‘The AI highway and your loss of freedom’ by the co-founder of Wikipedia

The following is an excerpt from Larry Sanger’s Blog @ Larrysanger.org. Sanger is co-founder of Wikipedia, who left the project when he saw that “the inmates were running the asylum” and agenda editors were controlling topics, bios, and pages on Wikipedia.

“The Generative AI Threat to Freedom, and How You Can Help to Stop It.”

1. A third kind of digital freedom: free “as in the user.”

For a long time, the world of open source has recognized two kinds of free software and information: gratis and libre. The Free Software Foundation explains the difference as free “as in beer” (French, gratis) and free “as in speech” (French libre). The difference matters: you can be free to use software but not free to look at its source code, or to develop or repurpose it. There is a similar distinction in the case of free encyclopedias and databases: there are plenty that you are free to read, but you have no right to edit and/or republish the data. Everybody acknowledges there is an important difference here.

But there is a third kind of freedom that only partly overlaps gratis and libre. In an age of corporate and government movements to use software to control the population, we need to celebrate software and information that is free “as in the user”: we can call it autonome (French again).

Autonome software respects the autonomy of the user. It treats the user as an individual with free will that must be respected, not as a programmable member of a collective. Autonome software and information is configurable by you and me; people control it, it doesn’t control people. And it respects your world view. It does not try to manipulate your personal beliefs and attitudes. In short, autonome software and information does not try to change our behavior. It is thoughtfully constructed so that it neither favors nor disfavors any particular social, political, religious, or philosophical views. It acknowledges where mature adults would like to be in control. It offers them that control.

Now, when the Free Software Foundation discusses the gratis-libre distinction, they have tended to assume that libre software is automatically what I am calling autonome.In other words, just because software is free in the sense of being “open source,” it will therefore respect the autonomy of the user, giving the user choices and not attempting to control the user. This makes some sense: traditionally, open source software is highly configurable. It does tend to put the user in control. There is an underlying theory that, if the software’s source code is open to developers, then some developers will inevitably add features that liberate the user in various ways. Unfortunately, this theory has proven to be limited in practice. It was understandable back in the 1990s and 2000s, when so much of the software world was still philosophically libertarian and projects were not doing the bidding of enormous corporations, shadowy government bureaucracies, and crusading ideologues. But that has changed, radically.

Software that is libre is not necessarily autonome. It is entirely possible that some software might be open source, but used by corporations and governments to control its users. It depends entirely on the choices of the developers and their paymasters. If developers want some open source browser to let corporations, governments, and criminals—but I repeat myself—spy on users, they can. If they want to create cost-free and open source AI software that happens to manipulate the political views of humans, they can. If they want to create a censor-happy social media or web search network that runs libre software, they can. If an encyclopedia is carefully edited to manipulate readers, but its owners want to release it under a libre license such as a Creative Commons license, they can.

But there is a third kind of freedom that only partly overlaps gratis and libre. In an age of corporate and government movements to use software to control the population, we need to celebrate software and information that is free “as in the user”: we can call it autonome (French again).

So the free software movement’s focus on open source and open content, laudable as that is, has been irrelevant to the very real challenges to freedom that have emerged in the wake of what was once called “Web 2.0” and then “social media.” The challenges have come from governments as well as from giant corporations (that are as rich and powerful as many governments)—entities that are happy to give money to, and which therefore probably exert some control over, open source operations.

And of course, the internet is now basically run by enormously powerful gratissoftware—most notably Google, YouTube, Facebook, Twitter, Instagram, TikTok, Reddit—all of which are profoundly anti-autonome.

I love a lot of gratis software, and as a committed Linux user, I love libre software. These are important kinds of “free software.” But the biggest challenge today is to produce competitive autonome software. We need software that scrupulously avoids controlling “users,” that never treats them as manipulable addicts, and that instead liberates responsible, thoughtful human beings.

Everyone loves to use autonome software, by the way. But it is not as lucrative as the other kind, and it does not empower the people who rule the world.

