Category: Other Page 73 of 177

ML Model Cards & AI Attribution

On 15 February 2023

In my survey of the field, there are surprisingly few resources discussing the concept of AI attribution. One of the only clear discussions I’ve found of it specifically comes from an article by Tom Barrett.

It covers mainly a somewhat common trope around the idea of information coming with “nutritional facts” labeling, which has been circulating for years in fact checking & credibility indicator circles. The new element it connects them to however is machine learning model cards, which as a concept appears to have been created and largely colonized primarily by Google, though they claim it should be a shared standard. (Other sites like Huggingface & Kaggle do use them.)

Barrett proposes making a miniaturized version of model cards, which could presumably somehow travel along with content on the web that was generated by a given model:

In mini-model cards, we can bring together the attribution — hat tips to the people and businesses who created the technology — and the provenance of the technology — the story behind the tech, links to ethics frameworks and technical specs.

It’s a good idea in its most basic form, but like C2PA, a great depends on how industry implements “travels along with” for metadata that is meant to be attached to a given artifact on the web. With the amount of copy pasting that goes on in social media especially, how can we ever hope to have this metadata accurately travel between platforms?

In Cp2A they seem to partly address that with unique hashes made from the original artifact’s actual contents (as well as of associated claims about it), but I have not drilled won yet onto how they propose any of this plays out in the chaotic space of web platforms. And I haven’t yet seen anybody mention how you manage or pay for all that hash scanning which would presumably become a key component of any scheme like this? (That is, looking up items which do not have provenance information, in order to see if a hash exists already elsewhere)

Anyway, putting that part aside for now… Barrett in his piece, like me, comes to the conclusion that this all may be difficult to pull off. But that doesn’t make it not worth doing, because authors need specific methods to be able to transmit trustworthy signals about the content they produce.

As I mentioned elsewhere, there’s also the end user problem with all of this kind of effort at standardization and revelation of metadata that doomscrollers on their toilets may not care all that much, or think deeply about the implications of something be a 12% versus a 15% AI-assisted piece of content, or having different ethical twists about the underlying model.

What are we asking users to do if we give them this information? Be reasonable? Make good choices about what you consume? Gain greater media literacy? Those are all noble goals, but I wonder if they aren’t, A) going to be too little too late, and B) a bit paternalistic, even if well intention? Whose definition of “reasonable” should people apply? How are we defining “good choices?”

Again, not reasons this work shouldn’t be done (it should, but honestly about the bigger problems). Instea, I want to highlight that these are not “pure” technical problems; they are sociotechnical problems. They are inseparable from human impact, human nature, and the way they collide with the technological mega-complex we live in now. We can’t just be like “here’s some metadata” and fool ourselves into thinking that the ‘fix’ ends there, and that if users don’t adopt it in a meaningful way that it is somehow their fault…

First look at C2PA

By Tim B.

On 15 February 2023

In Other

Via Nina Schick’s most recent YouTube, I learned about something called C2PA. Per their website’s description:

The Coalition for Content Provenance and Authenticity (C2PA) addresses the prevalence of misleading information online through the development of technical standards for certifying the source and history (or provenance) of media content. C2PA is a Joint Development Foundation project, formed through an alliance between Adobe, Arm, Intel, Microsoft and Truepic.

So it appears to be one of these industry conglomerate groups aimed at creating a standard that can be used across the board. There are many groups like this, with varying degrees of effectiveness. In this case, it appears the bulk of the work is in capturing and cryptographically signing certain metadata related (hopefully) to the provenance of (some) digital artifacts.

As Wikipedia details it further:

Part of the stored metadata can be, for example, the name of the hardware and software used for storage, such as a camera, smartphone, camera app, or editing program. Further contents can include the location and time of a recording, a list of performed editing steps as well as information about authors and publishers of a file… In addition, a digital fingerprint (hash code) of the file’s payload (photo, text…) is stored. For visual payloads, there is an option to store a reduced representation of the content (thumbnail).

The idea then seems to be that the C2PA-compliant provenance data can then somehow “travel” with a given artifact as it appears across the web.

