AI Enhances Our Intelligence, According to Pioneer Terry Sejnowski

Terry Sejnowski, a prominent figure in the field of machine learning, has long been an advocate for the beneficial effects of artificial intelligence (AI). In his 2018 book, The Deep Learning Revolution, he claimed that "AI will make you smarter."

Since then, the rapid advancement of generative AI (Gen AI) has transformed our daily lives. In his recent publication, ChatGPT and the Future of AI: The Deep Language Revolution, Sejnowski explores the development of large language models (LLMs) and asserts, "AI is indeed making us smarter."

Understanding Intelligence

The question arises: how do we define being smarter? What does intelligence actually mean? Sejnowski explains that intelligence revolves around problem-solving. He mentions that tools like ChatGPT not only enable him to learn quickly but also introduce concepts he may never have considered. "It's opening up doors," he says.

He likens ChatGPT to a shovel; it enhances our ability to accomplish tasks more efficiently than we could without it. He notes that writers use ChatGPT to overcome mental barriers, thus improving their craft. According to Sejnowski, when he wrote his latest book, the assistance of LLMs allowed him to complete it in half the time he took for his earlier book.

Tool Usage and Improvements

Sejnowski emphasizes that we are currently in a phase of learning how to effectively utilize AI tools. While he acknowledges that ChatGPT can perform many tasks, he also believes it falls short compared to the best human efforts. However, he emphasizes that it often performs better than the average human.

Throughout his book, he integrates summaries generated by ChatGPT, hoping they provide clarity. While some of these summaries may feel simplistic, they don’t detract from the overall quality of the work. Sejnowski’s primary focus is not merely on praising ChatGPT, but on analyzing how generative AI can reshape scientific understanding.

AI and Neuroscience Connection

One of the intriguing points Sejnowski makes is the reciprocal relationship between AI and neuroscience. He suggests that insights gained from AI are helping neuroscientists learn more about the brain, while findings in neuroscience are also aiding the development of AI. This interplay creates a beneficial cycle for both fields.

Sejnowski, who has a distinguished career, serves as the Francis Crick Chair at The Salk Institute for Biological Studies and is a Distinguished Professor at the University of California at San Diego. He shifted his focus to neuroscience because he aimed to comprehend how the brain operates, diverging from the typical AI research paths.

Large Language Models and Human Memory

Sejnowski shares insights about how the functions of large language models, specifically their word prediction capabilities, may correlate with human memory. He points out that everything inputted into these models is turned into numerical data, creating a "context window" that serves as their working memory. This concept may parallel how human brains process stories and information.

He hypothesizes that understanding these processes could reveal more about brain functions, specifically through what he describes as "traveling waves" of neuron activity. These waves, which have been largely overlooked in brain science, potentially hold answers to how the brain constructs narratives.

The Future of AI

Looking ahead, Sejnowski explains that the continued evolution of AI is likely to involve integrating LLMs into larger systems—similar to how language has developed throughout human evolution. He highlights cooperative ventures between scientific fields and AI, suggesting that novel insights could emerge from this synergy.

Sejnowski also notes that improved models and frameworks could lead to breakthroughs comparable to those seen in classical physics. He believes current LLMs are like medieval cathedrals built through trial and error, and as we understand more about them, more sophisticated systems will follow.

Rethinking Intelligence

Through the use of generative AI, individuals are gaining a clearer understanding of their capabilities and limitations. Sejnowski observes that as users become adept at prompt engineering, the tools increasingly reflect their unique styles, allowing for self-discovery. This creates the intriguing concept that achieving true "artificial general intelligence" might not involve creating humanoid systems, but instead, redefining what we understand by intelligence itself.

He poses thought-provoking questions about the origins of general intelligence and how our social interactions may pave the way for its understanding. This invites us to reconsider the traditional definitions of intelligence in both humans and artificial systems.

AI, intelligence, neuroscience