There’s no viable case for publicly accessible artificial intelligence tools, while the case against consumer AI is significant, and is widely being overlooked by profit-driven tech developers.
Social media platforms continue to roll out AI tools that generate images and posts, and create stickers, videos and even companions. But none of these use cases provides a net positive experience, and in fact, these scenarios all erode the value of social media as a connection and communication tool.
One example of this is the rise of AI slop, which has often been perpetuated by the platforms themselves as they add various tools to generate content. But now, the influx of AI-generated junk is causing platforms to reverse course, and even eradicate AI-generated material in the hopes of reducing slop overwhelm.
Last year, Pinterest added an option that lets users reduce the amount of AI-generated content that they see in their feeds, after users complained about the influx of AI slop overwhelming their Pin displays. More recently, LinkedIn allowed users to report AI slop. The platform said that it’s working to limit exposure for AI-generated content, despite the fact that LinkedIn also once gave users a button within the post composer that allowed them to use AI to generate a post on their behalf.
Both Pinterest’s and LinkedIn’s AI-limiting options are seeing heavy use, and other platforms may soon add similar restrictions, as AI-generated junk continues to flood social feeds.
Yet the counterargument here is that AI tools enable expanded content creation and experimentation, and give people more tools to bring their creative visions to life.
Which is true. Through AI tools, anyone can generate whatever they can imagine, and some of those examples do end up leading to interesting results. But the negative impacts of the AI slop wave significantly outweigh the benefits of simplifying content production, and the oversupply of slop is not worth the potential benefits of AI use.
The current wave of generative AI tools is primarily beneficial for specific, niche use-cases, when overseen by people who understand the outputs and how to refine and improve on those results.
AI tools, for example, can generate a poster for an event, but it’s going to look low-quality or generic unless the creator knows the ins and outs of graphic design and can refine and improve the output.
The same holds true for business use cases. AI tools can highlight elements of concern on a patient’s medical chart, but a medical professional is required to assess those outputs and determine their relative significance. AI can improve emails, but users still need to ensure that the information is correct. AI processes can find the best deals online, but users still need to ensure that the exact requirements and fine print align with the final use case.
The current wave of generative AI tools can drive efficiency, and can help to produce better results. But they should only be applied in specific use cases, with effective oversight, which is why providing them to everyone is a flawed approach.
Even the providers themselves know this.
Every other day, there are reports of AI agents infiltrating systems or making false claims, potentially sparking major concerns.
Last week, the New York Times reported that AI agents filled out visa applications for non-existent people. Reuters recently reported that AI agents had infiltrated Australia’s government systems. Meanwhile, ABC News in Philadelphia reported that an AI model submitted a false tip about an unsolved murder.
Anthropic effectively acknowledged the potential for harm on this front last week, when it announced that it would be cutting off live web access for its internal evaluations after some of its projects exploited U.S. government agencies.
Even worse, Axios reported last week that top executives at the major AI developers are currently gaming out scenarios for a public and political revolt after a “catastrophic AI event.”
Developers know that putting these tools into the hands of people who don’t understand the outputs is potentially dangerous. Yet, at the same time, Mark Zuckerberg is advising people to let their Meta Muse agent undertake all sorts of tasks on their behalf, while over the weekend, Elon Musk advised that people should upload their credit card details to Grok and let it make purchases for them.
Which leads back to the foundational point that generative AI should not be widely accessible to the public, outside of specific niche use cases.
The push to get more people to use AI agents in order to make money from the massive investment into tech development is dangerous, and is going to lead to many instances of unwitting data exposure, hacking, exploits and unintended consequences.
And in the end, all society will have to show for it will be video clips of people reimagining themselves as superheroes.