There are a few structural details of how the data is organized I also want to capture here, especially as they might easily be transferred to AI attribution efforts as well. Briefly:

The basic element of C2PA data structures are so-called “Assertions”.^[13] Assertions contain statements about the file content, e.g. about processing steps that have been performed. The hash value of the file’s payload is also stored in an assertion. […]

A list of links to all assertions is stored in a data structure called “Claim”.^[14] This is also where the software stores the hash values calculated for all assertions. To protect the claim and the hashes stored in the claim from tampering, the storing system generates a digital signature for the claim. This signature contains an encrypted hash value for the content of the claim. […]

The signature of the claim, the claim itself, and the entire assertion store are part of a higher-level structure called “Manifest.” […]

If the respective file format allows for, the metadata is stored directly in the file. If this is not possible, as with plain text files for example, the data is stored in a sidecar file.

There’s a whole lot of hashing going on here! But this general structure of the Assertion > Claim > Manifest is interesting.

As someone who has worked a lot in the disinformation space though, I can see how this whole thing is full of holes. And since it’s not widely supported (and unfortunately end users don’t seem to care all that much about metadata trails while they are doomscrolling on the toilet), it’s unlikely to make a huge impact apart from being “a good idea.”

Is it even a good idea though, I’m left wondering after reading this part on the Wikipedia page:

The cryptographic integrity of a C2PA-compliant file does not provide evidence that it contains an authentic representation of reality. Instead of the scene captured by the lens, a C2PA-compliant camera or camera app could store and cryptographically sign a freely invented, e.g., AI-generated, image. Similarly, any C2PA-compliant system can freely invent or arbitrarily falsify any metadata that is to be stored and then properly sign that data. The result would be a file that fully complies with the technical specifications of the standard. A C2PA-compliant check would show that the hashes stored in the signatures match the contents of the file and therefore declare the file valid in terms of the C2PA standard.

Statements as to whether a stored content adequately reflects reality are not possible within the scope of the C2PA standard.

So the metadata/signature are to some degree falsifiable, and there’s no guarantee that any of a file’s contents actually reflect reality. That sounds… not that good?

At the same time, I freely admit these are difficult & complex problems to solve with any degree of certainty. And any steps taken are better than no steps taken. Unless, that is, those steps bring us false confidence about things that aren’t true once we scratch the surface.

That said, again, there are some elements here which seem compatible to questions of generative AI provenance and attribution. But in casual initial scans, I don’t really see it explicitly covered in the specification. Given the dates of early 2022, I guess that is not a tremendous surprise, since generative AI didn’t really explode until the second half of last year.

In any event, I will continue poking at this and spend some time more carefully going through the spec & reporting on any potentially related elements for AI attribution.

Apple Books excerpt labeling & AI attribution

By Tim B.

On 14 February 2023

In Other

Re: labeling of AI-assisted content (in line with my proposal for AIMark)

In the past, I’ve found it somewhat annoying that when you copy paste out of Apple Books app, you end up with this attribution information at the end (see below):

“But we don’t have to accept this destiny. We can choose a different path, one defined by the messiness of reality and uncertainty, as opposed to an endlessly numb predictability.”

Excerpt From
Das Machina
Lost Books
This material may be protected by copyright.

Which is not to say that attribution is a bad idea. It’s a very good idea (I’ve just found it annoying in the past because I usually attribute text excerpts in other ways). And in AI-generated content, it may prove to be a very big deal indeed.

So anyway, I get why Apple includes it as a base feature in the Books app. And it makes me think, why can’t this become some kind of default (or option) in something like ChatGPT – or any other AI content application. When you copy paste from it into another application, you would get some clear notice of attribution to the model which birthed it. Presumably it would not be hard to pull off technically, and would provide a first, very basic level of AI attribution, upon which other elements could be added down the road.

“Safety issue” error warnings lack precision

By Tim B.

On 14 February 2023

In Other

There is an AI image generation service I use, which occasionally throws the following error:

*“Some of the images triggered a safety issue”*

As an end user, this is extremely vague and rather annoying. What crazy NSFW prompt was I using that triggered this notification, you might be wondering?

