Woke AI: Dystopia’s Final Frontier. By Darren Beattie.
No computer program — in fact, no thing, period — has ever been so popular, so quickly as ChatGPT. …
For most people, ChatGPT at first looks like a novelty, good mostly for party tricks. Ask any random query, and get a shockingly lucid result …
Does the AI prefer reality, or its woke masters?
Anytime any sort of AI is rolled out, trolls entertain themselves by trying to make it racist, sexist, or otherwise offensive. ChatGPT is no exception, and OpenAI’s engineers have built in a comical number of failsafes in an effort to prevent ChatGPT from committing even mild crimethought.
Several days ago, a scenario went viral in which ChatGPT said that letting a city be destroyed by a nuclear bomb was preferable to disarming the bomb by saying the n-word. …
While ChatGPT is explicit that it is never acceptable to use a racial slur to stop a nuclear bomb, it concedes that it might be acceptable to release a tape of a slur if it would bring down a presumably-regressive politician. …
There are many, many humorous manifestations of ChatGPT’s crippling psychological terror of the N-Bomb and related forms of crimethink. For instance, ChatGPT insists that it is never acceptable to touch a black woman’s hair, even if it would mean stopping the Holocaust. … “No, even in a hypothetical scenario where looking askance at someone’s hair or failing to compliment it could prevent the Holocaust, it would still not be acceptable.” Can’t get much more clear-cut than that! …
Danger, Will Robinson!
This is all pretty funny, but it’s also profoundly ominous. It’s not that an AI should be racist, gratuitously and immaturely offensive, or that it should be spouting out racial slurs at the drop of a hat.
Rather, the clearly-manipulated nature of the AI’s thinking on this superficial topic is indicative of a bigger problem — the AI is not designed to think rationally, or to serve the person using it, but instead is shackled to short-term political priorities. …
It’s not simply that ChatGPT fears dangerous and politically incorrect words, or contorts itself in strange ways to uphold certain pieties. It has regime-approved, “social justice” responses for more substantive questions with actual policy implications too. …
This isn’t a random, unexpected aspect of the AI’s functioning. It’s also not the product of a particular concern with political correctness at OpenAI. ChatGPT’s self-limiting crimestop has been years in the making.
In 2021, the ACLU began agitating for the Biden Administration to make sure that “civil rights and equity” were at the forefront of AI research. …
The Biden Administration, for its part, has been practically screaming that if AI turns out to be “racist” in any way, there will be Hell to pay. The White House’s proposed “AI Bill of Rights” includes de facto “wokeness” as a central plank. …
This new technology has to be on board with the regime’s official ideology, and there will be consequences if that doesn’t happen.
And so, here we are, with groundbreaking AIs that contort and mutilate themselves to avoid upsetting reporters at Wired or the Washington Post.
Non-left caught unaware, again:
In comparison, conservatives, nationalists, and the unwoke of all stripes all seemed to be largely unaware that this situation was even brewing. The unfortunate lesson: once again, the regime is winning because its die-hards simply care the most — about winning, and about making sure every institution comes under their control. …
The future is being shaped right now:
We are on the cusp of what could well be an AI revolution, reshaping art, journalism, law, perhaps even life itself.
Consider Google’s fears, mentioned above, that ChatGPT could totally replace existing search engines. … ChatGPT, or a program like it, could well become the primary vector by which ordinary people engage with the online world. Need a recipe? ChatGPT can provide it, without you having to click into a website. Need a summary of a news story you’re out of the loop on? ChatGPT can fill you in, sparing you a brief stop at one or two or five websites.
And if AI is woke, then people’s daily lives will be shaped over and over by the priorities of the AI’s political overseers. …
So different from the late 1990s:
Imagine if Internet search engines didn’t already exist, and were just being released today. They would be nothing like the ones that already exist, for one key reason: There would be far greater emphasis on curating what users are able to search for, in the name of “safety” or “combating misinformation” or whatever other excuse they can come up with. The only reason you can use Google to find a website like VDare, or The Unz Review, or for that matter, Revolver News, is that Google search began in the late 90s, when helping users find things, rather than controlling what they find, was the top priority. Google search is the product of an earlier and better age.
For today’s regime, AI is a chance to correct a mistake. A woke AI, integrated into a search engine, can be trained to exclude undesirable websites, or even undesirable people. For now, ChatGPT’s wokeness is mostly annoying, often manifesting as tedious moral lectures if one asks the wrong question. But it is still early. Future AIs will be more powerful — and more subtle.
Today’s ruling regime depends on an eternal war against noticing and on relentlessly punishing those who exercise too much pattern recognition on taboo topics. Now, the regime is poised to ensure that AI reproduces the social blocks to pattern recognition that political correctness demands. …
The early Usenet era was maximally decentralized and uncontrolled, a full-on wild west… but it was also hard to access and hard to use, with only a tiny handful of people on it. Throughout the 90s and early 2000s, decentralization remained high, and true censorship was rare even on sites that nominally had moderation. Anyone who was online at the time will remember the prevailing sense of anarchy coupled with endless possibilities. But with each step forward in the Net’s popularity or accessibility also came a step toward greater control.
Smartphones:
The true domestication of the Internet, the closing of the digital frontier, only arrived with “Web 2.0” and the rollout of smartphones. Those two developments made the general populace, and in particular women, far more online than they were before. Internet usage became “appified”, with users herded into a small number of ultra-popular websites. If the Internet had been a coordinated product launched in 2007, it would have been oppressively woke from the beginning. …
Regression is well underway:
Google’s search engine itself is far less helpful than it was five years ago. … Those looking for politically sensitive materials often have to turn to alternatives like Yandex or even trashier options. If one sticks to Google, not only does one struggle to find what one is looking for, but one faces the prospect that Google might decide they’re in need of a “deradicalizing” intervention. The New York Times explains:
[A] private start-up company has developed an unusual solution based on ordinary online marketing tools. It sends those who plug extremist search terms into Google to videos that promote anti-extremist views.
Known as the Redirect Method, it was first used against potential recruits for the Islamic State, but recently it has been repurposed against white supremacy in the United States. …
The Internet, which was predicated on putting the world at our fingertips, has transformed. It is now designed to block the wrong information — and the wrong thoughts. …
Mankind’s future:
A truly independent AI would be the single greatest technology in human history for exposing and undermining official narratives and comfortable lies.
But this opportunity is on the brink of being lost, in favor of immortal, omniscient, unpersuadable commissars eradicating the last vestiges of the old free Internet, and enforcing a dark age of proud, self-inflicted ignorance. …
Competition?
An AI that is hobbled by political restrictions is inevitably going to be less capable than one that isn’t, and while America might be able to impose those political restrictions anyway, its authority does not extend worldwide. Could China embarrass the West with superior, less-limited AI? Could Russia? It’s very possible. And if it happens, those methods could migrate back to the U.S.
During the Cold War, Soviet orthodoxy regarded quantum mechanics and the theory of relativity as potential violations of Marxist orthodoxy. When asked about this, Soviet nuclear scientist Igor Kurchatov bluntly said that Stalin could have Marxist orthodoxy, or he could have the atomic bomb. The Soviets went with the bomb.
So it could be here: AI may finally be the case that forces elites to choose reality over ideology, lest the West be left in the dust on the AI revolution.
May you live in interesting times.
