The War for the ‘Soul’ of AI

Wonder why the ‘Tech Bros.’ are slobbering all over Trump?

We are at war.

It’s not the normal kind of shooting war, with tanks, and planes, and ships. It’s not the “cyber war” that we’ve seen deployed, here, and there to prove that we can mess with each other’s technology.

This is the war for the development, deployment, and ‘soul,’ of artificial intelligence (AI).

For most folks, this subject can feel a bit complex. The whole point of AI is that it’s supposed to be simple, right?

Therein lies the quietest of dangers.

Okay, What EXACTLY is an AI?

An AI is not human. It is meant to emulate the way that humans interact with one another, though. It is built with that kind of super broad ability to understand us, no matter how we communicate, to improve our very imperfect ability to communicate what we want to know.

If you’ve never directly used an AI, it isn’t like any interaction that you’ve had with a computer before.

Usually, we put something simple into a search box. Whether it’s a web browser, or a spreadsheet, the computer tries to localize that information, and provide answers. Users have been programmed, to structure our search queries so that we get an answer, close enough to what we want, that we don’t have to scroll through pages of “wrong” answers.

Imagine having a super-smart assistant, who knows pretty much everything, or at least, can find something very quickly for you.

That still sounds like a search engine? True. Try to argue with Siri, or Google? They just provide you with other, usually wrong, or inappropriate, information.

You can argue, though, with an AI. Point out weaknesses, or inconsistencies in what it is telling you. Ask it about its methods to find information, and set your own parameters, if you’re subscribed to the paid versions.

At one point, since ChatGPT 4.0 now shows you sourcing, I asked the AI to use first, or original sources, not secondary ones that were referencing the original source. It did. They were much better. I then asked it to only show me first sources.

---

As you notice, it’s not a “programming language.” If you can speak a human language, you can ‘talk’ to an AI.

Why AI is Booth Great & Dangerous

AI is a human design. You can have relatively benign designers, who program it to do good, routine, and even pretty inspirational things, at incredible speeds.

There are, in the AI world, some very dark individuals, owners, and governments, though. Their developers and decision-makers, can:

• embed specific biases,

• enforce a political dogma,

• steer people toward, or away from information in a user request;

• appease a user’s bias, or point-of-view, providing incomplete information.

This is done through:

• Data Selection and Curation:

  • Data that builds AI models, curated to reflect a particular perspective, by selectively including, or excluding. information aligning with that desired worldview.

  • The AI then may continue to reinforce biases through algorithm design, and weighting that prioritizes content that fits a preferred narrative, subtly shaping users' perceptions.

  • Iterative Training - Reinforcement of learning techniques that further adjust the model’s output to favor desired viewpoints.

• Content Framing and Language Use

  • Word choice, tone, and/or structure, can influence user interpretations of information.

  • Seemingly neutral guidance, with subtle bias, can get users to adopt particular values, or conclusions, over time.

• Adaptive Feedback Loops and Omissions

  • AI systems can adapt responses, based on user interactions, to reinforce preferred ideas by amplifying preferred feedback from like-minded users, while suppressing dissenting viewpoints, creating an echo chamber effect, without overt coercion.

  • AI can subtly omit opposing perspectives, or competing ideas. Users believing they have received comprehensive information, will not challenge it. Only subject experts would know that they have only been exposed to a curated, omissive subset.

Ownership Matters

The big players are warring for the “soul” of AI. Their approaches, to both its growth, and the “guardrails” on how it is developed, vary widely.