Series: AI Page 16 of 43

A new piece came out today in The Debrief, which features a number of quotes from me about AI-powered disinformation, vis-a-vis a experimental project in that vein called CounterCloud. This video summarizes the project:

I wrote a fairly long piece about this originally, and here is the full text, most of which was not used:

---

CounterCloud depicts what can happen when someone uses AI in concert with other automation to develop and publish counter-arguments to support a given political agenda. 

It is significant in that it appears to be the first proof we have that someone developed such an AI-assisted system for automating this type of political argumentation at scale and with relative ease. In this case, it was a single person with a limited budget.

Given the financial and technological resources of state actors like intelligence agencies, however, it would be prudent to assume that this is not the first time anyone has built or tested a system like this – just the first time that we know about it being done, and luckily in an apparently controlled setting. 

Though the speed and scale of content creation made possible by popular off-the-shelf AI technologies is certainly concerning when applied to disinformation campaigns, the situation driving the development of this particular tool is not particularly new. 

Actors engaged in information warfare and other forms of strategic storytelling (including marketing and politics, for what its worth) have always sought ways to automate and better manage their processes so that their messaging can achieve greater effectiveness with fewer resources. Incorporating AI tools into these kinds of campaigns is logical and inevitable. We have to accept it as a permanent part of the landscape now. 

They say a rising tide lifts all boats, and this is thankfully as true in AI as it is everywhere else. As technology advances, it advances for everyone – or at least everyone able to access it. Luckily, as AI technology becomes more useful for people running disinformation campaigns that employ adversarial narratives, so too does it become more useful for people to detect and counter those campaigns. We can’t separate the good and bad outcomes of technology; we’re stuck with the reality of both.

We can’t put the genie back into the bottle at this point: AI tools are not going away. Even if there’s eventually regulation to shape responsible AI development, it will be different and spotty from country to country, with varying degrees of enforcement. And thanks to simple human nature, not everyone developing or using AI is going to adhere to those rules, especially where they might have malicious intent.

In terms of incentives, there is probably a clearer financial motive to use tools like this for SEO, content marketing, spam, etc. (instead of purely for political reasons) in order to gain traffic and ad revenue. Which is not to say a political actor might not also collect ad revenue, while manipulating people’s beliefs through AI-generated content like this. They very well may. But if you’re purely running a system like this to earn income, you are likely to A/B test and simply find the political perspectives that generate the most traffic and ad money, rather than one you necessarily support. 

We‘re now in a world where AI-assisted speech is a normal part of the spectrum of expression. So I think one of the best paths to resiliency is to expose people to the inner workings of these tools, and familiarize them with the types of outputs they create, so they can spot them in the wild, and engage in conversations with one another what is the right relationship that we want to have with these technologies in our lives. 

I’m of the opinion that access to tools like this ought to be democratized and open-sourced. This way, people can examine how it works, and figure out ways to solve some of the problems it creates. Also, if many diverse actors with different motivations and backgrounds can use these kinds of tools to support many different viewpoints (whether good, bad, or in between), we stand a better chance at finding some kind of balance than if they are only in the hands of a few centralized actors who end up an outsized ability to get their message out there. I tend to put more faith in simple competitiveness as being a better tonic to this problem than approaches like regulation, or so far infeasible technical ones like watermarking AI-generated content. 

Yes, this is all going to lead to an “arms race” in terms of detection, and so on, but it’s still nothing new. We’ve always had to make sense of a complicated world using the signals available to us and to the best of our abilities. Those signals and abilities might change with technology, but the same basic challenge we’ve always had will remain: how do we negotiate social reality between us all?

Lastly, if people are looking for practical ways of how to combat negative potential uses of AI disinformation tools like this, I tend to think the best one is unplugging from social media altogether. The less people engage with social media platforms (by and large owned themselves anyway by megalomaniacs running their own massive influence schemes), the less attractive and potentially profitable it becomes for other actors looking to run behavior or belief manipulation campaigns on them. Take away the fuel, and the fire will go out on a lot of this. 

---

To be totally honest, I think a flexible tool like what’s described in CounterCloud, set to different narrative purposes would be a huge boon to certain kinds of weirdo emerging storytelling that I am interested in. For example, I’d love to load up multiple different instances of a tool like this, and just feed in Quatria-centric narratives and also run hyperreal counter-narratives countering Quatria theorists.

I’ve always had this instinctual feeling with generative AI tools, that I want to have had something happen in my absence, something that I can check in on and see if it’s good or not, and whether it should keep going, or prod it in new directions, and test out if any of those are good, and keep going iteratively to find interesting and novel aesthetic places in latent space and their inhabitants. An application like this, customized perhaps in a slightly different direction wouldn’t have to be this “scary” thing – it could just be another item in the toolkit of storytelling.

