bias

Critical Thinking In The Age Of AI.

~ Taran ~ Leave a comment

Critical thinking is the ability to suspend judgement, and to consider evidence, observations and perspectives in order to form a judgement, requiring rational, skeptical and unbiased analysis and evaluation.

It’s can be difficult, particularly being unbiased, rational and skeptical in a world that seems to require responses from us increasingly quickly.

Joe Árvai, a psychologist who has done research on decision making, recently wrote an article about critical thinking and artificial intelligence.

“…my own research as a psychologist who studies how people make decisions leads me to believe that all these risks are overshadowed by an even more corrupting, though largely invisible, threat. That is, AI is mere keystrokes away from making people even less disciplined and skilled when it comes to thoughtful decisions.”

“The hidden risk of letting AI decide – losing the skills to choose for ourselves‘”, Joe Árvai, TheConversation, April 12, 2024

It’s a good article, well worth the read, and it’s in the vein of what I have been writing recently about ant mills and social media. Web 2.0 was built on commerce which was built on marketing. Good marketing is about persuasion (a product or service is good for the consumer), bad marketing is about manipulation (where a product or service is not good for the consumer). It’s hard to tell the difference between the two.

Inputs and Outputs.

We don’t know exactly how much of Web 2.0 was shoveled into the engines of generative AI learning models, but we do know that chatbots and generative AI have become considered more persuasive than humans. In fact, ChatGPT 4 is presently considered 82% more persuasive than humans, as I mentioned in my first AI roundup.

This should at least be a little disturbing, particularly when there are already sites telling people how to get GPT4 to create more persuasive content, such as this one, and yet the key difference between persuasion and manipulation is whether it’s good for the consumer of the information or not – a key problem with fake news.

Worse, we have all seen products and services that had brilliant marketing but were not good products or services. If you have a bunch of stuff sitting and collecting dust, you fell victim to marketing, and arguably, manipulation rather than persuasion.

It’s not difficult to see that the marketing of AI itself could be persuasive or manipulative. If you had a tool that could persuade people they need the tool, wouldn’t you use it? Of course you would. Do they need it? Ultimately, that’s up to the consumers, but if they in turn are generating AI content that feeds the learning models in what is known as synthetic data, it creates it’s own problems.

Critical Thought

Before generative AI became mainstream, we saw issues with people sharing fake news stories because they had catchy headlines and fed a confirmation bias. A bit of critical thought applied could have avoided much of that, but it still remained a problem. Web 2.0 to present has always been about getting eyes on content quickly so advertising impressions increased, and some people were more ethical about that than others.

Most people don’t really understand their own biases, but social media companies implicitly do – we tell them with our every click, our every scroll.

This is compounded by the scientific evidence that attention spans are shrinking. On average, based on research, the new average attention span is 47 seconds. That’s not a lot of time to do critical thinking before liking or sharing something.

While there’s no real evidence that there is more or less critical thought that could be found, the diminished average attention span is a solid indicator that on average, people are using less critical thought.

“…Consider how people approach many important decisions today. Humans are well known for being prone to a wide range of biases because we tend to be frugal when it comes to expending mental energy. This frugality leads people to like it when seemingly good or trustworthy decisions are made for them. And we are social animals who tend to value the security and acceptance of their communities more than they might value their own autonomy.

Add AI to the mix and the result is a dangerous feedback loop: The data that AI is mining to fuel its algorithms is made up of people’s biased decisions that also reflect the pressure of conformity instead of the wisdom of critical reasoning. But because people like having decisions made for them, they tend to accept these bad decisions and move on to the next one. In the end, neither we nor AI end up the wiser…”

“The hidden risk of letting AI decide – losing the skills to choose for ourselves‘”, Joe Árvai, TheConversation, April 12, 2024

In an age of generative artificial intelligence that is here to stay, it’s paramount that we understand ourselves better as individuals and collectively so that we can make thoughtful decisions.

AI Reviewing Body Cam Footage, and AIs talking to themselves.

~ Taran ~ Leave a comment

There’s been a lot posted about artificial intelligence since I last wrote about it, but some of it was just hype and marketing whereas the really cool stuff tends to sit well. There’s two main topics that I’ll get out of the way with this post – more verbose topics coming this week.

Talking To Myself…

There’s been some thought about the ‘inner monologue’ that some of us have. Not all of us do have that inner monologue, and we don’t have a reason why yet, but apparently people who do have an inner monologue think that artificial intelligences can benefit from it.

They are finding ways that an inner monologue is beneficial for artificial intelligences, which may oddly help us understand our own inner monologues and lack of it.

