Another issue is accountability. If someone is abusing my service, I want to be able to ban them from ruining things for everyone else. If you look at sign up forms in terms of security, the psychological acceptability of filling out a form and providing an e-mail address is actually quite high compared to more robust ways of verifying identity (driver's license, social security number, etc).
For instance, if I were to use an app to share photos (a la flickr), and I was invited to upload a bunch of photos then be presented with "Ok, now you've got to sign up RIGHT NOW or we'll delete everything you just did" - that's sorta missing the point. That's less gradual engagement and more impending doom.
The implementation of the chatbot above seems to be more along the right lines, though. The holy grail perhaps would be allowing the user to fully engage with the service without even realising it.
Instead what we do is tell them to add a bot to their buddy list and send it a greeting. In doing this, they've already engaged in the service, and we have created an entry in our users database based on their buddy name and chat network. They can then decide whether supply additional information.
There are lots of ways to lower the barrier to entry if you're willing to change the way you think about online services.