Hacker News Reader

Comment by bunderbunder | Hacker News Reader

The original statement pointed out that it’s just matrix math as a way of indicating that the technology itself is ethically neutral, and then went on to say that what we really need to pay attention to is how it’s being used. The “connect to” choice of words is clumsy and I don’t love it, but it does at least get at the point that AI doesn’t inherently become responsible or irresponsible until you hook it up and actually start doing things with it.

Which, if it’s not exactly the same point, I’m pretty sure it at least cuts quite close to the point of your rhetorical question, right? That the problem isn’t with the ones and zeros, it’s with what certain people are doing with the ones and zeros?

View on HN

> The original statement pointed out that it’s just matrix math as a way of indicating that the technology itself is ethically neutral, and then went on to say that what we really need to pay attention to is how it’s being used.

I think the idea that this technology is ethically neutral needs a discussion.

Would you care to elaborate?

A matrix is a mathematical abstraction, a technique for representing and manipulating relationships between quantities. You can call matrix math a technology, but then the term loses most of its meaning. I'd reserve the term "technology" for designed artifacts and systems: cars, guns, social media platforms, LLMs. Design requires intentionality, and these things are built by particular people, for particular purposes, under particular constraints, and those purposes get built into them. A gun is designed to kill and that purpose doesn't disappear depending on who holds it. Ad-funded social media platforms are optimized for engagement, and features like infinite scroll exploit known cognitive tendencies to achieve that. With LLMs, developers decide what data the model learns from, what it will and won't do, and whose values shape it's behavior. Of course, how we use technology matters too. But building something is a deliberate process, in which the builder makes a series of decisions about what to build and how. Many of those decisions involve trade-offs that affects people. That's why I find it unconvincing that technology is ethically neutral.

It sounds like we agree on everything but some semantic distinctions.

Because I would argue that the thing that might make an LLM harmful or not is not the underlying technology. It’s the ways its training data were acquired, the purposes for which the model was trained, the applications for which it’s being deployed, etc.

The fixating on “it’s just matrix math” doesn’t feel useful to me. I’m an NLP person and agree that it’s a huge oversimplification. But the original commenter’s purpose was not to educate people about how LLMs work in detail, and so being pedantic about it only serves to steer the conversation in an unedifying direction.

I think the problem is the ones and zeros or the linear algebra if you want to use that metaphor. I believe "progress" in AI needs to be stopped. AI researchers need to find something else to do with their lives.

My first comment was an attempt to show the "it's just linear algebra" argument in an absurd light.

Reply on news.ycombinator.com