By Michael Zolotov, AI expert, Co-Founder and CTO of Razor Labs
A year and a half after ChatGPT burst into our lives, artificial intelligence is now high in the awareness of everyone. Hundreds of millions of people have used an AI application at least once, and the topic is at the core of the activities of companies, giant corporations, and nations. Where is it expected to advance in the coming years? What new heights, as far as we can predict, is it set to conquer? And whether (and where) does it pose a threat to us, as those who are supposed to make a living on this planet?
Tool Limitations
First, it is important to recognize that although it may seem like artificial intelligence has seen and knows everything, this is not exactly the case. The latest versions of ChatGPT and similar models are familiar with about 70% of all written content available on the internet, challenging the ability to continue training and improving them. However, they rely only on public information—what has been published and exposed to the public. This means their database lacks enormous amounts of non-public information, such as private organizational knowledge of companies. As a result, their impact within organizations is currently limited.
ChatGPT and its counterparts have led a true revolution, making vast amounts of information easily accessible. However, it is important to remember that they are still inferior to human intelligence. Unlike ChatGPT, which primarily processes text, humans see and hear their environment, giving us a deeper understanding of the world. For example, ChatGPT can easily summarize complex scientific documents but lacks basic knowledge that we consider elementary. For instance, a person can learn to drive after 30-40 hours of learning, whereas Tesla requires millions of kilometers, and its driving is still not perfect. The structure of ChatGPT and similar models does not allow them to “think deeply” or plan the answer in advance, inherently limiting the complexity of tasks they can perform.
In my estimation, it will take at least a decade to overcome these limitations and enable tools to achieve capabilities that begin to approach human intelligence. In light of this, I predict that in the near future, we will see some awakening and a better understanding of the tools’ limitations.
The Future of AI: All-Powerful Model, Specialized Niche Expertise, and Organizational Autopilot
We expect AI to use a huge variety of sensors and capabilities.This raises a significant question: Are we heading towards a supermodel that knows “how to do everything”—doctor, lawyer, accountant, mathematician, etc.? Or are we looking at a dedicated AI model tailored for a specific task? The second approach argues that if I need a dermatologist, it’s less critical for them to also have knowledge in autonomous driving or in identifying and intercepting UAVs making their way from Iran to Israel. Especially since “knowing everything” requires an enormous amount of energy—not something that fits into a small chip on a missile (and let’s remember, that missile is not exactly connected to the internet).
It seems that both approaches, the general model and the specialized AI, will coexist depending on the application. There will be applications of general “autopilot,” like ChatGPT, which is supposed to know everything about everything and, over time, will get to know us better and better. It will not only be familiar with our writing style but also with who we are and what our dreams are. It will be able to help us write the next email in our precise and unique style, teach us a new field, and even advise us at crucial junctures in our lives, ensuring we don’t repeat past mistakes. On the other hand, there will be models with a specific purpose that specialize only in that. Guiding an interceptor missile is a great example, but also assisting doctors in diagnosing a particular disease. In the niche of specialized expertise, each such model will be more successful than the general model that knows “everything about everything.”
Another future direction, which has already begun to materialize in some places, is the “organizational autopilot.” This AI function performs actions based on organizational data and has access to all the organization’s information—from CRM and service call contents to the code itself. Such a tool can significantly enhance the organization’s operations across all levels—optimizing sales, automating customer service, training personnel, development, and more.
Will AI Replace Us?
The short answer is “no,” at least not with current technology. But it’s a bit more complex than that. A more accurate way to examine the issue is not necessarily in replacing us in jobs but in specific tasks. Broadly speaking, there are tasks that take 3 seconds or 3 minutes (such as writing an email), and those that require 3 hours or 3 days (such as strategic planning, thinking about innovative work plans, etc.). Currently, AI can assist with the former type of tasks but not with more strategic ones. This means it can help with almost any job but specifically with low-level, repetitive tasks that require shallower thought. High-level tasks that require creativity and planning will remain with humans.
Therefore, AI allows us to work at a higher level of abstraction. It gives us the opportunity to invest our time and energy more in planning and strategy and less in the precise formulation of the next email. The more a job includes a significant component of the “strategic category,” the less reason its holder has to fear AI.
Additionally, for jobs that involve significant interpersonal physical interaction, like sales or doctors who need to give personal attention to patients, AI will assist the person but not replace them.
If You Can’t Beat Them …
The latest developments in artificial intelligence are a real revolution. They have the potential to improve our productivity and increase the GDP of nations on a global scale. My recommendation is to leverage these innovative tools—while being aware of their limitations—to focus on the complex, creative, and strategic aspects of your role. Artificial intelligence can make us smarter, like superhumans surrounded by omniscient experts. By effectively integrating these tools, we can reduce the risk of being replaced by AI. And as the saying goes: “If you can’t beat them, join them.”
Michael Zolotov, 33, is an AI expert with a master’s degree in electrical engineering. He is a Co-Founder and CTO of a group of leading AI companies, including the publicly traded Razor Labs, which develops artificial intelligence technologies for asset-intensive industries; Axon Vision, which develops AI solutions for the defense market; and more. Michael also co-founded the Future Learning school, the first Deep Learning training academy for AI engineers in Israel.