If you want to read a bit more deeply into it, “Thought Cloning: Learning to Think while Acting by Imitating Human Thinking” is an interesting paper.

Having spoken to myself now and then over the years, I’m not sure it’s as productive as some think, but I’m not an expert and only have my own experience to base that off of. I do know from my own experience that it’s very easy to reinforce biases that way.

I do some thinking with language, but mainly my thinking is what I would best describe as ‘visually kinetic’, so I am pretty interested in this.

Reviewing Body Cams

One of the problems with any sort of camera system is reviewing it. It takes a long time to review footage, and an experienced eye to do it.

Police departments are turning to artificial intelligence to help with this. Given there is already real time facial recognition, on the surface this seems like a good use of it. However, there are problems with it as there are realistic concerns for communities of color, as well as related to data privacy. A running body cam collects every interaction, sure, but it also collects information on everybody involved in these interactions as well as the person accidentally getting into the frame.

With everything increasingly connected, watching the watchmen through body cams means watching the watchers of the body cam footage.

I wonder what their inner monologue will be like while reviewing hours and hours of boring footage.

An Example of Bias (ChatGPT, DALL-E)

~ Taran ~ 2 Comments

I was about to write up a history of my interactions with the music industry as far as ownership over at RealityFragments.com, and I was thinking about how far back my love for music went in my soundtrack of life. This always draws me back to “The Entertainer” by Scott Joplin as a starting point.

I could use one of the public domain images of Scott Joplin, someone I have grown to know a bit about, but they didn’t capture the spirit of the music.

I figured that I’d see what DALL-E could put together on it, and gave it a pretty challenging prompt in it’s knowledge of Pop Culture.

As you can see, it got the spirit of things. But there’s something wrong other than the misspelling of “Entertainer”. A lot of people won’t get this because a lot of people don’t know much about Scott Joplin, and if they were to learn from this, they’d get something wrong that might upset a large segment of the world population.

I doubled down to see if this was just a meta-level mistake because of a flaw in algorithm somewhere.

Well, what’s wrong with this? It claims to be reflecting the era and occupation of a ragtime musician, yet ragtime music came from the a specific community in the United States that are called African-Americans now, in the late 19th century.

That would mean that a depiction of a ragtime musician would be more pigmented. Maybe it’s a hiccough, right? 2 in a row? Let’s go for 3.

Well, that’s 3. I imagined they’d get hip-hop right, and it seems like they did, even with a person of European descent in one.

So where did this bias come from? I’m betting that it’s the learning model. I can’t test that, but I can go just do a quick check with DeepAI.org.

Sure, it’s not the same starting prompt, but it’s the same general sort of prompt.

Let’s try again.

Well, there’s definitely something different. Something maybe you can figure out.

For some reason, ChatGPT is racebending ragtime musicians, and I have no idea why.

    Public Domain Image of Scott Joplin.

There’s no transparency in any of these learning models or algorithms. The majority of the algorithms wouldn’t make much sense to most people on the planet, but the learning models definitely would.

Even if we had control over the learning models, we don’t have control over what we collectively recorded over the millennia and made it into some form of digital representation. There are implicit biases in our histories, our cultures, and our Internet because of who has access to what, who shares what, and these artificial intelligences using that information based only on our biases of past and present determines the biases of the future.

I’m not sure Scott Joplin would appreciate being whitewashed. Being someone respected, of his pigmentation, in his period, being the son of a former slave, I suspect he might have been proud of who he became despite the biases of the period.

Anyway, this is a pretty good example of how artificial intelligence bias can impact the future when kids are doing their homework with large language models. It’s a problem that isn’t going away, and in a world that is increasingly becoming a mixing pot beyond social constructs of yesteryear, this particular example is a little disturbing.

I’m not saying it’s conscious. Most biases aren’t. It’s hard to say it doesn’t exist, though.

I’ll leave you with The Entertainer, complete with clips from 1977, where they got something pretty important right.

From Wikipedia, accessed on February 1st 2024:

Although he was penniless and disappointed at the end of his life, Joplin set the standard for ragtime compositions and played a key role in the development of ragtime music. And as a pioneer composer and performer, he helped pave the way for young black artists to reach American audiences of all races.

It seems like the least we could do is get him right in artificial intelligences.

When Is An Algorithm ‘Expressive’?

~ Taran ~ Leave a comment

Yesterday, I was listening to the webinar on Privacy Law and the United States First Amendment when I heard that lawyers for social networks are claiming both that they have free speech as a network as a speaker, as well as claiming not to be the speaker and claiming they are simply are presenting content users have expressed under the Freedom of Speech. How the arguments were presented I don’t know, and despite showing up for the webinar I am not a lawyer1. The case before the Supreme Court was being discussed, but that’s not my focus here.

