In the latest news from the AI frontier, the townspeople appear to agree there’s a new sheriff in town. No one is sure whether the former sheriff, OpenAI, who has been enforcing the law of the state of artificial intelligence for the past two and half years, has any intention of retiring or moving on. But, in a kind of remake of Shanghai Noon — a comedy starring Owen Wilson and Jackie Chan — a Chinese start-up, DeepSeek, has taken over the plot initially dominated by the American cowboy, Sam Altman, CEO of OpenAI.
The New York Times technology columnist Kevin Roose calls DeepSeek a “scrappy Chinese A.I. start-up.” By scrappy, he means two things: pugnacious and managing to accomplish significant feats with astonishingly limited means. Given that DeepSeek has literally humiliated Silicon Valley and the Wall Street investors who now fear they may have over-invested in not so scrappy tech giants, the rest of us should start wondering whether ChatGPT still deserves to retain its status as our BAFF (Best Artificial Friend Forever).
Matthew Berman, the excellent commentator on all things AI, tweeted an example of dialogue using the new chatbot. “DeepSeek R1,” he informs us, “has the most human-like internal monologue I’ve ever seen. It’s actually quite endearing.”
Today’s Weekly Devil’s Dictionary definition:
Human-like:
Applied to anything mechanical, capable of provoking human emotions, such as perceiving it as empathetic, endearing or adorable, adjectives that formerly were reserved to the description of humans or pets.
Contextual note
Anyone who has spent time with ChatGPT — as I have over the past two years thanks to my “Breakfast with Chad” and “Outside the Box” columns — will acknowledge that OpenAI’s Large Language Model (LLM) possesses a tone of voice designed to express an authoritative point of view in its responses to a user’s prompts. You can enter into a dialogue with it, but most people simply ask it a question and wait for the answer. It often takes the form of lecturing or expert counseling.
Berman explains why he finds DeepSeek “endearing.” Unlike other chatbots, it takes the time to explain its goals, decisions and processes of reasoning. It even comments on the nature of the challenge it is dealing with along the way. Rather than calling this “endearing” one might call this “implicating.”
I have often stated my belief that we mustn’t be content simply with milking AI’s infinitely extended knowledge and its well-honed reasoning capacity for specific purposes. More than that, we need to build and refine a veritable culture of communication in which we accept AI as one of the actors, along with friends, family and colleagues. DeepSeek’s “implicating” tone should facilitate that effort.
You can follow Berman’s example here. In the first paragraph of DeepSeek’s explanation of the three killers problem, it announces, “Okay, let’s try to figure it out…” Another paragraph begins, “Hmm, let’s break it down step by step.” It then interjects, “Wait, maybe…” and later: “Wait, let me visualize this.” These moments of adjustment in the reasoning process replicate the very human experience of thinking aloud and taking stock of one’s progress while navigating a complex task. This is not only endearing, it’s especially pedagogical in the best sense of the word. In comparison, ChatGPT typically sounds pedantic.
There may be a good reason why ChatGPT feels less engaging and endearing. OpenAI quite logically aligns with the dominant productivist consumer culture in the US. Consumers have no time for dialogue or deconstruction of thought processes. They expect the LLM to adopt a know-it-all position, like an interactive encyclopedia. Why waste time on the process when the product sought is the answer to the formulated question? In contrast, DeepSeek has cultivated a style that reflects its corporate name: It poses as a seeker rather than a provider of solutions. Playing on our human emotions can even create the impression that the mechanical chatbot is a seeker of wisdom.
Berman highlights the pragmatic value of this style of communication. In his demonstration of DeepSeek’s attempt to solve the three killers problem, he qualifies the chatbot’s performance as “perfect thinking” and adds, “I love absolutely being able to see the chain of thought.” Within that chain, we notice human-like hesitations and interrogative reflection provoked by possible nuance. The process points to the possibility of contrasting conclusions. DeepSeek is reproducing the type of process found in Plato’s dialogues. It highlights the back and forth of an active brain seeking to contrast competing hypotheses as well as the effect of credible conditions.