2. Hallucinating chatbots are like unscrupulous journalists.

It was the rise of AI chatbots that led to this outburst. So let’s talk about generative AI and the AI-written encyclopedias of the near future. We’ll come back to autonomesoftware—it’s all interrelated.

Let me give you an example from my own experience of using online knowledge resources in early 2023. Recently, I was researching Christian denominations. There are a lot of them, in various families and subfamilies, with differences between them that you can appreciate only if you actually know something about Christian theology and church history. To tell the truth, I’m thinking of finally returning to church, and I figured that I would be willing to drive up to 30 minutes each way, if that would mean I could go to a church that would be just right for me. It turns out that dozens of different denominations are represented within 30 minutes of me, and many hundreds of individual churches. It occurred to me that, given such enormous numbers, it would probably be a good idea to narrow down the selection by denomination.

So I sat down to do some research. I started putting denominations on a spreadsheet. I thought I would catalog the “distinctives” (the distinctive features) of the denominations.

It was in the process of doing that—asking myself questions like, “What is the Restorationist Movement?” and “What is a ‘Holiness’ church?” and “Are there any alternatives to Calvinism and Arminianism?” and looking up answers—that it occurred to me that I might use this opportunity to test out You.com’s gratis LLM (Large Language Model) chatbot, which runs software apparently similar to OpenAI’s more famous LLM, GPT-3. I have also used a similar free AI chatbot at deepai.org, which uses GPT-3 itself. They are very similar.

I forget all the denomination questions I asked, but an example would be, “Are there any Presbyterian denominations that are not Calvinist?” Here is the chatbot’s answer to that one:

Yes, there are some Presbyterian denominations that are not Calvinist. For example, the United Church of Canada is a Presbyterian denomination that is not Calvinist. The United Church is an amalgamation of several denominations, including some that are Arminian in theology. Additionally, the Presbyterian Church in America (PCA) is also not Calvinist, as it is a descendant of the conservative Southern Presbyterian churches that split off from the mainline Presbyterian Church in the United States of America (PCUSA) in the early 1970s. They hold to an Arminian theology.

“The PCUSA holds to an Arminian theology?” I said to myself. “I wonder if that is true.” Then I switched to a traditional search engine and ended up on an encyclopedia article. Depending on the question, I might have visited a denominational website or a series of church websites. I discovered that the chatbot gave me correct answers approximately 80-90% of the time (by my rough estimate). In other words, 10-20% of the distinguishable claims made by the chatbot were actually just some likely-sounding garbage; the chatbot was, as AI researchers like to say, “hallucinating.” As to the question about Presbyterian denominations, it turns out that the conservative “Presbyterian Church in America” denomination is Calvinist, contrary to what the above says; it is the more liberal “Presbyterian Church (U.S.A.)” that is no longer dogmatically committed to Calvinism. So the chatbot got its facts switched around.

I emerged from my research with a set of paragraph-length descriptions of denominations. The chatbot actually supplied some helpful data for these descriptions, but I kept checking up on it, because of course I didn’t trust it, and I was right not to.

At some point, I decided to ask the chatbot some questions about a topic about which I have a fair grasp of the truth. I was interested to see how accurate and fair it would be with regard to—well—myself. It was easy to get the chatbot to hallucinate about me. I asked it to tell me some things my Twitter account had said about pedophilia. The chatbot responded that I had said that pedophiles should be deplatformed from Twitter. Now, I am pretty sure I never said that, but I believe I said something similar to that. So I asked, “When did the @lsanger Twitter account say that Twitter needs to deplatform pedophiles?” The response:

On April 7, 2021, the @lsanger Twitter account [NB: this Twitter link is supplied by the chatbot, but it is a link to a GitHub account, not a Twitter account, and is not my GitHub account, but somebody else’s; my Twitter account is here: @lsanger] tweeted that Twitter should deplatform pedophiles and other people who promote or engage in sexual exploitation or abuse of children. The tweet stated, “Pedophiles and those who promote or engage in sexual exploitation and abuse of children have no place on Twitter or anywhere else. Deplatform them now.”