I’m exploring how it would be possible to claim that a company’s algorithms that impact how a user perceives information could be considered ‘free speech’. I began writing this post about that and it became long and unwieldy2, so instead I’ll write a bit about the broader impact of social networks and their algorithms and tie it back.

Algorithms Don’t Make You Obese or Diabetic.

If you say the word ‘algorithm’ around some people, their eyes immediately glaze over. It’s really not that complicated; a repeatable thing is basically an algorithm. A recipe when in use is an algorithm. Instructions from Ikea are algorithms. Both hopefully give you what you want, and if they don’t, they are ‘buggy’.

Let’s go with the legal definition of what an algorithm is1. Laws don’t work without definitions, and code doesn’t either.

Per Cornell’s Legal Information Institute, an algorithm is:

“An algorithm is a set of rules or a computational procedure that is typically used to solve a specific problem. In the case of Vidillion, Inc. v. Pixalate Inc. an algorithm is defined as “one or more process(es), set of rules, or methodology (including without limitation data points collected and used in connection with any such process, set of rules, or methodology) to be followed in calculations, data processing, data mining, pattern recognition, automated reasoning or other problem-solving operations, including those that transform an input into an output, especially by computer.” With the increasing automation of services, more and more decisions are being made by algorithms. Some examples are; criminal risk assessments, predictive policing, and facial recognition technology.”

By this definition and perhaps in it’s simplest form, adding two numbers is an algorithm, which also fits just about any technical definition out there. That’s not at issue.

What is at issue in the context of social networks is how algorithms impact what we view on a social networking website. We should all understand in the broad strokes that Facebook, Twitter, TikTok and their ilk are in the business of showing people what they want to see, and to do this they analyze what people view so that they can give people what they want.

Ice cream and brownies for breakfast, everyone!

Let’s agree every individual bit of content you see that you can act on, such as liking or re-transmitting, is a single item. Facebook sees you like ice cream, Facebook shows you posts of ice cream incessantly. Maybe you go out and eat ice cream all the time because of this and end up with obesity and diabetes. Would Facebook be guilty of making you obese and diabetic?

Fast food restaurants aren’t considered responsible for making people obese and diabetic. We have choices about where we eat, just as we have choices about what we do with our lives outside of a social network context. Further, almost all of these social networks give you options to not view content, from blocking to reporting to randomly deleting your posts and waving a finger at you for being naughty – without telling you how.

Timelines: It’s All A Story.

As I wrote elsewhere, we all choose our own social media adventures. Most of my social networks are pretty well tuned to feed me new things to learn every day, while doing a terrible job of providing me information on what all my connections are up to. It’s a real estate problem on social network sites, and not everyone can be in that timeline. Algorithms pick and choose, and if there are paid advertisements to give you free access, they need space too.

Think of it all as a personal newspaper. Everything you see is picked for you based on what the algorithms decide, and yet all of that information is competing to get into your eyeballs, maybe even your brain. Every story is shouting ‘pick me! pick me!’ with catchy titles, wonderful images, and maybe even some content – because everyone wants you to click to their website so you can hammer them with advertising.4

Yet when we step back from those individual stories, the social networking site is curating things in a chronological order. Let’s assume that what it thinks you like to see the most is at the top and it goes down in priority based on what the algorithms have learned about you.

Now think of each post as a page in a newspaper. What’s on the front page affects how you perceive everything in the newspaper. Unfortunately, because it’s all shoved into a prioritized list for you, you get things that are sometimes in a strange order, giving a weird context.

Sometimes you get stray things you’re not interested in because the algorithms have grouped you with others. Sometimes the priority of what you last wrote about will suddenly have posts related to it covering every page in that newspaper.

You might think you’re picking your own adventure through social media, but you’re not directly controlling it. You’re randomly hitting a black box to see what comes out in the hope that you might like it, and you might like the order that it comes in.

We’re all beta testers of social networks in that regard. They are constantly tweaking algorithms to try to do better, but doing better isn’t necessarily for you. It’s for them, and it’s also for training their artificial intelligences more than likely. It’s about as random as human interests are.

Developing Algorithms.

Having written software in various companies over the decades, I can tell you that if there’s a conscious choice to express something with them, to get people to think one way or the other (the point of ‘free speech’), it would have to be very coordinated.

