You don't even need a nation state. 1 person could sufficiently weaponize this against multiple individuals by using it to mimic their writing style by training it on their social media/blog/articles they've written and then using it to start various accounts online as the target and slowly building up followers with content and at some point switching to saying highly controversial, or downright offensive, things to immediately cause people to change their opinion of the individual. Racist/sexist/etc content, allegations against someone else, fake content about a company/product/service that could cause varying levels of financial damages to the company etc.
It also is a new tool for blackhat SEO, instead of paying people to write crappy content, or running relevant articles through any number of commercially available software packages that will rewrite the article enough to fool search engines as being unique for weeks or even years, you can simply train the software on stacks of books of material about a given subject and have it churn out semi-intelligible-to-humans articles about the subject that could allow you to build a rather large amount of content to manipulate search engine rankings or even to just create contextual/affiliate ad farms.
If you have the software churn out smaller sets of content, then quickly edit it yourself, a single individual could churn out 10-20,000 words of content a day with banners/images/contextual ads inserted. A small blackhat SEO team in a country like Bangladesh or India could offer prices far cheaper than competition using more traditional methods, have articles that are more 'natural' in the customer's language, and be of a likely higher quality in a much higher volume.
This is the problem with entities like Open AI, they're all "AI is great, AI is good, AI is our future savior, yay AI" but are any of them going "well, here's the 16 ways I can think off, off the top of my head, how to wildly exploit this technology for personal/corporate/government gain"? AI doesn't have to be SkyNet or robotic killing drones to be exploited, an individual can benefit considerably (and cause considerable hardship for an entity) with stuff like this. Who at places like Open AI are asking these questions? Where are the employees/advisors/consultants that look at each project and offer real time feedback on ways to abuse the project in its current and near-future states?
It also is a new tool for blackhat SEO, instead of paying people to write crappy content, or running relevant articles through any number of commercially available software packages that will rewrite the article enough to fool search engines as being unique for weeks or even years, you can simply train the software on stacks of books of material about a given subject and have it churn out semi-intelligible-to-humans articles about the subject that could allow you to build a rather large amount of content to manipulate search engine rankings or even to just create contextual/affiliate ad farms.
If you have the software churn out smaller sets of content, then quickly edit it yourself, a single individual could churn out 10-20,000 words of content a day with banners/images/contextual ads inserted. A small blackhat SEO team in a country like Bangladesh or India could offer prices far cheaper than competition using more traditional methods, have articles that are more 'natural' in the customer's language, and be of a likely higher quality in a much higher volume.
This is the problem with entities like Open AI, they're all "AI is great, AI is good, AI is our future savior, yay AI" but are any of them going "well, here's the 16 ways I can think off, off the top of my head, how to wildly exploit this technology for personal/corporate/government gain"? AI doesn't have to be SkyNet or robotic killing drones to be exploited, an individual can benefit considerably (and cause considerable hardship for an entity) with stuff like this. Who at places like Open AI are asking these questions? Where are the employees/advisors/consultants that look at each project and offer real time feedback on ways to abuse the project in its current and near-future states?