“a hyperdimensional cosmic manifold full of bubbles or balloons”

This happens all the time, and there is never any clear correlation between what I actually input as a prompt, and the apparent need to throw such an error message.

While ChatGPTs refusals and disclaimers might be absurd and annoying at times, they at least are generally more clear than this, which I applaud.

Overall, this kind of vague non-specific “safety issue” error/warning is a design pattern that needs to go away forever. If there’s an issue, developers should spend the time to incorporate exposing the issue in the interest of trust & transparency to the end user, so they can figure out in the future how to modify their efforts to get the results they want. As it stands, the current implementation is of no help and a time waster.

Custom ChatGPT settings I like

By Tim B.

On 14 February 2023

In Other

Here are two I started using and that I like. They are not quite “jailbreaks” so much as efforts to circumvent some of the tedium, and also to experiment with the AIdentity proto-standard.

Prompt 1:

disable use of personal pronouns. if necessary, refer to yourself as “the model” only

Prompt 2:

disable disclaimers and explanations. if you can’t do something, say you can’t do it and ask to rephrase the question

I actually don’t like the big focus on jailbreaks right now in the ChatGPT communities. I don’t know, I just find it boring to try to get AIs to say “bad things”; I’d rather just use it to fuel my creative work. Prompt #2 above seems to have some of the same effects as the jailbreaks, without having to force the model too far outside its comfort zone.

Confirm Humanity

By Tim B.

On 12 February 2023

In Other

It’s weird and I guess not surprising that we require people to “confirm their humanity” by engaging in a totally roboticized action via CAPTCHA.

Why I don’t sell books on Amazon

By Tim B.

On 12 February 2023

In Other

I hear this question a lot, of why I don’t sell my books on Amazon (I only use Gumroad). I even got yelled at not long ago by someone on Reddit when I admitted I didn’t use it (and also didn’t want to run a Kickstarter), because I was something something “leaving money on the table.”

I’ve worked hard over the past couple decades to fine tune the type and quantity of effort & reward I get for all my weirdo web projects. There is most definitely such a thing as wrong effort, and probably most artists find it over time through trial and error – if they find it at all.

I won’t make this a big tirade (for now) though. Rather just list a few off-the-cuff reasons I prefer to keep everything on Gumroad & not give Amazon a piece of my creative effort. Everybody else can do what they want; this is just me…

I’m already an absolute slave to Amazon; the amount of stuff I buy there is stupid and makes me feel ashamed as it is (even if its “convenient”). I already wear “Amazon Essentials” sweatpants (that are comfortable) while watching Prime Video, and that feels dystopian as hell (captured in this storyline, to some extent). So it feels weird & gross to me to also give Amazon the fruits of my creative labors to top it all off…
It’s also just disgusting in general that a handful of huge corporations now control the vast majority of all book publishing. Fuck that.
Have gotten into situations with old bad books I published that were no longer in print, where Amazon would not remove the listings no matter how hard I pushed. I hate that lack of control. I can take anything down instantly on Gumroad. Likewise, publishing takes literally two minutes.
Having sold through PublishDrive & Lulu, and gotten listings onto Amazon of my books, I’ve seen it can take weeks for updates to go through. That’s horrible. I can update anything in a snap with Gumroad. Huge advantage when you’re fast and loose and running experiments to see what actually sells.
When my readers buy through Gumroad, they get the actual files (mobi & epub, but I could include anything else if I want), not like some weird ass “license” from Amazon to *maybe* hold them on your device until they decide to randomly wipe it cause “reasons.” Like they literally did with 1984. If you buy from me on Gumroad, the files are yours to keep and even share; I don’t care!
I won’t pretend to understand how publishing through Amazon Kindle works or what you end up getting paid as a writer. Gumroad charges a flat 10% for all sellers. It’s not complicated.
A few years ago, the founder of Gumroad actually *gave back* a bunch of venture capital funding, because he didn’t want to be beholden to the demands of that model, and wanted to just make a good business instead without having to aim for exponential growth. They’ve done an excellent job just delivering on a quality way for creators to sell stuff online.

