The Peculiarity of Conversational Ads
ChatGPT introducing ads shouldn't have its focus stop at "another product needs to earn its keep." The real watershed is that ads have entered the conversation: the worries, plans, and hesitations you've talked over with the assistant can all become the basis for targeting, and the depth of this understanding is beyond any search box. Worse still is the form—a banner ad is recognizable at a glance, but when an ad takes the shape of "the assistant mentioning something in passing," recognizing it becomes hard. OpenAI says there'll be labeling during the test phase, but how prominent and how broad the labeling is has no detailed rules yet.
The Cost of Free and the Regulatory Vacuum
This step is nearly inevitable—top models' inference costs are high, and the huge subsidy for free users needs a monetization outlet, with ads the most ready-made answer. But do the math clearly: the inference free users enjoy is built with real money, and this subsidy has to find an outlet eventually, with ads just the most ready-made one. What's worth being wary of is regulatory lag—for "advertising disclosure in conversational AI," existing rules are basically blank, and what counts as an ad, whether the model can distort advice for profit, and whether private conversations can be used for targeting all have no clear answers. For users, the pragmatic attitude is a bit more wariness toward the model's "casual" product recommendations, and don't fully trust an assistant that may be pushing products with important decisions.
via: Hacker News