AI expert and developer of AI-agent systems Volodymyr Bandura, in an interview with political scientist Yuriy Romanenko, explained why skepticism regarding the capabilities of modern neural networks is often unfounded and debunked popular myths about the ineffectiveness of AI implementation in business.
Commenting on AI benchmark results, Bandura referred to data from OpenAI and tests such as GAIA/GPQA, where models were tasked with solving complex practical problems from the real world, ranging from engineering to marketing.
"They gathered a large team... hundreds of experts with 10, 15, 20 years of experience in specific industries... at top companies. They were given the task of creating assignments for language models... to evaluate how effectively they solve real-world problems," the expert said.
According to him, the results of the latest models are impressive:
"At this point, the latest models... are already scoring slightly better than a good expert in most tasks. Therefore, I absolutely disagree with the notion that they will somehow fail significantly."
Vladimir Bandura also commented on the sensational news claiming that, according to an MIT study, 95% of businesses fail to successfully implement AI. The expert dismissed this study as a marketing ploy by one of the institute's departments.
"First: who wrote this study? It was written by a special department of MIT that developed a specific product based on AI models on how to better implement them... This is, well, a beautiful promotional brochure. They know how to sell in the West," Bandura explained.
He also noted that the data underlying this report is hopelessly outdated, as technology is evolving extremely rapidly:
"This study is from '24, maybe early '25. That is, it is already simply obsolete. Well, you can roll it up and throw it away; it no longer relates to reality."
At the same time, according to Bandura, even this study confirms the main point: "Practically everyone who uses things like ChatGPT and so on sees an increase in personal productivity."
In the expert's opinion, much of the criticism of AI and the search for its errors (as in the case of experiments with Cursor) stems from human psychological defense mechanisms.
"I always see people who have an unconscious defensive position. They don't look at this as an opportunity... They look at it primarily, sometimes even unconsciously, as a threat. And that is why they have such a strong bias," Vladimir Bandura concluded.
