Anyone else feels many LLMs are heavily biased towards consumerism these days?
Consumerism is the idea that encourages the continuous acquisition of goods and services by ordinary plebs. Consequently, the solutions to a problem many LLMs present are also geared towards maximum spending, they won't present a budget, non-premium or a free solution even if one exists. They might elaborate if you ask about it specifically but rarely come up on their own. The default option is always the most pricey or premium option. Did any of you notice that with specific LLMs? I tend to specify oss/free/self hosted solutions when looking for things and it usually get appropiate results. I think if commercial options are more popular and more present in training the bias is kinda expected. Eg when asking for c compilers vs mobile video editor apps it's clear where they gonna fall. For now it's my experience but ads surely will start getting in the way. OpenAI is explicitly selling ads. Google’s whole business is ads, so I’d be unsurprised for that to show up in Gemini, too. And then you have to consider how much of the training data is probably scraped content marketing. Adding "include free and open-source options" to system prompts fixes this almost entirely in my experience. Shouldn't be necessary, but it works. Yes it's very clear chatGPT's free models are concerned with general public / household problems / personal problems. It's not really suitable for business tasks.