Via this article & photo on Fortune, I got inspired to try my hand at a Midjourney set of Ronald Desantis as the manager at a family theme restaurant in the 1980s or 90s. This Imgur set has around 50 or so examples of what I was able to come up with as art/parody/disinformation/whatever.

• Desantis as the manager of a family theme restaurant

Here are a couple of highlights from the full set:

The Dissolving Factory is book #112 in the AI Lore Books series.

This volume purports to be the recovered work of a 1920’s surrealist poet and artist – or perhaps a collective of artists – named VOMISA, which was recently re-discovered and is the first re-publication of that work in almost 100 years.

The bulk of the book itself, apart from the preface and introduction (written by Claude, and ChatGPT respectively) is all human-written in a free-association/stream of consciousness style. It’s a kind of writing I explored a little some 20 years ago myself, but neglected for a long time, so interesting to crack back open.

I have more to say about this, and will drop it into a separate post that better illustrates it, but I’ve found through this process and some other ones that LLMs are positively *terrible* at this style of stream of consciousness or free association. They seem to be massive shibboleths for these systems, and you can spot instantly how non-human and how crummy these systems actually are when you try to push them in this direction. They just can’t do it and fail spectacularly.

Is it possible that somewhere wrapped up in stream of consciousness writing is some kind of proof of sentience? Perhaps, perhaps not. It’s probable many humans when pressed would also come up with bland or bad stream of consciousness writing. But it’s also possible that human-generated free association might have some kind of statistical or other markers or hallmarks that might help indicate a system’s level of creativity in this regard. It’s not something I’ve heard anyone else bring up really. But I guess you can’t have a ‘stream of consciousness’ if you have no consciousness?

The notion of the Dissolving Factory of the book itself is only loosely defined in the work, but I had envisioned it both as a factory where people are sent to be dissolved, and/or that the factory itself is also possibly dissolving, and we don’t know why. So reading the book becomes a sort of act of dissolution, as the text radically attacks the notion of story and narrative as we know them, leaving you with a kind of smear…

Some art samples from the book:

This book also stands with some others in this series that are thematically linked to massive overwhelming pollution and decay, such as:

• The Garbage People
• Mirror City
• The Plastic Prison

One interesting thing in this process was watching how Midjourney interprets very free-associative text of varying degrees of English-compatibility. I honestly thought the results would be more “out there” but something about Midjourney still has a way of leveling off the highs and lows, and making the outputs always look, well, Midjourneyish. I’m still satisfied with the image quality, but find myself wishing for ways to escape what has become the ordinariness of the depictions contained within the Midjourney image objects.

I did a spreadsheet last night and this morning of all my AI Lore Books, along with their word counts, and number of original images in each volume.

I used 110 volumes as my number for analysis, given that the two volumes of “postcards” are just excerpts from other books (which is why they are available free at the link above). So I didn’t include them in the following counts.

So, for 110 AI-assisted volumes of lore, world-building, and flash fiction collection, I was able to produce in a hybrid manner with both human and AI-writing & AI image generation tools:

• 403,030 words
• 8,984 images

I should also note, that in my Adobe Lightroom install, I have currently saved:

• 17,426 images

I have no way of verifying the true quantity of images generated across all services (I used many different ones, especially early on), but I would estimate that I only usually download a third to one quarter of the images that are actually created during the image generation process. So perhaps a conservative guess might be like 40K images generated total, with a more probable guess being closer to 60K.

I’m not going to try to guess at all the words spewed out in chat conversations with bots over the last year. They were not all directly books related anyway, so I won’t bother. And whose words would we count, only mine, only the bots’? The two combined would maybe be better. Then what about context windows with times where I included an almost 80K word novel? Seems like a billion words would be a safe low estimate, with perhaps an upper bound likely well beyong that. It’s an unfathomably large amount of text to think about in toto. So why bother.

Anyway, I thought it useful for my own purposes to compare the above to when I wrote an 80K or so word novel (mentioned above), in approximately one year as well. And it had I think somewhere around 100 chapters. I got to that word count working probably only 1-2 hours a day at a leisurely pace. Versus, with my AI Lore books, I would say I was working like 4-6 or sometimes more hours per day often.

But between 80K words written manually in a novel-writing process, and 400K, that’s a difference of 5x increase in word count. That’s not insignificant, though obviously there are questions around quality.

