AI.
“Not only content can it write, but also AI. So it’s AI-generated AI. It’s when one AI system develops another AI system,” Tal-Botzer said as he discussed what he described as the next revolution in artificial intelligence.
The presentation traced AI’s development from the early work of Alan Turing and the invention of neural networks to modern large language models. Tal-Botzer said the latest advances are enabling AI to design and optimize other AI systems with minimal human intervention.
He also warned that while AI can uncover patterns in massive datasets, statistical relationships should not automatically be interpreted as proof of cause and effect.
“There is, of course, no causation, but there is correlation,” Tal-Botzer said, explaining that the era of big data makes it easier for AI systems to discover misleading relationships unless the results are carefully interpreted by humans.
The discussion also addressed cybersecurity and data privacy amid concerns that malicious actors could use advanced technologies to gain access to mobile phones and personal information.
Tal-Botzer described the danger as real, particularly when such technology is used by terrorist organizations or authoritarian regimes.
“It can begin with terror organizations that can penetrate your phone and things like this, and also dictatorship regimes,” he said. “So yes, it’s a real, real threat, and there is some kind of a trade-off.”
Companies providing AI-related services must meet multiple security standards, he added.
“We need to comply with many, many different requirements and many different levels of security,” Tal-Botzer said. “And we do that, but there is a trade-off.”
He cited Israel’s Nimbus project, which he described as the government’s cloud infrastructure, as an example of efforts to keep sensitive information within the country.
“The machines, the computers, the servers of the cloud—of this Nimbus cloud, of the Israeli government cloud—are physically here in Israel, and they are guarded by soldiers,” Tal-Botzer said.
Keeping the infrastructure inside Israel helps reduce the danger of sensitive government data reaching foreign actors or critical systems being compromised, he explained.
Tal-Botzer also demonstrated how evolution-inspired computing is transforming AI development through genetic algorithms, in which thousands of possible solutions compete, evolve and improve over successive generations.
“Genetic algorithms simulate the process of evolution,” he said, explaining that stronger solutions are selected while new “offspring” solutions inherit the best characteristics from previous generations. This allows AI systems to improve continuously through experimentation.
The technology is already finding practical applications in industry. Tal-Botzer cited a project involving drones that monitor construction sites for safety hazards, including workers not wearing helmets and other dangerous conditions.
Because actual accidents are too rare—and too dangerous—to use as training data, researchers instead created synthetic construction sites using technology similar to that used in the gaming and film industries.
“There were not enough people falling from the roof,” Tal-Botzer said, explaining why his team generated thousands of simulated construction scenarios to train AI models safely.
After repeatedly refining those simulations through an evolutionary feedback process, the system achieved what Tal-Botzer described as nearly perfect accuracy in detecting certain safety violations without relying on real accident data.
“This is one of the first in history that was able to give almost 100% success with no data, just by simulating the data,” he said.
The AI briefing formed part of an innovation and technology immersion program organized through the Embassy of Israel in Manila under the leadership of Israeli Ambassador to the Philippines Dana Kursh and Deputy Chief of Mission Ariel Seidman. Attended by members of a Philippine media delegation, including representatives from NET25, the initiative sought to strengthen relations between the Philippines and Israel by giving Filipino journalists firsthand exposure to Israeli expertise and encouraging deeper cooperation in technology, innovation, security, healthcare and other areas.
