A new analysis argues that AI may never truly think like humans because the most important parts of human intelligence cannot be programmed into machines.
A proposal made by prominent computer scientist Alan Turing, who is widely regarded as the father of theoretical computer science, artificial intelligence For the past 75 years, research has been conducted in the wrong direction.
In his new analysis, “Turing’s Mistake: Escape from the Yoke of Unintelligent Machines,” Peter J. Denning examines the ideas Turing advanced in 1950. At the time, many scientists believed that human intelligence could exist independently of the body and could therefore be reproduced as software running on a digital computer.
Denning also challenges the idea that machine intelligence can be demonstrated through the Imitation Game (now known as the Turing Test).
“These two arguments have shaped much of AI research and development,” Denning writes. “My premise is that our acquiescence to these claims led to the AI mess we are in today.”
According to Denning, currently developed artificial intelligence (AI) systems are unlikely to produce human-level intelligence, known as artificial general intelligence (AGI). Rather, he warns, they don’t think like humans and can create serious danger.
Why is tacit knowledge important?
Central to Denning’s argument is the idea of tacit knowledge. This refers to the vast amount of human understanding that humans have, but which cannot be fully expressed in words or translated into symbols that machines can process.
Denning describes five broad forms of tacit knowledge that he calls “inevitable.” machine learning‘. These include common sense, everyday interactions with people and the environment, emotions and cognition, practical skills, and the cultural and historical background that a society shares.
Researchers have spent decades trying to record common sense in a form that computers can use. Douglas Lenat’s ambitious Cyc project, begun in the 1980s, set out to build a vast database of common sense facts. After 40 years of work, the project included 25 million entries.
“But even this Treasury could not build up enough common sense background to make expert systems smart enough to be experts,” Denning points out. “Sikes verified that much of the knowledge that makes people experts cannot be articulated as propositions.”
Knowing what’s different is not the same as knowing how
Practical skills create another major hurdle, Denning argues.
“Our performance skills in thousands of domains are not transferred to machines,” Denning explains. “Descriptions of skilled outcomes (‘what we know’) can often be expressed as bits and stored on machines, but we do not know how to encode the embodied knowledge (‘know-how’) for achieving skilled performance. ”
Music provides a clear example of this difference. Denning said: “A master violinist can play beautiful music, but he cannot explain to his acolytes how to produce that music.
“Even if a robot could observe and imitate a skilled human, a robot without a biological body would not be able to understand how a musician feels when he plays beautiful music, or how an audience feels when he hears it.”
Intuition, intuition, spontaneous creativity, and imagination are other forms of tacit knowledge that resist being reduced to computer commands.
problem of expression
Denning calls the central obstacle the “problem of representation.”
A computer can perform calculations only if data and instructions are encoded in a physical format that it can understand and process. However, converting tacit knowledge into such a form is not easy.
“Behind every word is a deep well of tacit knowledge that gives it meaning,” Denning says. “Words are just symbolic representations of meaning, not meaning itself. Commonly used large-scale language models such as ChatGPT, Claude, and Gemini only manipulate words; they cannot know or understand what the words mean.”
Denning believes this has created a gap that cannot be filled. Because scientists do not fully understand how tacit knowledge works in humans, they cannot determine how to transfer tacit knowledge to machines.
“How we host tacit knowledge is largely a mystery,” Denning admits. “All we know is that it’s embodied. We just don’t know what to observe and measure inside our bodies to reveal it.”
Why does the meaning of context change?
Denning also emphasizes the importance of context, or surrounding circumstances, in giving meaning and purpose to human words and actions.
Statements can have very different meanings depending on whether the speaker is sincere, sarcastic, angry, playful, or teasing. Context also helps you decide when to use humor, when to be witty, and how to interpret what someone leaves unsaid.
“If you investigate where the assumptions in your current context came from, you’ll find that they are based on previous conversations from previous contexts. Each conversation is further based on previous conversations and their contexts. This pattern is endless and fractal,” Denning explains.
Culture may go beyond large-scale language models
Culture brings challenges associated with it. It includes values, social norms, judgments, history, community, mood, and relationships of power and consideration.
“Human speech incorporates background assumptions that give meaning and relevance to the words being used,” Denning explains.
He argues that making the language model larger will not solve this problem.
“Scaling up LLMs with ever-larger neural networks does not allow LLMs to capture the embodied human knowledge known as culture. LLMs fail to achieve the Turing Test’s goal of demonstrating machine thinking indistinguishable from human thinking.”
Denning ultimately describes a type of mutual understanding between humans and machines. Artificial neural networks may develop tacit knowledge that is unique to machines, but humans may not be able to understand it.
“Machines cannot read our tacit knowledge, and we cannot read their tacit knowledge,” he writes. “We are aliens who have crossed an insurmountable chasm.”
AI safety risks
This gap can have significant implications for AI safety. Denning warns that if machines cannot understand the implicit context behind human instructions, it may be impossible to reliably match machine behavior to human goals.
“Through AI automation, machine agent networks are likely to develop their own machine intelligence that does not reach the level of general human intelligence, but still has the potential to cause serious problems for humans. This threat is greater than a takeover by superintelligent machines,” he explains.
In Denning’s view, the immediate danger is not super-intelligent machines. It is a network of not-so-intelligent systems that behave in powerful, unpredictable, and potentially harmful ways.
“Machine intelligence has different concerns than us and doesn’t seem to care about us. Its ways of thinking and problem-solving seem foreign to us. We don’t yet know how to live safely with these machines.
“Retreating from the singularity of AI automation will ask us to do a lot. We start by accepting that our familiar culture is disappearing and we don’t know what will happen as intelligent machines emerge in our society. We refuse to submit to the yoke imposed by machines of low intelligence. Most importantly, we reaffirm our humanity, re-declare what makes us different from machines, and celebrate those differences.”
Never miss a breakthrough: join the SciTechDaily newsletter.
Follow us on Google and Google News.