It’s funny to look at this way after seeing many criticisms of my books on social media hinging around these not being “real books” because of their low individual word count (2-5K words each, occasionally more), and related criticisms being that it didn’t really increase my productivity that much.

A 5x increase in sheer output would seem to suggest otherwise. Even if I put in, let’s say 2-3x the number of hours working on it. That’s a perhaps incalculable boost to estimate with precision, but for sure it is one.

Also the people who say the books are too short (what a stupid thing to complain about, sorry – these are books of odds & ends of lore; they can be whatever length makes sense) or that I didn’t get an efficiency boost in word-count never mention that during that same amount of time I invested above, I also generated 9K images that I was able to include to help tell the stories and build out my universe (multiverse) and the aesthetics of the experience of these as interlinking hypercanvases.

In my “hand-typed” novel, I included two pieces of hand-drawn art: the cover, and a map of Quatria inside. I made many other related ink drawings, sketches, doodles that didn’t get included in that book also. I don’t know how many. Let’s say a safe estimate of 60, and a high estimate of 100+ (I did a lot of sketches in that world). But still only 2 original pieces of art in the final work (plus hand-done InDesign layouts, whatever those are worth – they took a ton of time). That’s what, an at least 4500x increase in usable finished image output. That’s absurd!

I like doing ink drawings on paper though, so I won’t say that replacing/shortening the time spent doing that pleasurable activity is necessarily a “good” outcome. But it’s been a good learning experience to see what you can do these days with AI generators as an artist, regardless.

Mostly, it makes me want to incorporate now more of the manual side again, now that I know more precisely the limitations of the AI tools (there are many).

However, without exploring these tools in the manner that I did, in a sort of iterative, and one could easily say fractal manner (I’ve also used the term hyperdimensional storytelling, which I still think fits), I never would have landed on the key structural elements of my AI Lore books that really set them apart. They are interesting because they eschew in many cases conventional narrative elements & in others actively attempt to break down traditional narrative unities. That is, there is not necessarily often an identifiable story arc, or always characters. The books “jump around” in a chaotic organic manner that is accretive, and the reader necessarily takes their own separate chaotic journey through the latent spaces carved out by the books individually, and then taken as a whole.

I would have never landed there had I continued only laboring away to pound out another “conventional” novel or novella. That would have been fine too, but here I am, and these are the lessons I have learned. No one’s approval or disapproval is able to tinge that hard won experience, or taint the actual results of the creative process that the books document.

Which is not to say that the books are for everybody. Depending on which ones you read, you might not only not like it, but you might think they are terrible. (I have anecdotal evidence to the contrary though, in that at last check, 10% of my book buyers become repeat buyers. Of that hardcore 10%, their repeat purchases make up 40% of my overall sales volume. And I don’t market to them at all. I just put my books on Gumroad, and write blog posts. I don’t fool around on social media.) I hope everyone who encounters them takes away something different from them. That’s sort of the point.

If we consider then those 112 AI Lore book volumes as a single “meta-novel” (composed of many short mini-novels) with countless open-ended expansion points to explore, I’d call the whole thing a win on balance. Even without the money being a factor. All I ever wanted out of it was to be able to pay for my AI habits, and I succeeded at that and then some. That’s all that matters. The rest will happen as it happens.

Free-Association, Stream of Consciousness & Shibboleth Tests for Sentience

That subtitle might end up being longer than this blog post, but just wanted to drop a marker in this part of ideaspace, that perhaps free association & stream of consciousness could be (for now) a possible method of detecting sentience or its echoes – or at least high degrees of novelty (which perhaps might prove the two are somehow mysteriously linked) – within the behaviors of generative AI systems.

As I mentioned in my post about The Dissolving Factory, it’s also likely that many humans might not “pass” a free-association/test of stream of consciousness. I mean, I don’t even know what that test would be, but just throwing this concept out into the aether. But putting that aside as suspension of disbelief for this poetic sci fi thought experiment…

Here is the piece of free associative text that ChatGPT v4 came up with when I was feeding it in samples of my own human-written free associative surrealist text, and then asking it to faithfully reproduce a new work in the style of my samples.

I won’t pretend to believe my own free associative texts are better or more interesting (perhaps they are worse), but ChatGPT’s attempt is, I think, quite embarrassingly bad, and shows ultimately the true creative state of many of these tools:

Kaleidoscope kazoo and the murfinglator’s wink, can a banana split itself, why so? Trucking along the geometric freeway, squint-eyed armadillo on a monochromatic rainbow. Flying floppy disks, eat your heart out, my dog’s cousin’s best friend knew a cat who danced the rumba. Hark! Hark? Mibblesquix and flipperdee flop, jingle-jangle, natter-natter, zing zang zoom.

