I find the concept of an AIM client with artificial intelligence to be vaguely disturbing. Suppose it started having my conversations for me? That could get messy.
I woke up (actually woke up, as opposed to "woke up" in a dream) to this in my inbox this morning. This probably isn't much to worry about, at least for now ... but part of computing is understanding the risks. [Read it before you go any further.]
Now, hashing is not the same as cryptography. So if you use, say, SSL to send your credit card over to Amazon, or PGP to encrypt email to your SO ... you're probably still reasonably safe. (Even PGP signatures are safe, because those use SHA1, which is a different--and stronger--algorithm which hasn't yet been broken.)
However, a lot of password systems (including most standard Linux boxen) use MD5. As far as I can tell from looking at their current CVS sources, LiveJournal uses MD5 for storing passwords internally.
But, for the most part, people don't need to worry. Those MD5 hashes are most often stored in a reasonably-protected database behind a firewall. So it would be very difficult for an attacker to even get the hash in the first place. And really, if someone can get enough access to LJ's database to get your password hash, they can probably read all your private entries anyway so it doesn't really matter.