As the AI development race heats up, we’re getting more indicators of potential regulatory approaches to AI development, which could end up hindering certain AI projects, while also ensuring more transparency for consumers.
Which, given the risks of AI-generated material, is a good thing, but at the same time, I’m not sure that we’re going to get the due diligence that AI really requires to ensure that we implement such tools in the most protective, and ultimately beneficial way.
Data controls are the first potential limitation, with every company that’s developing AI projects facing various legal challenges based on their use of copyright-protected material to build their foundational models.
Last week, a group of French publishing houses launched legal action against Meta for copyright infringement, joining a collective of U.S. authors in exercising their ownership rights against the tech giant.
And if either of these cases results in a significant payout, you can bet that every other publishing company in the world will be launching similar actions, which could result in huge fines for Zuck and Co. based on their process of building the initial models of its Llama LLM.
And it’s not just Meta: OpenAI, Google, Microsoft, and every other AI developer is facing legal challenges over the use of copyright-protected material, amid broad-ranging concerns about the theft of text content to feed into these models.
That could lead to new legal precedent around the use of data, which could ultimately leave social platforms as the leaders in LLM development, as they’ll be the only ones who have enough proprietary data to power such models. But their capacity to onsell such will also be limited by their user agreements, and data clauses built in after the Cambridge Analytica scandal (as well as EU regulation). At the same time, Meta reportedly accessed pirated books and info to build its LLM because its existing dataset, based on Facebook and IG user posts, wasn’t adequate for such development.
That could end up being a major hindrance in AI development in the U.S. in particular, because China’s cybersecurity rules already allow the Chinese government to access and utilize data from Chinese organizations if and how they choose.
Which is why U.S. companies are arguing for loosened restrictions around data use, with OpenAI directly calling for the government to allow the use of copyright-protected data in AI training.
This is also why so many tech leaders have been looking to cozy up to the Trump Administration, as part of a broader effort to win favor on this and related tech deals. Because if U.S. companies face restrictions, Chinese providers are going to win out in the broader AI race.
Yet, at the same time, intellectual copyright is an essential consideration, and allowing your work to be used to train systems designed to make your art and/or vocation obsolete seems like a negative path. Also, money. When there’s money to be made, you can bet that corporations will tap into such (see: lawyers jumping onto YouTube copyright claims), so this is seemingly set to be a reckoning of sorts that will define the future of the AI race.
At the same time, more regions are now implementing laws on AI disclosure, with China last week joining the EU and U.S. in implementing regulations relating to the “labeling of synthetic content”.
Most social platforms are already ahead on this front, with Facebook, Instagram, Threads, and TikTok all implementing rules around AI disclosure, which Pinterest has also recently added. LinkedIn also has AI detection and labels in effect (but no rules on voluntary tagging), while Snapchat also labels AI images created in its own tools, but has no rules for third-party content.
(Note: X was developing AI disclosure rules back in 2020, but has not officially implemented such).
This is an important development too, though as with most of the AI shifts, we’re seeing so much of this happen in retrospect, and in piecemeal ways, which leaves the obligation on such to specific platforms, as opposed to implementing more universal rules and procedures.
Which, again, is better for innovation, in the old Facebook “Move Fast and Break Things” sense. And given the influx of tech leaders at the White House, this is increasingly likely to be the approach moving forward.
But I still feel like pushing innovation runs the risk of more harm, and as people become increasingly reliant on AI tools to do their thinking for them, while AI visuals become more entrenched in the modern interactive process, we’re overlooking the dangers of mass AI adoption and usage, in favor of corporate success.
Should we be more concerned about AI harms?
I mean, for the most part, regurgitating information from the web is largely, seemingly just an alteration of our regular process. But there are risks. Kids are already outsourcing critical thinking to AI bots, people are developing relationships with AI-generated characters (which are going to become more common in social apps), while millions are being duped by AI-generated images of starving kids, lonely old people, innovative kids from remote villages, and more.
Sure, we didn’t see the expected influx of politically-motivated AI-generated content in the most recent U.S. election, but that doesn’t mean that AI-generated content isn’t having a profound impact in other ways, and swaying people’s opinions, and even their interactive process. There are dangers here, and harms being embedded already, yet we’re overlooking them because leaders don’t want other nations to develop better models faster.
The same happened with social media, allowing billions of people to access tools that have since been linked to various forms of harm. And we’re now trying to scale things back, with various regions looking to ban teens from social media to protect them from such. But we’re now two decades in, and only in the last 10 years have there been any real efforts to address the dangers of social media interaction.
Have we learned nothing from this?
Evidently not, because again, moving fast and breaking things, no matter what those things might be, is the capitalist approach, which is being pushed by corporations that stand to benefit most from mass take-up.
That’s not to say AI is bad, that’s not to say that we shouldn’t be looking to utilize generative AI tools to streamline various processes. What I am saying, however, is that the currently proposed AI Action Plan from the White House, and other initiatives like it, should be factoring in such risks as significant factors in AI development.
They won’t. We all know this, and in ten years time we’ll be looking at how to curb the harms caused by generative AI tools, and how we restrict their usage.
But the major players will win out, which is also why I expect that, eventually, all of these copyright claims will also fade away, in favor of rapid innovation.
Because the AI hype is real, and the AI industry is set to become a $1.3 trillion dollar market.
Critical thinking, interactive capacity, mental health, all of this is set to impacted, at scale, as a result.
