“Open” “AI” Isn't | Hacker News Reader

“Open” “AI” Isn't

pluralistic.net

> But "open source" itself is an offshoot of a more foundational movement, the Free Software movement, whose goal is to promote freedom, and whose method is openness.

Yeah, no. Free Software is a subset of open source software, and "Open Source" (as written) is yet another subset of open source software. No wonder people have gotten all this mixed up...

Ultimately, people are right to be mad at this, but they're choosing the wrong stones to lob. It's not "FreeAI" or "FOSSAI", it's "OpenAI". Open can mean anything, and doesn't even need to correlate with the qualities of their products.

> The paper's authors start by suggesting that the "open" in "open AI" is meant to imply that an "open AI" can be scratch-built by competitors (or even hobbyists), but that this isn't true. Not only is the material that "open AI" companies publish insufficient for reproducing their products, even if those gaps were plugged, the resource burden required to do so is so intense that only the largest companies could do so.

This is pretty wrong too! Even "Free Software" like GPL-licensed programs don't demand a non-existent resource burden for the end-user. If you cannot run the GNU Compiler, you cannot demand the upstream send you a custom binary. That's not what Open Source means, or open source, or Free Software.

> like Meta's PyTorch and Google's TensorFlow. These tools are indeed "open source," licensed under real OSS terms. But they are designed and maintained by the companies that sponsor them, and optimize for the proprietary back-ends each company offers in its own cloud. When programmers train themselves to develop in these environments, they are gaining expertise in adding value to a monopolist's ecosystem, locking themselves in with their own expertise.

Sorry, you had 50+ years to cry foul over proprietary API interfaces and now we are here. Even as a FOSS super-advocate I cannot feel empathy for this author.

> No one should be surprised to learn that the products of the "open" wing of an industry whose products are neither "artificial," nor "intelligent," are also not "open."

I understand hating on OpenAI for not being open, I agree. But to argue their products aren't AI-related is tenuous at best. Yes, they are trained by humans. But does that really exclude these models from living under the umbrella of AI? I just think the author should have backed up this portion of their argument more.

I think this touches on the fundamental issue with most AI technology in its current stage of development.

The word “intelligent” means something very different than what most people understand the word to mean.

Yes, it holds a specific meaning relative to AI products, but I think the point is generally that the industry is using these words flippantly, and in ways that influence public perception to these company’s benefit.

Interesting point. I agree the word "intelligent" can be a sticking point in this area but just looking at the definition that comes up when I long press the word on my Mac:

intelligent | inˈteləj(ə)nt | adjective • (of a device, machine, or building) able to vary its state or action in response to varying situations, varying requirements, and past experience.

I would describe ChatGPT as intelligent based on those terms. Is it AGI? Absolutely not. But I would call it intelligent because I can give it unique input and it will give unique and helpful output relevant to the context I provided it.

I tend to agree that these are "intelligent" tools, and we've been calling various categories of tech intelligent since long before LLMs entered the public consciousness. But when LLMs entered the public consciousness, it was perceived as (and implicitly sold as) a new kind of intelligence. Something fundamentally different than what has come before. And to be fair, there's some truth to that, and the experience of interacting with these tools validates that view. They feel intelligent, and it's easy to anthropomorphize. Blake Lemoine and Kevin Roose come to mind.

As a result, I think the broad perception is that this new kind of intelligence is AGI-adjacent, even if it's clearly not yet AGI. But this is a distinction that is likely lost on just about everyone who doesn't work in or deeply understand technology.

I think these companies are pretty happy to leave that misconception as-is. They'd rather the public think these things are AGI-like. The reason I'm splitting hairs here is because this all plays into whatever happens next in terms of public policy/regulations, and I think the current level of understanding outside of our bubble is rather poor.

And to bring this full circle, that's why I'm defending the use of scare quotes in this piece. I think they highlight something that we haven't figured out how to talk about accurately yet.

If you have to say it, it is less likely to be true.

Fancy ketchup.

People's Republic of <whatever>.

OpenAI.

Premium sound system

Your call is very important to us We're taking these allegations very seriously We believe in transparency

It’s fascinating how one man can so finely hone his ability to be insufferable. Even if he’s right, his style and overconfidence are incredibly off-putting to me.

There was a time when "open" didn't mean "open source" but rather, roughly "available under reasonable and non-discriminatory (RAND) license terms". The Open Software Foundation of the late 80s and early 90s took their name from this meaning.

Reply on news.ycombinator.com