In the beginning...
it may be worthwhile to offer several, shorter, topics for each session. Doing so has several advantages (and a few risks)
- The presentations remain relatively general, in a 25 minutes presentation, one doesn't get in enough detail to bore folks that are not actively practicing the underlying langage/technology etc.
- It is easier to find volunteers for short presentations
- Folks will more readily show up if there's at least one topic that seems of interest to them and the chances of this happening increase with say three presentations rather than one.
- After each session you can collect feedback, and see what works, and what doesn't (and possible why) and hence fine tune things are the group's interest clusters become more evident.
Except for basic presentations aimed at presenting a particular language, i.e. giving a flavor of what the syntax looks like, what the general strengths and purposes of the language etc, I'd stay away from language-specific topics, as a way to be more inclusive. Instead I'd try topics like Artificial Life, various sub-topics of linguistics / NLP, game design, Web technologies, history of computer hardware...
Another cool feature to have as you try and build up a group, is "jam sessions", maybe the last 30 minutes of the meeting, where a few folks each time have the floor, for a maximum time of (say) 9 minutes to discuss/demo any topic of their choice (with possible censorship needed as per University guidelines, but otherwise very open and wierd is cool: open people's mind, and otherwise, not bother them for more than 9 minutes).
The main risk with this generally "light"/generic regimen is that some of the folks may want to go more in-depth, a get tired of this topic butterflying. Maybe after a few sessions you can have better feel for clusters of interest (indeed a session could be about the CS and mathemtics behind social networking, recommendation systems etc..) and you can start "putting more meat on the bone", in specific areas, be they language-driven (a Python interest group) or topic oriented (Natural Language Processing or Game development or ....). Also you can/should be up-front about this trial-and-error approach, so that folks understand the content's depth may increase over time.
Finally when you have a better feel for what the group likes and dislikes, you should seek to have outside participation, from local industry players, from faculty outside of CS/IT, possibly there'd be nothing wrong with bringing in, say, a anthropoloist/linguist to discuss Lakoff's prototype concept and its relation to categorization/filtering/abstraction, for example. Aside from networking opportunities (and I'm not talking LAN parties lol), this outside outlook could provide additional interest from the group.