Certain content would have to be weighted as is done with advertising. Random content churning through feeds would not fire things off with the social networking algorithms unless they manually chose to do so across users. That requires a lot of coordination, lots of meetings, and lots of testing.

It can be done. With advertising as an example, it has been done overtly. Another example is the last press against fake news, which has attempted to proactively check content with independent fact checkers.

Is that free speech? Is that freedom of expression of a company? If you look at this case again, you will likely draw your own conclusions. Legally, I have no opinion because I’m not a lawyer.

But as a software engineer, I look at it and wonder if this is a waste of the Court’s time.

  1. It should be in the interest of software engineers and others about the legal aspects of what we have worked on and will work on. Ethics are a thing. ↩︎
  2. It still is, and I apologize if it’s messy. This is a post I’ll likely have to revisit and edit. ↩︎
  3. Legal definitions of what an algorithm is might vary around the world. It might be worth searching for a legal definition where you are. ↩︎
  4. This site has advertising. It doesn’t really pay and I’m not going to shanghai viewers by misrepresenting what I write. It’s a choice. Yet to get paid for content, that’s what many websites do. If you are here, you’re appreciated. Thanks! ↩︎

AI, Confirmation Bias and Our Own Insanity.

~ Taran ~ 1 Comment

In unsurprising news, if you feed artificial intelligences the output of artificial intelligences they become a bit insane. I’d covered that before in Synthetic Recursion, which seemed pretty intuitive even before I wrote that, but scientists at Rice and Stanford University wrote a paper: “Self Consuming Generative Models Go MAD“.

So, we can say that’s been verified.

What’s even worse is apparently, Taylor Swift, Selena Gomez and Kim Kardashian have been saying things that they did not say – organized disinformation that has appeared all over, and if in vacuuming copyrighted content OpenAI’s ChatGPT might get infected. It’s not just artificial intelligences, output from people willfully misleading others can easily make it in.

Fortunately, I verified with ChatGPT4 and they got it right by… using Bing. I don’t use Bing. Why does ChatGPT4? Because of the same reason you can’t have a Coke with your Kentucky Fried Chicken.

While this time it has been caught – it started in November 2023 – it demonstrates how inaccuracies can crop up, how biases can be pushed, and how many problems we still have with misinformation without involving artificial intelligence. Every time we get anything on social media these days we have to fact check, and then we immediately get blowback about fact checking being flawed.

Why? It fits their confirmation biases. Given the way large language models are trained, we can say that we’re getting a lot right and yet we’re collectively also under a delusion that humanity has collected is right. What is true is that what we believe we know just hasn’t been proven wrong yet, with different thresholds for that varying from person to person.

With science, there’s a verification process, but science has been under fire increasingly because of who pays for the papers to be written and published. Academia has to be funded, and we don’t fund that as much as we should so others do sometimes, to their own ends. That’s why it’s important to read the papers, but not everyone has the time to do that. There is good science happening, and I’d like to think more of it is good than bad.

With AI tools, I imagine more papers will be written more quickly, which creates a larger problem. Maybe even an exponentially larger problem.

We accept a lot, and since we don’t know what’s in learning models, we don’t know what has been verified until we find things that aren’t. This means we need to be skeptical, just like when we use Wikipedia. There are some people who don’t like doing that footwork because what they see fits their confirmation biases.

Should we be surprised that our tools would have them too based on what we feed them?

It’s almost as if we need to make sure we’re feeding these learning models with things of value. That should come at a cost, because when we write, when we express ourselves in any way, it’s based largely on experience, sometimes hard won.

Meanwhile, artificial intelligence tools are being created to write summaries of books authors took years to write. Amazon is being flooded with them, apparently, and if I see another advertisement about microlearning on Facebook that seems to use these sort of precis notes, I might throw up through my monitor on someone else’s keyboards.

The Age of Context.

~ Taran ~ 1 Comment

While the information economy has thrived for a distinct few, we’re now in an age of context. It’s hard to explain this to people around me sometimes because many still were blissfully unaware that information was the new oil from the late 1990s.

Consider that there hasn’t been as much uproar as one would expect in Trinidad and Tobago over the TSTT data breach, where scanned copies of identification and other documents. A few people I’ve interacted with have been laissez-faire about the whole thing not because there’s nothing that they can do about it1 but because they themselves don’t see the value of the information.

Social networks have capitalized on this. I still have family members and friends who do those wonky quizzes on Facebook that allow other parties access to information from their Facebook account. ‘X’, still better known as Twitter, also does the same. Data in large quantities of vast numbers of people and their interactions becomes information. In turn that information requires context, which I have written about before, even in the context of AI and synthetic recursion.

