Data Robots! GPT, "AI", and Slaves

"Robata": is it slave work, or drudge work?

Historically, it means both kinds of bothersome work. It's work that the well-off had (and perhaps have) the serfs do. It's also the Czech word that became the worldwide word "robot" after its was used by Karel Čapek in the play R.U.R. to describe manufactured slaves. The abbreviated title refers to "Rossum’s Universal Robots", the lead encorporation (as in creating bodies with sufficient "liveliness" to be useful slaves) corporation in the play. It's founder's name translates to "Mr. Brain", an automation optimist with a belief that productivity is the measure of Man's* worth. The type that gets big funding in silicon valley and from industry.

As a data gorilla, I see a lot of data robata. And there's finally something to do it. Generative Pre-trained Transforms (GPT) are the robots, in the manner of Čapek, for my field.

There are a lot of people convinced that with the arrival of GPT Large Language Models (LLM), that "general AI" has arrived.

These models are impressive. And powerful. And transformative.

These are not general AI.

And are not structured like human, or even mammalian, intelligence: maybe vaguely like some particular glob of neurons, but they don't grow the right way. These are not a mirror to our minds, although they can mirror our discourse. They are a piece, a sub-system.

They model language; they do language processing. Yes, some models do image output, creating a picture from a prompt. For reasoning and conversation, these are intellectually similar in their role and impact. These models are communication engines. Without a deep audit of the "communication engine" analogy, I will state that clearly the models are practical for communication processing.

And when  trained on the "entire" internet, they "remember" it all, and can sound authoritative in real time, with all the dependability of internet advice. Like, or better than, a trope-fulfilling used car salesman. Despite the connotations, this is an amazing advance, and a powerful, powerful tool for many of us persons, natural and incorporated, that work with language, languages, imagery, or data.

• To grasp this change, there is a lot to consider and discuss.
  
• - What are these GPT-LLM?
• - How to reason about their use.
• - How to use for value.
• - Jobs these open and what training those jobs need.
• - Impact on economic and social stability.
• - Reasoning about the kind of things these are in Law.

My current summary of the GPT-LLM currently buzz-worded as "AI":

Generative models bullshit,
In an immensely useful way.
As unceasing as their electric feed,
And never ask for pay.

I am personally very excited about having and mastering the use of these tools.

* Man capitalized to indicate classical usage as a respectful pronoun for all humankind. This time, it's not just the common name of a species. And yes, you're included, whatever others may call you, or that you may call yourself. Contrast to Nature (what came about without us; what God wrought), and Machine (our creations both domestic and feral).

Thoughts on Archival Quality Systems

Sometimes a bit of perspective can be had by considering a new viewpoint, and today I'm having a bit of that experience as I consider the process of repairing, refurbishing, or upgrading interactive installations for museums. The idea of a concrete archive of knowledge, trying to pass itself on to the future, is really interesting. As someone who does a lot with code and algorithms, and a fair amount with electronics and small carpentry, I can see an installation as a set of these parts, but it seems wrong - the real point is the preserving, explaining, and sharing knowledge. All of the technical and material aspects of the design need to be subordinate to this purpose.

There are two big challenges with installations: people, and time.

People are the whole point, but people are also really rough on things that are not their own. Particularly the young people who are still learning, and since they are still learning, are the usual target audience for museum installations. So there has to be a lot of both durability and maintainability in an installation piece.

Time is an essential trait for most museums: most have preserving knowledge as a significant part of their mission. Unlike an artifact, as time goes on, an installation is expected to change and stay current with the latest knowledge. Fortunately, it is not usually expected that they will do this without human intervention (Whew!), but this does mean that they need to be fully prepared for upgrades and replacing parts.

Funding is almost always fairly short term for installation development, so they should be cheap to maintain. Further, due to the vagaries of staffing in non-profits, they should not require any particularly esoteric skills to maintain. Assume that at some point in the future, some poor volunteer is going to be called on to determine if "the old can be set up or fixed for ." Part of your goal as the original designer should be to make this person successful, and thus perpetuate the knowledge that the installation represents.

In my next post, I plan to elaborate these thoughts into steps that are a bit more concrete. I hope to see from that if this point of view is particularly supportive of particular design methods, or if it has any surprises about what to emphasize.

I support the Oxford comma.

With a highly divided election approaching, I thought I'd make my political stance clear: I support the Harvard, Oxford, and serial comma.

