Currently, artificial intelligence has demonstrated really impressive capabilities. However, it may still be in its “infancy” compared to what it could do in the future. The next evolutionary step in the segment is AGIs, for which some intermediate steps have already been taken. The “agentic” capabilities of some AI platforms are just an example of AGIs’ capabilities. However, these “AI entities” are expected to match humans in any task.
OpenAI, the precursor to the current artificial intelligence revolution in the tech industry, has as its main goal the development of AGIs in the future. This is not easy, and the company is aware of this. In fact, there isn’t even a real consensus among AI experts on when the first real AGIs might arrive. There is even a step beyond AGIs: “artificial super intelligence.”
AIs that match human capabilities in any task will arrive “over the next five to 10 years,” Google DeepMind CEO says
Demis Hassabis, CEO of Google DeepMind, offered his vision on the matter. “I think today’s systems are very passive, but there are still a lot of things they can’t do. But I think over the next five to 10 years, a lot of those capabilities will start coming to the fore, and we’ll start moving towards what we call artificial general intelligence,” he said.
Basically, Hassabis believes it will take us a decade to see the full potential of AGIs. However, these “AI entities” won’t suddenly arrive in 10 years. We’ll gradually see their capabilities over the next decade until a single AGI integrates them all. For Hassabis, an AGI is “a system that’s able to exhibit all the complicated capabilities that humans can.” Some call them AI agents, although this can lead to confusion with current AI platforms also called “agents” that focus on specific fields.
The tech industry expects AGI to be a form of AI that can match human capabilities in virtually any task. AGIs could also be much more autonomous than current platforms. Some chatbots, like Google Gemini, ChatGPT, and Perplexity, have integrated “Deep research” capabilities, which is an “agentic” feature. When you use Deep Research, the AI can independently research information on a specific topic, interpret it, summarize it, and organize it into a white paper-like report. During the process, the AI ”made decisions” based on its “reasoning ability.”
If you think this is impressive, it’s just the beginning of what AGI could offer. “We’re not quite there yet. These systems are very impressive at certain things. But there are other things they can’t do yet, and we’ve still got quite a lot of research work to go before that,” Hassabis said.
Experts disagree on when AGIs and “superintelligences” will arrive
Of course, developing something like this is far from easy—or cheap. This is why estimates of when real AGIs might arrive vary widely among AI experts. Robin Li, CEO of the Chinese giant Baidu, believes AGIs are “more than 10 years away.”
Others are somewhat more optimistic. Dario Amodei, CEO of Anthropic AI, projects that a form of AI that is “better than almost all humans at almost all tasks” will emerge in the “next two to three years.” Jeetu Patel, Cisco’s Chief Product Officer, is even more optimistic, stating that an example of AGI could arrive this year.
However, Patel doesn’t seem to be referring to a “full” AGI, but rather something similar. Additionally, he says there are “three major phases” in AI development. He calls the current phase “basic AI,” while the next is “artificial general intelligence (AGI),” a stage where AI platforms receive reasoning capabilities comparable to those of humans. The third stage, “superintelligence,” will be something beyond anything we’ve seen or will see in the short term. While he doesn’t give a specific timeframe, Patel states that superintelligence is “at best, a few years out.”
Hassabis has also referred to so-called “superintelligence” or “artificial super intelligence” (ASI). The CEO of DeepMind believes that these types of “entities” will surpass human intelligence. However, he doesn’t dare to give a timeframe for the arrival of these developments. “No one really knows,” he said when asked.
What is needed to achieve artificial super intelligence (ASI)
Hassabis also opines on what it will take to reach that point for AI. Basically, AI platforms need to analyze and understand the context of the real world. While it sounds simple, it’s a rather complex goal to achieve. A true superintelligence should be able to understand everything around it and act autonomously accordingly, according to its goal.
2025-03-18 15:10:19