A recent paper by computer scientists asserts that AIis progressing toward artificial general intelligence, which is another way of saying that machines can perform any task that the human brain is capable of. The paper has been greeted with skepticism at a time when AI-related worries are on the rise.

ALSO READ:Four Ways to Stop AI From Guessing Passwords

"Human-level performance is one thing," Nick Byrd, a professor at Stevens Institute of Technology who studies the operation of reasoning, said in an email interview with Lifewire. "Human reasoning is something entirely different. Long ago, digital calculators approached (and surpassed) human math performance. However, I doubt that their calculations operate similarly to those of the majority of humans."

Artificial General Intelligence

The recent Microsoftpaper's claims that AI chatbots powered by OpenAI's GPT-4 think like humans broke new ground.

In the abstract of their paper, the researchers write, "We demonstrate that, beyond its mastery of language, GPT-4 can solve novel and difficult tasks spanning mathematics, coding, vision, medicine, law, and psychology, without any special prompting."

ALSO READ:AI GPT 4 May Enable Filmmakers to Be Even More Creative

"Moreover, in all of these activities, GPT-4's performance is startlingly close to human-level performance and frequently outperforms earlier models like ChatGPT.Given the breadth and complexity of GPT-4's capabilities, we believe it could be considered an early (yet insufficient) artificial general intelligence (AGI) system."

Large language models (LLMs) that power the current iteration of artificial intelligence (AI) chatbots have been compared to an advanced form of autocorrect. Consequently, many observers were startled by Microsoft's new paper.

Anagha S. Nadkarni, CEO of AI Detector Pro, a company that identifies AI-generated content, stated in an email that the current iteration of chatbots cannot describe specific facts using human-level reasoning.

"For instance, I once asked it to explain the Bengal Famine of 1943, and it did a poor job; its response is circular and weak," Nadkarni explained. "Neither did it delve deeply into the effects of colonialism on India and Bangladesh, nor did it bother to identify which nations comprised the famine-affected region."

"Human performance is something different. Human-level reasoning is an entirely different proposition."

Byrd stated that mimicry rather than reasoning is a better method to describe how current AI chatbots function.

"When I ask a chatbot to write a press release about Microsoft's latest AI achievements, the chatbot generates a series of words that statistically best match the types of phrases and sentences found in similar press releases from its training data," he explained. In the grand scheme of things, this fake-it-until-you-make-it mimicry may resemble what we do, particularly when we are still learning a new skill.

Alexandra Mousavizadeh, CEO of Evident, a company that analyzes the implementation of artificial intelligence, stated in an email interview that GPT-4 does not mimic the diversity and adaptability of human thought.

She added, "It is multimodal and incredibly impressive for a system that doesn't require special prompting to solve problems, but we shouldn't overlook its limitations, such as fundamental logical flaws, hallucinations, and biases."

AI Is a Mimic, Restricted by Limits

Lou Bachenheimer, the CTO of the technology company Blue Prism, stated in an email that while AI may not be able to reason in the same way that humans do, this does not preclude the software from achieving results comparable to brainpower. Bachenheimer suggested requesting generative AI to design a billboard.

"It would design something to fit the size constraint, a rectangle in this case," he explained. "In contrast to AI models, however, humans are capable of thinking beyond specified constraints. In this example, a human billboard designer may consider adding a cardboard cutout to the billboard and arguing that it should not be a simple rectangle. While the AI is capable of designing, it relies on prebuilt assets and will struggle to be as innovative as humans."

ALSO READ:Conversational AI such as ChatGPT may soon have a human-like face.

In the future, even Microsoft paper skeptics believe that AI may be able to approach human reasoning capabilities.

"It clearly will in some domains," Mousvizadeh declared. "In some cases, it will be superhuman, such as in chess, error detection, and programming. We should anticipate the emergence of human-like reasoning in which the structure, adaptability, and complexity of human consciousness are imitated. That is not really present at the moment."