On ‘Strawberry,’ Reasoning, and Intuition

The launch of OpenAI's new reasoning-based models shakes the market

The AI community is abuzz this week surrounding the launch of OpenAI’s new reasoning-based artificial intelligence model ‘Strawberry.’

The model is designed to “think through” options and planning procedures before providing a response. The idea is to produce a model that performs better on tasks that require planning. It also appears to succeed at explaining its own “thinking” process in ways that are eye-opening and different than previous models.

As an English teacher, I do not feel equipped to provide an analysis of the model itself. As such, I’d like to share information from researchers and thinkers that I trust. If you are looking for a more thorough breakdown, I would recommend the following:

Ethan Mollick - Something New: On OpenAI’s Strawberry Launch

Benjamin Riley: The new version of ChatGPT might be a big deal

Nick Potkalitsky’s LinkedIn Post

But, the news of this launch also reminded me of a profile of the “Godfather of AI,” Geoffrey Hinton, last November in The New Yorker.

At the time that I first read it, I was just diving into AI and the article proved to be formative for my own perceptions and ideas around what the advancement of this technology meant for me and my students. If you have not read it and have access to The New Yorker, I highly recommend it.

I’d like to share a paragraph from the article that I felt was poignant and relevant. Pay close attention to Hinton’s final comment:

How should we describe the mental life of a digital intelligence without a mortal body or an individual identity? In recent months, some A.I. researchers have taken to calling GPT a “reasoning engine”—a way, perhaps, of sliding out from under the weight of the word “thinking,” which we struggle to define. “People blame us for using those words—‘thinking,’ ‘knowing,’ ‘understanding,’ ‘deciding,’ and so on,” Bengio told me. “But even though we don’t have a complete understanding of the meaning of those words, they’ve been very powerful ways of creating analogies that help us understand what we’re doing. It’s helped us a lot to talk about ‘imagination,’ ‘attention,’ ‘planning,’ ‘intuition’ as a tool to clarify and explore.” In Bengio’s view, “a lot of what we’ve been doing is solving the ‘intuition’ aspect of the mind.” Intuitions might be understood as thoughts that we can’t explain: our minds generate them for us, unconsciously, by making connections between what we’re encountering in the present and our past experiences. We tend to prize reason over intuition, but Hinton believes that we are more intuitive than we acknowledge. “For years, symbolic-A.I. people said our true nature is, we’re reasoning machines,” he told me. “I think that’s just nonsense. Our true nature is, we’re analogy machines, with a little bit of reasoning built on top, to notice when the analogies are giving us the wrong answers, and correct them.”

I love analogies and have used them heavily in my arguments in favor of the concept of anthropomorphizing AI, which means approaching and interacting with it more like it is a human than a technology.

I have received both support and pushback on this stance, but I believe it is an effective tactic for the same reasons stated above by Bengio and Hinton. It leans on analogies that the average person can understand - like ‘Stranger Danger.’ 

However, Yale Researcher Luciano Floridi wrote a research paper detailing the risks of engaging in “conceptual borrowing,” a phrase that essentially means using analogies to connect one discipline to another. That paper strikes a cautious tone with regard to over-use of analogies, but then ends on an optimistic note that subtly supports my own beliefs and approach.

In the context of ‘Strawberry’ though, my initial instinct is worry – not only because we are once again experiencing a possible leap in technological capabilities for which we are not cognitively prepared – but also because this shift to an emphasis on ‘reasoning’ takes the technology away from how our brains actually work, according to Hinton.

In The Singularity is Nearer, author Ray Kurzweil goes to great pains to emphasize the importance of analogies in our cognitive processes and subsequent advancements in knowledge and intelligence. Charles Darwin, he notes, stumbled onto The Theory of Evolution from an analogy to geology.

Specifically, he leaned on the conclusions of Scottish geologist Charles Lyell, who argued that canyons were not God-given creations - as was the prevailing view at the time - but were created by rivers that appeared first and slowly impacted the rock into its subsequent formation.

“[Darwin] took Lyell’s concept of the significance of a river eroding one small grain of stone at a time and applied it to one small genetic change over a generation…defending his theory with an explicit analogy,” Kurzweil writes.

In The Origin of Species, Darwin wrote, “As modern geology has almost banished such views as the excavation of a great valley by a single diluvial wave, so will natural selection, if it be a true principle, banish the belief of the continued creation of new organic beings, or of any great and sudden modification of their structure.”

These points emphasize the value of our intuition and the use of analogies. They lead me to wonder if our collective love of reasoning and logic actually betrays a deeper truth that both Kurzweil and Hinton seek to describe; That’s not actually how our brains work. 

The question is, Do we want — or need — a reasoning machine?

Before I leave, I’ll provide one anecdotal example that drives towards an optimistic note. When I sit down to write an essay, I rarely engage in a reason-based planning process. Some people like to outline, but I find that outlining often hamstrings the creative process. The logic gets in the way. Utilizing logic first, I might develop a better flow of ideas in my first draft, but I will never get to the deepest truth, which is of utmost importance to me.

Instead, I sit down and intuitively lay down my thoughts on the page. If my ideas are good enough, they tend to connect to one another without running into too many logical snafus. If they are not good ideas, they run into quite a few of these traps, sometimes leading to my decision to abandon the idea. My logical review of the intuitive creation sometimes highlights enough gaps to call the intuition into question, but only after the fact.

This mirrors the process that Hinton describes in his quote. Said differently, analogies help us build, logic helps us trim down. 

As a result, there might be reason for great optimism as it pertains to ‘Strawberry.’ A better logic machine might free us up to engage in deeper, more intuitive, creative work, subsequently acting as the “logic trimmer” on our quest for the deepest truths. 

What do you think?