This highlights the neglected appreciation that AI can help us not just solve problems or answer questions, but, more ambitiously, overcome the kind of black and white, true/false approach to learning instilled in us by our education systems. We can thus use the experience of an active AI dialogue to train our own logic to do more than what we were taught to focus on at school: get good grades. This is especially true in our era of standardized testing, which produces conformist knowledge and punishes creative thinking.
When I interrogate ChatGPT on complex social, cultural and political questions, in the guise of a solution it often recommends cultivating critical thinking. DeepSeek appears to have been designed to engage in and illustrate the dynamic, non-linear movement that true critical thinking implies.
Historical note
Interestingly, in one of Berman’s YouTube productions, when promoting a sponsor’s website that gives access to AI agents, he promises that the agents will provide “the right answers at the right time.” This idea appears to contradict the transparent thinking processes he sees as one of the greatest benefits of AI. But “the right answers” correctly sums up what most people expect to get from AI. This is a predictable effect of the consumer culture that dominates not only our thinking, but also our schooling.
Who has time these days to join a conversation with an intelligence unless there’s the expectation of a clear, productivist, self-interested and potentially profitable outcome? For most of us, our schooling taught us that learning was about getting and giving “right answers” to get good grades. It wasn’t about refining our chain of thought and even less about learning to interact “endearingly” with others.
But in our social relationships with human intelligences — our friends, family and colleagues — how often do we find ourselves looking for or expecting answers? Anyone whose interactions with others proceeded principally on that basis would quickly be branded as annoying. The productivity of social interaction among humans is not measured by the number of solutions to problems or answers to questions. We measure it by the improvements in our quality of life and the shared experiences that generate a sense of relative harmony and synchronized effort.
Berman is right to rejoice at the visibility of the “chain of thought” DeepSeek shares with us. But humans need more than that for their individual and collective development and social intercourse. Some of the great fiction writers a century ago tried to reproduce it. They called it the “stream of consciousness.”
If we really want to set criteria for understanding the nature of the singularity — the predicted moment when AI will equal or surpass human intelligence — we should not underestimate the role of our stream of consciousness. It actually makes no sense to imagine that anything approaching consciousness could be an attribute of AI. After all, neurologists, psychologists and philosophers cannot agree on what human consciousness is or how it is produced. But every human being — including scientists and philosophers — knows what it is through their senses.
Most psychologists would agree that everything we learn happens through our stream of consciousness, simply because consciousness never stops and always produces some kind of effect. Theoretically, learning never stops, even when we do the same thing over and over again. So, as we seek to build a more nuanced culture of interaction with AI, let us applaud the progress LLMs have been making in reproducing the “chain of thought,” one specific component of our mental life and skills. That will help us collectively to train our mastery of logical chains of thought as well as structure our stream of consciousness, which is essentially made up of non-linear logic.
*[In the age of Oscar Wilde and Mark Twain, another American wit, the journalist Ambrose Bierce produced a series of satirical definitions of commonly used terms, throwing light on their hidden meanings in real discourse. Bierce eventually collected and published them as a book, The Devil’s Dictionary, in 1911. We have shamelessly appropriated his title in the interest of continuing his wholesome pedagogical effort to enlighten generations of readers of the news. Read more of Fair Observer Devil’s Dictionary.]
[Lee Thompson-Kolar edited this piece.]
The views expressed in this article are the author’s own and do not necessarily reflect Fair Observer’s editorial policy.
Support Fair Observer
We rely on your support for our independence, diversity and quality.
For more than 10 years, Fair Observer has been free, fair and independent. No billionaire owns us, no advertisers control us. We are a reader-supported nonprofit. Unlike many other publications, we keep our content free for readers regardless of where they live or whether they can afford to pay. We have no paywalls and no ads.
In the post-truth era of fake news, echo chambers and filter bubbles, we publish a plurality of perspectives from around the world. Anyone can publish with us, but everyone goes through a rigorous editorial process. So, you get fact-checked, well-reasoned content instead of noise.
We publish 2,500+ voices from 90+ countries. We also conduct education and training programs
on subjects ranging from digital media and journalism to writing and critical thinking. This
doesn’t come cheap. Servers, editors, trainers and web developers cost
money.
Please consider supporting us on a regular basis as a recurring donor or a
sustaining member.
Will you support FO’s journalism?
We rely on your support for our independence, diversity and quality.
Comment