The best social networks are those developed in-person at conferences. They can be perpetuated via technology, but I have made very few "friends" via ResearchGate that I interact with regularly and feel that I know, beyond a Q&A context.
Academia.edu and ResearchGate are the most popular, generic ones, with LinkedIn trailing, because it is more professional-oriented. ResearchGate is more dynamic, but they have different audiences and memberships, so you might want to use both until open access Scholarly Hub gathers momentum. ORCID is not a social platform, but rather a way publishers disguise to get us check that they use our identities right for free. Using their ID is useful, especially if you have an Anglosaxon name, but don't expect them to do anything really for researchers, such as clarifying corret name usage for other systems (e.g., Arabic, Chinese, Hispanic, Indian, Portuguese, Arabic). You might also consider Google Scholar. It is very modest as to the social aspect of their offer, but many institutions do consider their impact report in their reviews.
The best social networks are those developed in-person at conferences. They can be perpetuated via technology, but I have made very few "friends" via ResearchGate that I interact with regularly and feel that I know, beyond a Q&A context.
Generic, mediated networks (like LinkedIn, Academia.edu, ResearchGate, Mendeley) are good but to be relevant to you and sustainable they need to include people with whom you can actually work on projects. Any of them are more successful when you build them the other way around: starting with people that you know and work well to others that you follow to either take inspiration or to keep up to date in your field. So attending conferences, panels, summer schools, exchanges, reaching out to colleagues to carry out collaborative research is the best way to build a strong network. Technology will help you keep in touch.