A recent post on TikTok declares a commitment to AI transparency and introduces new efforts to do so. The social media company says it will expand AIU’s literacy efforts with “continued investment in new educational resources and trusted expert partnerships.”
“We know from experts that education is key to empowering people to take control of their experience with AI,” the blog reads. TikTok’s new guide aims to promote responsible use of AI tools, with a new in-app hub teaching “practical skills for finding AI-generated content when searching for AI-related terms.”
The company also promises to test an improved detection system targeting so-called AI spam and vulgar accounts that post “topics that may pose a risk to public trust or well-being.” In this, we move towards similarity detection protections that other companies have introduced to prevent deepfakes.
Finally, I will join the C2PA Steering Committee to further strengthen our efforts in labeling and content authentication.
Please specify: AI-A or AI-B?
In general, this post takes care to ensure readers understand that TikTok is working on AI. Deep learning and generative AI, “when used transparently and responsibly, can transform the way people share their creativity, discover new passions, and stay safe on our platform,” the company said.
The company says these updates are part of a “broader effort to protect and enhance positive AI experiences on TikTok.”
This statement highlights the emerging need to distinguish between types of AI. This umbrella term, which stands for artificial intelligence, has little use outside of science fiction and marketing. It is imprecise, vague, not at all descriptive, and is burdened with a semantic history. As the technologies involved become more widely applied, a line will need to be drawn between machine learning systems deployed for on-platform security and systems that “transform the way people share creativity.”
The latter generally refers to generative models and AI chatbots that prompt users to generate media content based on what they have captured as training data. It is almost always sensory in nature, i.e. audiovisual. These have so far had a poor track record.
Security AI and its less obvious cousin, LLM, used as a coding tool (or fraud engine), operate at the system level. Their purpose is fundamentally different. These foster the kind of creativity that TikTok doesn’t talk about empowerment or “new passions.”
The question arises: what kind of AI is TikTok working on? And who will be tasked with classifying so-called good AI from so-called bad AI?
Tick, TikTok: genAI deluge could disrupt platform
TikTok is currently under investigation by British regulator Ofcom over whether its machine learning-based system for estimating users’ ages qualifies as a highly effective age guarantee. Like many big tech companies, it touts AI as the future and doesn’t want it to be regulated outside of its own pitch and the many use cases it has spun off for its own model. This means that, like Meta, it operates a doublespeak script that combines security and privacy efforts with evangelization and evasion.
But TikTok’s commitment to AI transparency, particularly its promised action against AI spam accounts, raises concerns. “Every day, we see amazing ways that AI is creating new storytelling opportunities for communities.” “However, as this technology evolves, it could also be misused to create mass spam that excludes genuine creators.”
The stated goal is to remove accounts that spread false opinions about politics, current events, financial advice, medical content, and more. However, there is a core of understanding in it. The law of scale suggests that over time, platforms like TikTok are likely to become overwhelmed with randomly generated, fake AI content that no one wants to see. The AI boom says embrace change, but TikTok may not like what its platform looks like on the other side.
As it stands, it’s the last major social app to garner massive attention. It was released 10 years ago. Glyphs are on social media screens. Ultimately, all traditional social platforms may be doomed to follow Facebook’s path into a swamp of AI posts about vegetables and special cookware that only old people care about.
Article topics
AI fraud | Deepfake detection | Deepfake | Generated AI | Similarity detection | Social media | TikTok