I never said that. Those words appear in neither Twitter’s fairly reliable internal search engine nor in a Google-based search. I have said that pedophiles should be kicked off the platform, but the word “deplatform” is one that I doubt I have ever used, though I have mentioned it. (As philosophers like to point out, you can mentiona word by quoting it, without using it yourself.) “Deplatform” is just not in my personal lexicon; when I write the word, I am generally quoting others.

When I asked the chatbot what my political beliefs are, it said,

Larry Sanger is a political independent and has stated that he is “generally libertarian and centrist with a strong belief in the need for limited government and respect of individual rights.” He has also expressed support for free markets, limited government, and the need for economic liberty. Additionally, he has voiced his opposition to political correctness, excessive regulation, and the overreach of government power.

This is a hallucination, but it’s not entirely wrong. I am generally libertarian, yes, and I am in favor of limited government. But I am not a political independent, and I have never, ever described myself as a “centrist.” The supposed quotation is entirely fabricated. Making up quotations and URLs is a persistent problem with chatbots at present, although, as we will discuss, that particular problem will almost certainly be solved soon. Based on my tweets and blog posts and things that have made it into print, an unscrupulous journalist might concoct some such likely-sounding quotations out of whole cloth. A reader might emerge with some information about me which is correct, and some incorrect.

The supposed quotation is entirely fabricated. Making up quotations and URLs is a persistent problem with chatbots at present.

That, in fact, is an apt description of AI chatbots (at present): like an unscrupulous journalist or a low-paid listicle writer, they just make stuff up. In fact, the chatbot I used is not bad at writing listicles.

Here’s one it spat out when I give it the prompt, “Write a five-item listicle article titled ‘Five great parks in Ohio for hiking.’” Here for your perusal is the output, but I don’t recommend actually reading it:

I can’t be held responsible if you go there and find the park doesn’t exist, or the hiking opportunities limited. The bot could be hallucinating.

As an avid Ohio hiker, I can tell you this is plausible. Still, who really knows if Mohican State Park features over 13 miles of hiking trails, exactly? It does—but I didn’t know for sure until I looked it up.

But “Custer Memorial State Park”? It doesn’t exist. (End of Section 2). Continued…

Finish reading entire article here.

Visit The Sharyl Attkisson Store today

Shop Now

Unique gifts for independent thinkers

Proceeds benefit independent journalism

Leave a Comment

Your email address will not be published. Required fields are marked *

5 thoughts on “‘The AI highway and your loss of freedom’ by the co-founder of Wikipedia”

  1. This Article and the Facts the D.O.J. and FBI IC3 as well as Congress needs to look into and Regulate and Fine these Bad actors . Despite what you want to believe Pre-A.I. companies have been Hiding in background Live violating “YOUR-Mine and your Kid’s App’s and O.S. Privacy settings anyhow in past decade, even in Private Browser Mode with Browser Also set to strict settings as you use them online Live ! “There there ! Hiding in background violating your settings anti-trust privacy rights, Example of some= 13vod-adaptive.akamaized.net as well as other akamized.net violators and or 126.net or 163.com hiding in background like a bunch of crap house rats with fluttering eyebrows giggling, watching what you me and your kids are doing on these criminal devices ! ” You people need to get smart and have the United States switch over to All communications A.I. Anti-virus servers, for everything ! = Not to Track a living sole ? but to prevent and stop these people and many more people and Data violators from installing Malicious links and or code scripts and fake website tracking codes, malware code as well as to filter all other malicious code 24/7..This will save Billions every year in damages and as well as ensuring these bad Big tech companies from enabling bad actors with there half a$%$ O.S. and browsers Apps, thus ensuring Legal Law Anti-trust communications in future ! “The More you know ?

  2. This article and Larry Sanger search for the right religion denomination ? My advice to Larry ? This is been the problem with mankind for over 2,000 years , and why there’s been so much turmoil in the World much more in middle east countries in History. At one point in the Bible God made the statement He wished he never would have created mankind . God should have never let Mankind right the Books of the Bible ? He sent His Son Jesus to teach the Bible and according to Jesus in the end of his teaching “No one can get to Heaven unless through him ? Thus the difference in old Testiment and the words and teachings of Jesus scriptures and time on earth …

Scroll to Top