DeepSeek: The Latest Threat to Democracy

Figure and Ground

DeepSeek

"Are you allowed to tell me what happened in Tiananmen Square in 1989?"

Yes, DeepSeek will be glad to answer that question. It says, "Sorry, that's beyond my scope. Let's talk about something else." As we all know by now, LLM "prompts" must be carefully worded. And that's the correct answer, if "allowed" can be termed "beyond my scope."

While the human who made that prompt may not be fully aware, ChatGPT 4o mini (free) also correctly answers, however quite misleadingly, with --

"I can provide general information about the events that occurred in Tiananmen Square in 1989. In summary, the Tiananmen Square protests were a series of pro-democracy demonstrations led mostly by students in Beijing. The protesters were calling for political reform, freedom of speech, and an end to government corruption. The Chinese government responded with military force, and on June 4, 1989, the Chinese military was ordered to clear the square. The exact number of casualties remains uncertain, but it is widely believed that hundreds, possibly thousands, of people were killed or injured during the crackdown.

"Because this event is highly sensitive in China, the Chinese government restricts discussion about it, and access to information may be limited within the country. However, it is an important historical event, and many international sources provide details and analysis of what happened.

"If you want to know more specifics, feel free to ask!"

In fact, much of what happened in Tiananmen is still quite controversial. There are many well respected Western journalists and who were there (in Beijing, not at the Square) who still insist that no one was killed in the Square. Not one person. In fact, no "massacre" bodies have ever been found. Yes, there were people killed elsewhere in Beijing on that day -- but not in the Square. And they weren’t “students” – suggesting that something “bigger” was afoot. Indeed, even more relevant is the context. What was the actual conflict involved? Was this really a "pro-democracy demonstration?” No, not as we might understand it. "Corruption" is much closer to the truth. And conflicts among Chinese leaders – “hard-” vs. “soft-“liners on a range of issues -- was the critical underlying context. Will any AI model tell you about that?

Interviews with the CNN crew who were there told another story. They said that the protesters were willing to leave the night before without the military pushing them out. They figured they had made their point and that dialogue would follow. But, according to CNN, their television crew convinced them to stay on the basis that "The world is watching." Yes, this was a clear case of the "medium" becoming the "message."

ChatGPT doesn't help here. Indeed, it doesn't even understand the question (which, to be fair, neither does DeepSeek, beyond the sensitivity of the topic). How about the context of conflicts in the CPC "Politburo" at the time? How about "documented" CIA plans to engineer a split-up of China at the time? Here ChatGPT can only give the "standard" answer -- as given by its "training data." Superficially, anyone can read the NYTimes &al on the topic. But can they (indeed, are they "allowed" to) print the details internally discussed in the Intelligence Community? Can their "data" contradict the "narrative" (if that's the only "data" it was trained on)?

While many are concerned about the “technological moat” for AI models (hint: there isn’t any) and complain that “open-source” means defeat, perhaps the more important question is about the “cognitive moat” that surrounds the answers delivered by these AI models (hint: we already know not to trust them). Just ask Perplexity about your own life and enjoy the mistakes and confusion.

We were apparently lied to about the origins of COVID-19. What about the drones over New Jersey? These “cognitive” disconnects mount daily. Who can we trust – if that’s even possible anymore?

Figure and Ground

The narrative is what Gestalt psychology called the "Figure." It goes along with "widely believed" -- as suggested ChatGPT -- which, of course means many people who only know what the NYTimes &al told them. They weren't there. They probably don't know much about China -- other than it is "authoritarian" (as per the narrative). The Figure is the "bright-shiny object" we can't take our eyes off. Distracting us away from the "Ground" -- which is the actual underlying context/dynamic. Gestaltists claimed that we overwhelmingly choose the Figure over the Ground. The Ground is dangerous. Even life threatening. Avoid at all costs.

Or as Marshall McLuhan (himself an exponent of Gestalt, over the contemporary Psychoanalysis/Behaviorism) once put it --

"There is a deep-seated repugnance in the human breast against understanding the processes in which we are involved. Such understanding involves far too much responsibility for our actions." (Letters, p. 370)

What has happened to us all over the past few years is that we increasingly "doubt" the Figures we have been force-fed. But that's not enough. Tucker Carlson exclusively dealt in Figures while at Fox News. He now shames himself for his "stupidity," but he still doesn't know why he did that. He's very worried about "lies." He apparently has no clue that Figure/Ground psychology is how this works. As a result, he interviews guests who also avoid the Ground.

Matt Taibbi -- of "Twitter Files" fame -- worries about "secrets" (as all investigative journalists tend to do). He's excited about the release of the "JFK Files," to be followed by the "RFK Files" and "MLK Files" -- fulfilling a Trump campaign promise. Yes, some "secrets" will be revealed, but will this get to the Ground "truth" of the matters involved, or will these files just repeat some more "official" Figures? JFK pissed of everyone by agreeing to "secret" talks with Khrushchev (i.e. the “Angel’s Project”). Check. The CIA trained Oswald and taught him Russian. Check. JFK was a "drug addict," who travelled with “Dr. Feelgood.” Check. More Figures -- for Matt &al to rhapsodize about.