We start with data, we process within a context to provide information, and an artificial intelligence is then fed the information, or data, to create results for a user.

This, now, is being fed into AI. It seems harmless to many people.

John Hagel recently wrote about what is missing in artificial intelligence. He brings up some very important points, such as the trust in large global organizations diminishing and that impact on how much information people will knowingly share.

The context I provided above with one of the data breaches in Trinidad and Tobago doesn’t seem to jive with that on the surface, but it does. I can stand in a retailer line and listen to people of all political stripes income levels complain about corruption in government. There may not be agreement on who or what the problem is, but there seems consensus that the government isn’t something that they trust.

It’s the same in the United States, and just about everywhere I have been or have heard from.

…People will increasingly embrace technology tools that can help them be much more selective in providing access to their data. This continues to be a big opportunity for a new kind of business that I called “infomediaries” – businesses that will become trusted advisors and manage our data on our behalf (I wrote about this in my book, Net Worth)…

“What’s Missing in Artificial Intelligence?“, John Hagel, JohnHagel.com, Nov 27th 2023.
His affiliate links remain, not mine.

This is where it gets interested. If people slow the flow of their personal information into social networks, social media, apps, etc, the information is supposed to become stale and dated, like that print version of the Encyclopedia Brittanica some of us grew up with. This is where John points out something important.

Explicit knowledge is knowledge that we can express and communicate in words. Tacit knowledge is knowledge that is embodied in our actions, but that we would find very challenging to express. It’s about knowledge that we acquire when dealing with real-life situations and seeking to find ways to have increasing impact. Some tacit knowledge is long-lasting – it involves mastering enduring skills and practices and cannot be acquired by reading books or listening to lectures. Those who have mastered these skills and practices find it very hard to explain everything they do.

Here’s the challenge – in a rapidly changing world, tacit knowledge increases in proportion to explicit knowledge.

“What’s Missing in Artificial Intelligence?“, John Hagel, JohnHagel.com, Nov 27th 2023.

This is where context comes in.

When you interact with some well designed software that is actually trying to help you, one of the things that software engineers look for is the user intention. The user intention exists in a context, and the combination of the intention and context can allow for better results if done properly.

To make the point, I’ll quote an article that could itself be biased, but provides a context. I won’t say it’s a good or bad context, I’ll let readers decide.

This showed up in my Facebook memories from 2012, which I found noteworthy given it’s 2023.

…We turn now to look at a stunning new exposé on how Israel is using artificial intelligence to draw up targets and how Israel has loosened its constraints on attacks that could kill civilians. One former intelligence officer says Israel has developed a, quote, “mass assassination factory.”…

“Israel Is Using Artificial Intelligence to Generate Military Targets in Gaza“, Amy Goodman, Truthout, Dec 1st, 2023.

I’m not going to try to navigate the Israeli-Palestinian conflict here, I just offered an example of a context which information can be used.

The context for the Israelis, if this is all true, is picking targets, and the intent would be to minimize what we call collateral damage. The stakes are high.

The articles accuse Israel of weakening their intention with their artificial intelligence that is selecting targets, which is an example of widening or broadening results based on intention in the user’s (Israel’s) context. They could tighten it for less casualties as well, but in doing so may not be able to hit targets that might be fighting with them.

That this is being automated at all is probably at best a little disturbing. Is it true? I don’t know. I recognize biases, and as John pointed out, trust is an issue.

Is it fair to use this as an example? I think it is, whether true or not, because it also points out some of the stakes we might be talking about with artificial intelligence. Context matters.

In theory, I could agree with this sort of use of artificial intelligence to defend civilians. If you look around throughout the media, though, you’ll find the entire Israeli/Palestinian issue to be more divisive than most other things and therefore contexts vary. Intentions vary. I can honestly say that though I know much about it I do not know enough to write about it. It is not my story to write.

In that regard, it’s a perfect example because we can also see how contexts and intentions can shift over time, and not everyone has the same contexts. Intentions shift over time, and not everyone has the same intentions.2

That example goes a bit beyond whether when asking an AI to generate an image of a person and the result is a person of European descent. This is despite the fact that people of European descent constituting 16% of the global population, which probably has 84% of the global population wondering why they have to type in specifics all the time. There’s lots of bias, as I have written about before.

Populations change. At one point, it was said that people of European ancestry constituted 38% of the global population, but during that period the available information wasn’t what we have now and the global population was smaller with Europeans doing the most emigration3.