Consider this wish: "Live healthy, learn much, rock and roll." Super-stardom, or a rocking chair and rolling in your grave? Depends on how many things are in that list, right? For the Oxford comma camp there are three things, so "rock and roll"! For the non-comma there are four... "rock" and "roll" to you I guess.

In term of social acceptability, note that the Oxford Comma is the subject of a top-40 hit song, "ommission of serial comma" is not.

Before you dismiss me as an extremist, please note that I am forgiving in the case of the omission of the comma before the abbreviation "etc."

Privacy, Identity, and Third-Party Service Providers

"We may share information with third party service providers."

Ever read that (or equivalent) in a privacy policy?

It means: "Access to your personal information by this site, including information that you have not provided directly to us, is restricted only by expense, and anything you do here may be linked to widely available databases tracking your behavior." This is also true offline, and has been true since before the 1990's. And about combining offline and online data: this was never effectively restricted. As soon as your address existed in the database of an online company that allowed "third party services", they could buy the results of your warranty cards, vehicle registrations and other public records, magazine subscriptions, credit card purchasing profiles, and more. Surprised?

Would you be more surprised if you found out that knowing how often your doctor prescribes types of medicines is also available. And that availability has been considered constitutionally protected as part of the corporate "freedom of speech" for the BUYING corporation? Imagine if we as individuals could demand information because it might change how we say something! (Yes, that's a power that corporations have claimed as part of their rights due to their "personhood".)

You may rightly wonder how all this is available. If you read the fine print on various contracts you'll see a lot of "may be shared with third party service providers" and "may be used for marketing purposes". Those two together are the glue for the corporate identity markets: the former is taken to mean "I can share all the data I have with a data aggregater to get the information about you that I don't have". And the latter: "and it can be shared with any other company that also wants to market to you or any of the characteristics that have been attributed to you."

What data links these databases together? Anything they can get. An address (shipping, billing, receive our packet, free gift, ...)? Great! A phone number? Great! A cell phone number? Even better: location, and more individual than a house phone! A credit card number? Not so great: we'll have to pay the credit card company to give us the name/address/phone of the people who bought from us, because the card data is protected by law. But rock solid reliability, since they have your billing address (and probably your primary email address, and a couple of phone numbers - after all, you have to "call from your home phone to activate", right? Even if you never gave them your home phone before that...). And then we'll have to pay again to link it to any other databases. An email? Okay... they're unreliable, but as long as you've bought from someone and given them that email, or given that email to someone you do business with (electronic billing, maybe?) we can probably link it to your "real" identity.

Yes, with school and other events, this blog has, alas, been neglected. On the upside, I am remembering it again, and have a bit of a backlog of draft posts, like the above. The topics may broaden a bit, as I explore a bit more about teaching and communicating scientific and technical skills.

Scientific Grammar

Scientific writing is sometimes hard to read because of bad grammar, even more than because of strange abbreviations and technical terminology. This is sadly expected in journal articles, even though clear writing will make it more likely that someone will read far enough through your research to use it and cite it. It is also the reason that so many whitepapers are written by non-experts. The sponsoring organization wants people to read them, not fear them.

"The Science of Scientific Writing" (Gopen and Swan, American Scientist, Nov-Dec 1990) is an article that does a great job at documenting these problems and showing how to fix them. The article stresses a simple pattern: start with the familiar: end with the new. As they put it:

"In our experience, the misplacement of old and new information turns out to be the No. 1 problem in American professional writing today."

Gopen and Swan back their thesis up with "worked examples." Taking passages from published articles, they show how to revise them for clarity.

The article ends with seven rules to summarize what they have found. I am putting the rules here to remind me of them, and to entice the reader unfamiliar with them, to visit the original article and learn what they are reminders for.

1. Follow a grammatical subject as soon as possible with its verb.
2. Place in the stress position the "new information" you want the reader to emphasize.
3. Place the person or thing whose "story" a sentence is telling at the beginning of the sentence, in the topic position.
4. Place appropriate "old information" (material already stated in the discourse) in the topic position for linkage backward and contextualization forward.
5. Articulate the action of every clause or sentence in its verb.
6. In general, provide context for your reader before asking that reader to consider anything new.
7. In general, try to ensure that the relative emphases of the substance coincide with the relative expectations for emphasis raised by the structure.

This article won't replace "The Elements of Style" by Strunk and White, but it is a useful addendum for the scientific writer.