All the current dating sites (Plentyoffish, OkCupid, etc) started by first creating hundreds of thousands of fake profiles and activity to attract users.
And then slowly scaling back on those fake profiles when they had real users.
Is there a way to create a new dating site without resorting to such a tactic? Which is questionable at best because those photos they used were of real people without their permission.
Pre-signups and a staggered rollout to different geographies. If you can get people interested on your unique premise alone, and you can get exposure to a large number of people in a specific geography, you can have people do a pre-signup with a bare minimum of info (age, gender, location) to be notified when it's available in their area. Ideally, you get a picture and a couple sentences description from them as well, so their profile can go live as soon as they're granted access.
Once you reach a large enough queue of users, with a viable gender ratio, you open up access in that region and notify people their profiles are live and ready to be further customized. If you successfully already got some photos and descriptions from them, there may even be some messaging activity that starts right off the bat and helps draw them in.
There's also a model like what Jess, Meet Ken is using, where users are actively referring new users. This has been very challenging for most dating sites in the past (because of stigma around online dating, because of the social signalling issues), but there's always room for innovation.
Finally, a lot of startups build an audience through content before they offer an actual product. If you can get a large enough audience, you can convert a large number of them into your initial users, so long as your content and the product are relevant enough to each other.