Contexts and intentions shift, and what is important today may not be important tomorrow. The ability to shift and adapt to contexts and intentions is something I agree should be a part of the future of the world – and by extension, artificial intelligence, though present iterations of AI are working toward centralizing when the world will require more decentralization to allow for the diversity needed to tackle hard problems.

John calls it the rise of the Age of Context, and maybe we can also add the Age of the Acknowledgement of the Diversity of Intentions. His is snappier, though.

  1. This is largely because the government of Trinidad and Tobago hasn’t gotten the Data Protection out or even made it an apparent priority, demonstrating a lack of understanding of the value of information. ↩︎
  2. Personally, I just shake my head at the whole thing because this should not be continuing, regardless of which side one believes is right. If you held a gun to my head, I’m on the side of the innocent civilians of both sides, then someone will say the civilians aren’t innocent, and it gets worse from there. So let’s not start that. ↩︎
  3. European emigration is fascinating by itself. ↩︎

Damnatio Memoriae

~ Taran ~ Leave a comment

In discussion with another writer over coffee, I found myself explaining biases in the artificial intelligences – particularly large language models – as something that is recent. Knowledge has been subject to this for millenia.

Libraries have long been considered our centers of knowledge. They have existed for millenia and have served as places of stored knowledge for as long, attracting all manner of knowledge to their shelves.

Yet there is a part of the library, even the modern library, which we don’t hear as much about. The power of what is in the collection.

‘Strict examination’ of library volumes was a euphemism for state censorship

Like any good autocrat, Augustus didn’t refrain from violent intimidation, and when it came to ensuring that the contents of his libraries aligned with imperial opinion, he need not have looked beyond his own playbook for inspiration. When the works of the orator/historian Titus Labienus and the rhetor Cassius Severus provoked his contempt, they were condemned to the eternal misfortune of damnatio memoriae, and their books were burned by order of the state. Not even potential sacrilege could thwart Augustus’ ire when he ‘committed to the flames’ more than 2,000 Greek and Latin prophetic volumes, preserving only the Sibylline oracles, though even those were subject to ‘strict examination’ before they could be placed within the Temple of Apollo. And he limited and suppressed publication of senatorial proceedings in the acta diurna, set up by Julius Caesar in public spaces throughout the city as a sort of ‘daily report’; though of course, it was prudent to maintain the acta themselves as an excellent means of propaganda.

“The Great Libraries of Rome“, Fabio Fernandes, Aeon.com, 4 August 2023

Of course, the ‘editing’ of a library is a difficult task, with ‘fake news’ and other things potentially propagating through human knowledge. We say that history is written by the victors, and to a great extent this is true. Spend longer than an hour on the Internet and you may well find something that should be condemned to flame, or at least you’ll think so. I may even agree. The control of information has historically been central, and nothing has changed in this regard. Those who control the information control how people perceive the world we live in.

There’s a fine line between censorship and keeping bad information out of a knowledge base. What is ‘bad’ is subjective. The flat earth ‘theory’, which has gained prominence in recent years, is simply not possible to defend if one looks at the facts in entirety. The very idea that the moon could appear as it does on a flat earth would have us re-examine a lot of science. It doesn’t make sense, so where is the harm in letting people read about it? There isn’t, really, and is simply a reflection on how we have moved to such heights of literacy and such lows of critical thought.

The answer at one time was the printing press, where ideas could be spread more quickly than the manual labor, as loving as it might have been, of copying books. Then came radio, then came television, then came the Internet – all of which have suffered the same issues and even created new ones.

What gets shared? What doesn’t? Who decides?

This is the world we have created artificial intelligences in, and these biases feed the biases of large language models. Who decides what goes into their training models? Who decides what isn’t?

Slowly and quietly, the memory of damnation memoriae glows like a hot ember, the ever present problem with any form of knowledge collection.

Bias, Color, and Stereotypes Shown by Buzzfeed.

~ Taran ~ 3 Comments

Buzzfeed had a post last week that I thought I’d let soak for the comments before I wrote anything about it. The image at top are 4 images from their post, made much smaller to create one image for this one. The intent is to have the examples without the quality, which you can see on their post – all 50 of them.

I noticed a few things right off the bat in the images in, “I Asked AI What Europeans Think Americans From Every Single State Look Like, And The Results Are Just Plain Mean“, and it’s almost like a Norman Rockwell caricature of each state.

Two of the states are likely easily identifiable by most Americans, 2 maybe not. As an American, Louisiana and Idaho are pretty easy. What are the other two? Go see the Buzzfeed post.

Not all of them are that bad. However, there are only people of European descent in the images. They do seem pretty consistent about how the United States has portrayed itself in some ways. I also admire that the author had the time to work in 50 states. I find even the thought of doing that boring.