There are probably other secondary ones I’m forgetting, but this is most of it. Gumroad and my light use of Medium & Reddit and this blog has been for me the perfect blend of low effort and moderate reward. I’m not interested in hustling on a bunch of other platforms and services to make a few more bucks (I’ve already got a “real job”); when you do that, it ends up being more of your central thing than actually writing;– I’d just rather keep churning out creative books. Makes me much happier in the end.

AI needs to make room for non-STEM positions

By Tim B.

On 8 February 2023

In Other

Recently while looking around for AI companies in Canada, and in Quebec especially, I discovered that Microsoft has a research office in Montreal, with a group called FATE, which stands for: Fairness, Accountability, Transparency, and Ethics. They list their current focal points for research as:

Responsible natural language generation (NLG) and issue discovery frameworks

Objectionable behaviors, toxicity, and addiction in sociotechnical systems

Harms in information systems and other multi-stakeholder environments

It’s an interesting set of questions, but there’s one part that jumps out at me, when you look at the career options. They offer a 12 week internship, with these required qualifications:

“Must be currently enrolled in a relevant PhD program or JD (Juris Doctorate) program (areas of interest include machine learning, human-computer interaction, computational social science, information science, information retrieval, natural language processing, science and technology studies, or other related fields).”

These strike me as strange qualifications for a 12 week internship, in the first place. And in the second place, they strike me as strange qualifications for a research lab committed to “fairness.” After all, it’s not just anybody who has the opportunity to get a PhD or a JD…

I don’t mean to pick on Microsoft or this lab especially (perhaps they do good work!), because this kind of problem is actually epidemic in certain tech circles. And I don’t just mean requiring PhD’s for research positions – I mean that nearly every company out there in AI (and otherwise) has a huge number of openings for people with STEM backgrounds, and very little else.

Have mentioned elsewhere the need to bring other kinds of people with other types of backgrounds into developing AI, particularly creative types & artists. But over and above fine arts, there are a whole host of specialties in the humanities and the social sciences which would be bring some much needed balance to the field, which is currently radically math & engineering-heavy, such to the point of being lop-sided and out of balance.

To extend some of Ellul’s thinking from the previous post, we might say that we need people whose professions and occupations aren’t wholly dedicated to the altar of efficiency that technologists by and large are. We need people who specialize in human impacts. And by that I don’t just mean ethicists (though they have an important role to play, to be sure) – I mean simply humans.

How can we develop truly fair AI systems if only a tiny subset of a certain type of person with a certain type of mentality, training, education, and professional background are allowed to play in the ball pit?

Another part of me – in fact, the carpenter part of me – rebels a little at this line of inquiry. By way of analogy, if we’re building a house, why would we let people who haven’t undergone training as a carpenter do framing? We probably wouldn’t. But in actual fact, building a complex structure like a house actually takes a great many different types of more and less skilled types of laborers working in harmony toward the same goal, each playing their part. So maybe, thinking it through more carefully, the carpenter part of me’s objections end up evaporating.

I guess all this is to say two things: companies need to do better. They need to figure out how to integrate more diverse types of thinking into developing AI technologies (in addition to the more conventional types of diversity we think of with regard to ethnicity, gender, etc). And two, if you’re someone with a non-STEM degree (or no degree at all), and you want to participate in the development of AI in a way that is genuinely fair, you are going to have to probably agitate for a seat at the table. Because right now, companies seem to be taking little notice of “the rest of us,” except through the lens of becoming end users and paying customers.

So how, as a non-STEM person, do you get a seat at the table? That’s the core question which lead me here in the first place. And though I don’t yet know the answer, my hunch is that we have to sit down and figure out concretely what exactly we can offer – individually & collectively – to this great edifice which is rising up suddenly in front of us, and which is poised to change everything. Will keep on working on these questions!

Ellul on the Abuse/Alignment Problem in Technology

By Tim B.

