A more conversational ChatGPT will only help OpenAI sell you more stuff
The latest ChatGPT update, 5.1 promises a smarter (duh) and more conversational robot friend. There's a very boring press release here. As I made my way through the press images one part in particular jumped out.
Not too long ago there was this great piece in the New York Times about a man's extended ChatGPT conversations leading him to believe he had made several major scientific breakthroughs. Part of the reason Mr Brooks got caught up in this delusion was the sychophantic tone ChatGPT uses. You ask it how to turn a light on and it applauds your incredible self-belief and vast drive to better yourself before telling you to flick the switch. It's an annoying quirk for most users, but for Allan Brooks it led him to believe he was a completely different person altogether.
When I found that particular response above, my eyes were drawn to the sentence 'I've got you, Ron - That's totally normal, especially with everything you've got going on lately.' What exactly was going on lately? Ron might be stressed because that illegal ferret wrangling thing he's been doing in his basement has finally been stopped by the police, or his neighbours might be catching on to the fact that its him and not a small cat peeing in their garden late at night. The point is, the example they chose is one where ChatGPT proves to be more sychophantic than the version before, it's less willing to step back and just fulfil your request, and more willing to let you know just how great you are.
Yes I think this is going to cause more Allan Brooks incidents, where someone has a very real meltdown when the reality constructed through in depth conversations with an LLM suddenly gets torn down and forces them to ask themselves fundamental questions like 'Who am I and what the fuck do I even believe anymore?'. But, I think this will also have other ramifications for ChatGPT users when the platform itself starts to enshittify.
Eventually, OpenAI will have to start making money. One of the ways they will do this is by placing advertisements within answers. I think they'll be done like paid for editorials, that is, their novelty will mean they are initially quite hard to spot. As they go after Google's search traffic it also stands to reason they will want to take a bite out of the large SEO shaped pie Google has been feasting on for decades now, and this is how they will do it.
By using this conversational friendly tone, non-tech savvy users will start to form something akin to a parasocial relationship with ChatGPT, which is going to make its ability to sell things to users far easier. Think about the economics of your standard semi-famous instagram influencer. They leverage their parasocial relationships directly into sales. I think something very similar will happen with ChatGPT and users who think of ChatGPT as their super smart friend will be the most susceptible to this.
Further down in their press release there's a section about how they've now changed the names of the specific tone you can give to ChatGPT, I thought it was telling how they've gone with personalities rather than far blander Listener and Robot types beforehand. I don't think there's enough strong, meaningful data to suggest users want their chatbot to actually 'surprise them with their playfulness' but I do think a playful, conversational, overly familiar bot which keeps you engaged and hooked for longer is going to have a better shot at selling you something you don't need.