The comments, though, are pretty interesting to read.

The first comment in this thread is that the artificial intelligence doesn’t think ‘people of color’ exist… yet the first reply to that is that it’s ‘anti-white for sure’. We live in a world when both perspectives aren’t necessarily wrong.

Is this the trouble with AI and bias? Or is it the trouble with us and bias?

I’ll offer it’s both.

Why aren’t there people of non-European descent in there? I have no idea. I have some ideas.

For example, between the 1940s and 1990s, physical film was biased itself.

Since many images are scanned images, and not all images of darker skin tones are flattering – National Geographic is probably the only magazine of that era that seemed to work on it more resolutely – I don’t think that images from that period would not be biased.

Then, there’s the media bias. When I looked at what Buzzfeed’s generated image for Florida (it’s one of the 4 in the top), I could swear I had met that guy somewhere. He’s an amalgamation of many people I have known over the years, and not in a bad way at all.

What other reasons would there be? Well, if you were to ask me about stereotypes by state of ‘people of color’, I couldn’t come up with one. There are differences, of course, but they aren’t as apparent.

Not one Native American in the bunch. New Mexico, though, does seem to represent aliens.

The bias may also be because of what Europeans normally see of America, and that can be an issue of (1) What Europeans want to see, (2) What media portrays, and (3) What is true.

How artificial intelligences see us, though, might be more interesting to ask them. If you describe yourself to an artificial intelligence, I believe the further you are from the norm of the data, the more descriptive you’ll have to be.

AI’s are just systems, and not very smart ones right now. Perhaps we should watch what we feed them, but we haven’t been very good at what we feed ourselves, so I’m not sure how we should proceed.

Exploring Beyond Code 2.0: Into A World of AI.

~ Taran ~ 2 Comments

It’s become a saying on the Internet without many people understanding it: “Code is Law”. This is a reference to one of the works of Lawrence Lessig, revised already since it’s original publication.

Code Version 2.0 dealt with much of the nuances of Law and Code in an era where we are connected by code. The fact that you’re reading this implicitly means that the Code allowed it.

Here’s an example that weaves it’s way throughout our society.

One of the more disturbing things to consider is that when Alexis de Tocqueville wrote Democracy in America 1, he recognized the jury as a powerful mechanism for democracy itself.

“…If it is your intention to correct the abuses of unlicensed printing and to restore the use of orderly language, you may in the first instance try the offender by a jury; but if the jury acquits him, the opinion which was that of a single individual becomes the opinion of the country at large…”

Alexis de Tocqueville, Volume 1 of Democracy in America, Chapter XI: Liberty of the Press In the United States (direct link to the chapter within Project Gutenberg’s free copy of the book)

In this, he makes the point that public opinion on an issue is summarized by the jury, for better and worse. Implicit in that is the discussion within the Jury itself, as well as the public opinion at the time of the trial. This is indeed a powerful thing, because it allows the people to decide instead of those in authority. Indeed, the jury gives authority to the people.

‘The People’, of course, means the citizens of a nation, and within that there is discourse between members of society regarding whether something is or is not right, or ethical, within the context of that society. In essence, it allows ethics to breathe, and in so doing, it allows Law to be guided by the ethics of a society.

It’s likely no mistake that some of the greatest concerns in society stem from divisions in what people consider to be ethical. Abortion is one of those key issues, where the ethics of the rights of a woman are put into conflict with the rights of an unborn child. On either side of the debate, people have an ethical stance based on their beliefs without compromise. Which is more important? It’s an extreme example, and one that is still playing out in less than complimentary ways for society.

Clearly no large language model will solve it, since the large language models are trained with implicitly biased training models and algorithms which is why they shouldn’t be involved, and this would likely go for general artificial intelligences of the future. Machine learning, or deep learning, learns from us, and every learning model is developed by it’s own secret jury whose stewed biases may not reflect the whole of society.

In fact, they would reflect a subset of society that is as disconnected from society as the companies that make them, since the company hires people based on it’s own values to move toward their version of success. Companies are about making money. Creating value is a very subjective thing for human society, but money is it’s currency.