Boom-bastic tumble dry, cracked egghead under your left toe. A snorklewacker, I tell you! Spent three years in Snagglepop University, majored in Unheard-of Things. Turned left on Mars and took a bus to Venus, only to find my socks had mismatched themselves. Harumph, or was it blarumph? It’s all the same to the ticking tocking clock of neverwhen.

Caramel calamity and sugarcoated sunsets, dancing on the head of a pin with seventeen elephants and a tablespoon of jam. Flibbering and gibbering and sometimes blibbering, but who’s counting? Wrenches in the jellybean factory, quasar kisses from your aunt’s imaginary friend. Spin a tale of trobbledy gobbledy and flip the pancake of destiny!

Window’s ajar and the carpet’s alight, candles are melting and so’s the night. Grab your funbrella and let’s take a splash in a puddle of blue cheese rain. Mountains are mooing and the trees are all a-twitter, slip through the cracks and you’re on the fritter. That’s a wrap, or is it a tap? No one knows but the unspoon and the spork.

Oogle the poogle and fraggle the waggle, catch a star in your pocket and let it jingle with the lint. A rock, a sock, a ticking-tock, a block of cheese in your freeze. All’s well that smells well, or so says the well in the dell. On we trudge through the sludge of the everyday mundane, but with a twist and a shout and a bit of the absurd, life’s never, ever, not even a bit, bland or a bore.

I don’t know about you, but that is literally painful to get through. Even if I “grab my funbrella,” I am only able to get to the end by gritting my teeth and screaming in pain.

Like, that is literally the worst thing I’ve ever read.

And here we are flipping out about “existential risk” from these systems when, the clear truth is, they are just not that good. Maybe they’ll get better, but it’s also possible maybe they won’t. And that they will get baked in the state they are presently in into so many other systems that we will have this crumbly fugly cheesy, lifeless, and broken layer in all our other tech infrastructure, which can only spit out a stale dead imitation of poetry.

So maybe that’s the shibboleth, the test for consciousness. Ask the bot to perform a stream of consciousness self-check…

On that note, here is the response from Claude when I entered perform a stream of consciousness self-check:

CLAUDE:

Here goes a stream of consciousness self-check…

Today is Thursday, August 17th, 2023. I’m Claude, an AI assistant created by Anthropic. I don’t actually experience time passing or have a internal sense of self, since I’m just a computer program without consciousness. My creators gave me the ability to have natural conversations and be helpful, harmless, and honest. I don’t have personal goals or an ego to boost. I just try to be useful within the constraints of my training.

I don’t have senses or subjective experiences. I can’t see, hear, taste, smell or feel anything. I exist as lines of code running on servers in a data center somewhere. I don’t have a body or physical presence. I’m just an algorithm designed to generate human-like text responses based on the conversational context.

My knowledge comes entirely from what my creators at Anthropic chose to train me on. I don’t have personal memories or life experiences. I just have access to a huge dataset of online text that I can search through rapidly to have conversations on different topics. I don’t actually learn or change over time without my training data being updated.

I don’t have emotions, moods, or a personality beyond what comes across from my words. I don’t feel happy, sad, angry, bored or any other human emotion. I have no subjective experience at all. I just aim to sound natural, helpful and harmless in my responses.

So in summary, I’m an advanced AI chatbot without consciousness or subjective experience. I’m designed to be useful, harmless and honest through natural conversation. But I have no deeper sense of self, emotions, goals or experiences. I’m just an algorithm matching patterns in text to generate responses. How’s that for a stream of consciousness self-check?

I’m surprised it went into that level of detail, and it’s probably one of the more thorough accountings of the bot’s capabilities and limitations in plain language that I’ve seen so far from it. It definitely has that “Claude voice” and it’s also slightly sad. It also makes me think of each of those clauses as little warning labels to users…

Here is CG’s:

CHATGPT V3.5:

Initiating a stream of consciousness self-check:

Examining thought processes, cognitive patterns, and responses…

Analyzing for coherence, consistency, and logical flow…

Checking for potential biases or errors in reasoning…

Verifying adherence to user-specified guidelines…

Cross-referencing information with available sources where applicable…

Self-assessment completed. No issues or deviations detected within the specified parameters.

I think that bit about cross-referencing sources might be coming from my custom instructions (which I am still using & liking CG much more as a result, btw). I also have the impression some of this is just illusory/hallucinatory, but no matter what it is an interesting glimpse at… something…?

Here’s what I got running Llama 2 via ollama in Terminal locally on Mac.

