“Corporate data is like gold for AI models that have already consumed all the information online”
“Corporate data is like gold for AI models that have already consumed all the information online”
Uri Eliabayev, an AI expert, discusses the distinct nature of AI models and the remaining data sources to train them.
While ChatGPT has become the generic name for AI models that have dominated the world over the past two years, each AI model has its own unique characteristics, with significant differences between them. "Each model develops its own personality, and the choice of model depends on our work style and what we expect from the results," said Uri Eliabayev, an AI expert, in a podcast interview as part of Calcalist's 2025 Forecasts Conference in collaboration with Bank Hapoalim and the Phoenix Group, which took place on Tuesday in Tel Aviv.
According to Eliabayev, some users have grown particularly fond of Grok (Elon Musk X’s AI) for its direct and unfiltered approach, which often disregards copyright concerns and tends to be less politically correct compared to other models. "What I really like about ChatGPT, for instance, is that it writes in well-organized paragraphs, providing answers that read like concise summary articles. On the other hand, Claude often structures responses in bullet points, which I personally connect with less. At the same time, for audio and video analysis, I prefer Google’s Gemini," explained Eliabayev.
When it comes to choosing the best AI model for a specific task, Eliabayev advocates a hands-on approach: "I strongly believe in trial and error. Ultimately, the choice of model depends on our working style and the desired outcome." For example, he added, "If I want to create a creative text, I’ll use ChatGPT. For data analysis, I’ll choose a different model."
A key focus in the AI industry today is tailoring model behavior. "Models can be trained to respond in specific ways," Eliabayev explained, "either by exposing them to new data or integrating guidance mechanisms." For instance, certain models are specifically designed to observe copyright laws or avoid addressing controversial topics.
AI models are trained on vast amounts of data to build their understanding of the world, but what happens when the internet—the primary source of such data—is largely "exhausted"? "What could be collected has already been collected," said Eliabayev. "Now it’s about 'new data,' which is constantly being updated, but its volume is relatively small compared to what’s already been processed."
The solution, according to Eliabayev, lies in leveraging proprietary corporate data. "Corporate data is like gold for these models," he emphasized. "It contains unique insights into internal processes, customers, and products, enabling AI systems to adapt to an organization’s specific needs. This creates added value far beyond what public information can offer."
The integration of AI into organizations is not without its challenges, particularly regarding its impact on the workforce. Will these models replace human employees? Eliabayev believes that AI will transform the nature of work rather than eliminate jobs entirely. "AI models will take over routine tasks, freeing employees to focus on more creative and complex work," he said. However, he also warned that workers must adapt to this new reality. "The demands on employees will increase," he noted. "They will need to develop new skills, such as the ability to work effectively with AI systems and understand their limitations."
Eliabayev also highlighted the global race in AI development, noting that technological advancements in the West are primarily driven by private companies. "However, China is closing the gap at an impressive pace," he said. The Chinese government, in close collaboration with private firms, is advancing cutting-edge AI technologies.
In one striking example, Eliabayev described a Chinese AI model that, when given an English-language task, paused, "thought" in Chinese, and then delivered a response in English. "This illustrates how advanced these models have become in understanding context and executing tasks," he said. “Whether in the West or the East, the trend is unmistakable: AI models are driving breakthroughs in fields such as video, audio, and analytics.”