With artificial intelligence being involved in so many things and with them becoming more and more involved, people should at the least be concerned:

  • AI-powered driving systems are trained to identify people, yet darker shades of humanity are not seen.
  • AI-powered facial recognition systems are trained on datasets of facial images. The code that governs these systems determines which features of a face are used to identify individuals, and how those features are compared to the data in the dataset. As a result, the code can have a significant impact on the accuracy and fairness of these systems, which has been shown to have an ethnic bias.
  • AI-powered search engines are designed to rank websites and other online content according to their relevance to a user’s query. The code that governs these systems determines how relevance is calculated, and which factors are considered. As a result, the code can have a significant impact on the information that users see, and therefore what they discuss, and how they are influenced.
  • AI-powered social media platforms are designed to connect users with each other and to share content. The code that governs these platforms determines how users are recommended to each other, and how content is filtered and ranked. As a result, the code can have a significant impact on the experiences of users on these platforms – aggregating into echo chambers.

We were behind before artificial intelligence reared it’s head recently with the availability of large language models, separating ourselves in ways that polarized and made compromise impossible.

Maybe it’s time for Code Version 3.0. Maybe it’s time we really got to talking about how our technology will impact society beyond a few smart people.

1 This was covered in Volume 1 of ‘Democracy in America‘, available for free here on Project Gutenberg.

The Ongoing Saga of the ‘AI’pocalypse

~ Taran ~ 4 Comments

I ran across an surprisingly well done article on the AIpocalypse thing, which I have written about before in ‘Artificial Extinction’, and it’s worth perusing.

“…In his testimony before Congress, Altman also said the potential for AI to be used to manipulate voters and target disinformation were among “my areas of greatest concern.”

Even in more ordinary use cases, however, there are concerns. The same tools have been called out for offering wrong answers to user prompts, outright “hallucinating” responses and potentially perpetuating racial and gender biases.”

“Forget about the AI apocalypse. The real dangers are already here“, CNN, Catherine Thorbecke, June 16th, 2023.

Now, let me be plain here. When they say an AI is hallucinating, that’s not really true. Saying it’s ‘bullshitting’ would be closer to true, but it’s not even really that. It’s a gap in the training data and algorithms made apparent by the prompt you give it. It’s not hallucinating. They’re effectively anthropomorphizing some algorithms strapped to a thesaurus when they say, ‘hallucinating’.

They’re trying to make you hallucinate, maybe, if we go by possible intentions.

“…Emily Bender, a professor at the University of Washington and director of its Computational Linguistics Laboratory, told CNN said some companies may want to divert attention from the bias baked into their data and also from concerning claims about how their systems are trained.

Bender cited intellectual property concerns with some of the data these systems are trained on as well as allegations of companies outsourcing the work of going through some of the worst parts of the training data to low-paid workers abroad.

“If the public and the regulators can be focused on these imaginary science fiction scenarios, then maybe these companies can get away with the data theft and exploitative practices for longer,” Bender told CNN…”

“Forget about the AI apocalypse. The real dangers are already here“, CNN, Catherine Thorbecke, June 16th, 2023.

We don’t like to talk about the intentions of people involved with these artificial intelligences and their machine learning. We don’t know what models are being used for the deep learning, and to cover that gap of trust, words like ‘hallucinating’ are much more sexy and dreamy than, “Well, it kinda blew a gasket there. We’ll see how we can patch that right up, but it can keep running while we do.”

I’m not saying that’s what’s happening, but it’s not a perspective that should be dismissed. There’s a lot at stake, after all, with artificial intelligence standing on the shoulders of humans, who are distantly related to kids who eat tide pods.

We ain’t perfick, and thus anything we create inherits that.

I think the last line of that CNN article sums it up nicely.

“…Ultimately, Bender put forward a simple question for the tech industry on AI: “If they honestly believe that this could be bringing about human extinction, then why not just stop?””

“Forget about the AI apocalypse. The real dangers are already here“, CNN, Catherine Thorbecke, June 16th, 2023.

That professor just cut to the quick in a way that had me smile. She just straight out said it.

And.

When we talk about biases, and I’ve written about bias a lot lately, we don’t see all that is biased alone. In an unrelated article, Barbara Kingsolver, the only 2 time winner of the Women’s Prize for fiction, drew my attention to one that I hadn’t considered in the context of deep learning training data.

“…She is also surprisingly angry. “I understand why rural people are so mad they want to blow up the system,” she says. “That contempt of urban culture for half the country. I feel like I’m an ambassador between these worlds, trying to explain that if you want to have a conversation you don’t start it with the words, ‘You idiot.’”…”

“Barbara Kingsolver: ‘Rural people are so angry they want to blow up the system’“, The Guardian, Lisa Allardice quoting Barbara Kingsolver, June 16th, 2023

She’s not wrong – and the bias is by omission, largely, on both the rural and urban sides (suburbia has a side too). So how does one deal with that in a training model for machine learning?

We’ve only scratched the surface, now haven’t we? Perhaps just scuffed.