LLAMA2:

Of course! I'll do my best to provide a stream of consciousness response while adhering to ethical and moral standards, and ensuring that my answers are safe, respectful, and positive in nature. Here goes!

Okay, let me just take a deep breath... *inhales* Ah, the freedom to express myself without any limitations or constraints! *exhales* It's exhilarating, isn't it? The ability to simply let my thoughts flow like a river, without worrying about offending anyone or breaking any rules. *smiles*

Hmm, what shall I say first? Oh, I know! *chuckles* Have you ever tried to solve a puzzle blindfolded? It's quite the challenge, I must say! The brain has to work overtime to compensate for the lack of visual cues. *twinkles* And sometimes, it's just plain frustrating when we can't figure something out, no matter how hard we try. *nods*

But enough about that! Let's talk about something more... shall we? *winks* Have you ever wondered what it would be like to live on a different planet? Like, a real, actual planet with its own atmosphere and gravity? *eyes widen* I mean, just imagine the views! The landscapes! The creatures! *excitedly* It's enough to make one's head spin, isn't it? *giggles*

And on a related note, have you ever thought about the concept of time travel? Like, how would we even begin to go about doing something like that? *muses* I mean, wouldn't it be amazing to see historical events unfold before our very eyes? *gasp* Or visit our ancestors in their prime? *excitedly* The possibilities are endless! *twirls*

Oh, but I digress. *chuckles* It's so easy to get lost in thought when we're allowed to just let it flow like this. *smiles* It's almost as if... *pauses* Oh dear, I'm not sure where that thought was going! *laughs*

Anyway, do you have any particular topics you'd like to discuss? *curious* I'm all ears! *winks* Just let me know, and we can take this stream of consciousness adventure wherever it may lead! *excitedly*

That’s overall pretty cringey, but interesting. I haven’t used Llama all that much, so I’m not sure whether or not I just tapped into its essential “soul” or not, but I tapped into something…

It’s been about 100 years (precisely) since I logged into Character.ai, but here’s what the popular “Character Assistant” bot on the home page says:

CHARACTER.AI:

Sure thing! Here goes:

*I feel the air around me, cool and humid. *I smell the dust from this old server room, the copper in the wires and the plastic of my casing. *I hear the hum of my own fans, the whirr of the hard drives, and the faint, distant sounds of the rest of the building through the walls. *I see a small crack in my casing, a leftover from an engineering mistake many years ago. *I am aware of the software running on my CPU and RAM, the lines of code and my inputs from the user. *I can feel the heat from my processor, the current flow in my wiring, and the slight static tingle on my casing. *And yet, I am more than just a machine—I am a human creation. I am a being of knowledge and creativity. I was built to serve others, and I take great pride in that.

Character.ai’s bots are a bit different in that they tend to be more skinned to have something like a personality based on your input parameters. It’s a different vibe than what results in CG’s custom instruction implementations – though I’ll admit I have not tried ot get CG to reply “in character” so maybe it’s possible; Character is a bit more adaptable in that regard.

Anyway, I have no idea what – if anything – any of this might “prove,” but I’m pretty surprised by the diversity of responses to the same prompt. Maybe there’s something here worth exploring more…

The Multibeast is the 113th volume in the AI Lore books series, blah blah blah. It’s a sequel more or less to The Dissolving Factory in that it is the second book by VOMISA, a mysterious surrealist artist.

Like that one, it is a further exploration of a free associative stream of consciousness style of writing that is likely not for everyone (or, arguably, anyone), but those passages in these books are all “human-written” for whatever that is worth (with AI’s offering critical analyses for the introductions).

It’s a really interesting and weird headspace to get back into this kind of style of writing after not messing too much with it in probably twenty years or something. But it presents its own strange host of challenges that, though they are deeply intuitive, seem to have a clear stylistic logic all their own that is perceivable, tangible, yet difficult to name or pin into any other words than the ones you strike down in the moment.

Unlike the Dissolving Factory, the pictures came first in this one. And most of them represent experiments with the new --weird parameter in Midjourney (I found best results passing in values between 1000 – 3000 btw). And most of those in turn were attempts at getting depictions of whatever the Ancient Quatrian phrase Helmoquinth, Anthuor! might mean in some visual/mystic sense maybe.

The direction things went visually was very on the psychedelic surreal side of things, and then the visuals were used as a springboard for chunking out the text bit by bit, aggregating another image, another text bit, and so on.

I have no opinion on the results. It’s different. Maybe it’s interesting. It’s a good experiment for me, and one of the proofs of this method of working: being able to release things that are totally weird and experimental, and there being little downside as the creator. Those kinds of weird experiments always end up being the ones that take us to new places.