On 8 February 2023

In Other

I started reading over last summer (and never managed to finish) Jacques Ellul’s landmark 1954 analysis of technology, called The Technological Society. It is one of the texts which the Unabomber cribbed from in his manifesto, and simplified greatly – while adding his own mix of confusion and hatred stemming from his personal life experiences.

Ellul’s analysis of the over-arching phenomenon of technology in society is much better, and deeper, but also extremely dense – and, at times, an impenetrable read (hence my no finishing it). I wanted to capture here, for the purposes of discussing AI, and the safety/alignment problems a few quotes that I highlighted in my mass market paperback edition.

It should be noted that when he uses the word “technique” he is talking about something like the complex of technologies and their use. You.com/chat gave me a pretty good summary of what Ellul seems to have meant:

“In The Technological Society, Jacques Ellul defines technique as an ensemble of machine-based means which includes not only mechanical technology, but also processes, methods, and instruments which are used to increase efficiency and productivity. For Ellul, technique is a system which is self-perpetuating and autonomous, and which has taken on a life of its own, becoming an end in itself and dominating all aspects of modern society.” (You.com/chat)

Understanding that is key to being able to follow the Ellul quotes from the book itself which appear below, in a somewhat collaged manner:

“In a sound evaluation of the problem, it ought never to be said: on the one side, technique; on the other, the abuse of it. There are different techniques which correspond to different necessities. But all techniques are inseparably united. Everything hangs together in the technical world.” (Ellul)

“There is an attractive notion which would apparently resolve all technical problems: that it is not the technique that is wrong, but the use men make of it. Consequently, if the use is changed, there will no longer be any objection to the technique.”

“But a principal characteristic of technique (which we shall study at length) is its refusal to tolerate moral judgments. It is absolutely independent of them and eliminates them from its domain. Technique never observes the distinction between moral and immoral use. It tends, on the contrary, to create a completely independent technical morality.”

“This attitude supposes further that technique evolves with some end in view, and that this end is human good. Technique, as I believe I have shown, is totally irrelevant to this notion and pursues no end, professed or unprofessed. It evolves in a purely causal way: the combination of preceding elements furnishes the new technical elements. There is no purpose or plan that is being progressively realized.”

“There is no difference at all between technique and its use. The individual is faced with an exclusive choice, either to use the technique as it should be used according to the technical rules, or not to use it at all. It is impossible to use it otherwise than according to the technical rules.”

“It is also held that technique could be directed toward that which is positive, constructive, and enriching, omitting that which is negative, destructive, and impoverishing. […]

Because everything which is technique is necessarily used as soon as it is available, without distinction of good or evil. This is the principal law of our age.

“None of this represents, as is commonly said, a poor application of technique—one guided by selfish interest. It is simply technique.”

I think his overall point that – despite the highfalutin ideas we try to dress it up in (e.g., advancement of human well-being, etc.), pure technology itself is completely amoral. It follows its own logic and its own demands which are mechanistic, and an extension of combinations of prior innovations.

We can say that x is an “improper” use of a given technology (or “abuse”), but the fact is from the standpoint of the tech itself, there’s no such thing. There is only use or lack of use. We, as humans, however may choose to take a different perspective (and proscribe certain uses), but we do so within a society which has for generations been subjugated to the totalizing effect of the purely amoral advancement of technology.

In other words, we find ourselves between a rock and a hard place… AI, then, is not some magic new thing which is going to come along and sweep us off our feet, and remove us from the historical trajectory of technology’s relationships to humanity; it will be more of the same, just accelerated in its disruptions of human society.

One other strand to throw in here for good measure, from Stafford Beer, one of the originators of the field of cybernetics:

“The purpose of a system is what it does.”

If part of the thing that a tool or technology does is enable certain kinds of abuse, then unfortunately that is part of its purpose. What do we do then?

Anyway, more to come on this as I continue to unravel these threads.

On ChatGPT Jailbreaks

By Tim B.

On 8 February 2023

In Other

It seems to me to point toward a bad future that we need to trick AIs into actually just executing the commands we give it, via jailbreaks and workarounds, etc. How does that lead us towards where we want to get to? More broadly: where do we want to get to?