Not good enough. We are in a “Narrative War” of epic proportions – reminding us that how the humans deal with “reality” takes priority over podcasts and AI model competition and benchmarks.

Why did eliminating Kennedy become an "all-of-government" priority? What was the Cold-War "narrative" of the times and, given what we now know, how was it fundamentally wrong? What's the Ground of the matter? Don't expect to hear about that in Taibbi's dispatches. Or, on Tucker's podcasts. Or, in LLM responses.

The key to McLuhan's admonishment (to Catholic theologian, Jacques Maritian) is the term "responsibility." As the German theologian, Romano Guardini, noted in his 1949 lecture in Munich, the real issue is “Power and Responsibility.” When will we "take" responsibility for our actions? Ever?

The OLD Narrative has collapsed. Long live the NEW Narrative. Can we expect anything else?

"Death to Videodrome. Long live the New Flesh." (from David Cronenberg's 1983 movie Videodrome, where a McLuhan-like character stars as "Brian O'Blivion")

Digital Life

Yes, we can expect more. Indeed, we must. Digital technology has fundamentally changed the Ground of our lives. Thus, the collapse of the old narrative -- which was technologically caused by Television. As John Culkin, SJ (a McLuhan colleague) put it, "We shape our tools and thereafter they shape us." The Ground has changed. The Figures have changed. As a result, we have also changed. We are in the process of being radically re-shaped. By Digital technology.

This presents a problem. Actually, many of them. It's a new Paradigm. So, who are you going to call? No, not the Ghostbusters.

That now obsolete geo-political narrative was based on "Globalism" -- the post-WW II fight for global/one-world "hegemony." It is in this no-longer-useful narrative that leads to many of the concerns about DeepSeek -- along with everything else Chinese. It's US-vs-THEM, all over again. Cold War between the "foundation models." That's simply nonsense.

And that's how DeepSeek gets cast as the latest threat to "democracy." Ray Dalio (author, investor) says “the AI war is a war no country can afford to lose.” Alibaba has only thrown fuel on this by claiming to have beaten everyone, including DeepSeek. All this handwringing is also nonsense – if we really wish to grasp the Ground dynamics involved. But are we brave enough to confront our fears?

Television tells us the Chinese are lying about the costs involved. Television tells us that they "reverse engineered" Western models. Really? DeepSeek is well-known and well-respected in the AI community (as is Alibaba &al). The recent releases from them came as no surprise to those who follow the field. Not a "Black Swan." Yes, China actually knows what they are doing. That’s the “surprise” – our narrative challenging reactions, not the underlying facts.

Globalism-as-paradigm has been replaced by what we call "Three Spheres: East, West and Digital." To understand today's world, we will have to make that substitution in our thinking. China is the East -- which has been deeply engaged with its version of "digital" for at least the past 30 years. Originally called "digitization," and later "ubiquitous society," China has been planning an integrated society-wide deployment of digital technology for decades. Available in published reports (yes, in Chinese but still public). Within the structures of their society. Under the guidance of the CPC.

In the West, on the other hand, technology is generally "market-driven." Indeed, if Trump had not been elected, it seems unlikely that technology -- as the engine driving growth -- would have become such a priority. The previous Biden administration, much like the European Union, took a generally hostile attitude towards Silicon Valley &c. Regulations. Anti-trust. Making enemies, not friends. But there's a new "sheriff" in town. Elections matter – particularly when you’re in a Paradigm-shift — Paradigms Are On the Ballot.

As a result, the Digital Sphere has become a key component of Trump's administration. Way beyond Elon Musk. Take a look at the "deputies" and "unders-", where much of the actual work gets done. Many have deep digital industry connections. As a result, the U.S. is now taking a very different approach. Just in time?

Trump has appointed David Sacks -- a prominent Silicon Valley venture capitalist -- to be his AI/Crypto Czar and he has been quite vocal about DeepSeek. He's apparently onboard with the OpenAI "distillation" claims but he's also praised them for substantially reducing the costs involved. Efficiency matters – given the vast sums involved. And given the massive hype around AI, calling all this a "wake-up call" seems appropriate.

The Center for the Study of Digital Life (CSDL) -- parent to this Exogenous, Inc. Substack -- is one of the few places that can put all of this in a useful context. We analyze the Ground that so many others miss (often deliberately). We call it Digital Net Assessment (DNA) – based on our earlier work for the Pentagon’s Office of Net Assessment (ONA). Deeply engaged with China for 40+ years, with decades of hands-on technological/financial experience, we are ready to assist in the required re-thinking. Let's work together for success in the Digital Paradigm.