Javascript is by far the most useful of dynamic languages for real-world practical work - not only is it irreplaceable for "client-side" work on the user's browser, but Node.js is rapidly making it very interesting for server-side work, too. Sure, it has many issues, but a book such as Crockford's Javascript: the good parts will help you avoid many of them.
JS's syntax of course is quite different from that of dynamic languages such as Python or Ruby, which try to avoid braces and semicolons (which you'd better not avoid in JS: it tries to guess on your behalf but too often it guesses wrong!-). There is really no "syntax used by most of the dynamic languages" given these huge syntax differences (which grow if you throw into the mix Scheme, Erlang, Perl, PHP, Tcl, ...), so that part of your specs is moot.
Second most useful today is probably Python -- as Allison Randall (program chair of OSCON and a well-known Perl guru) put it, Python has surprisingly become something of a "default language" in many fields. For example, the SEC is considering a regulation to mandate publication of algorithms used in stock trading, and their initially proposed language for such a publication is "of course" Python. As this post explains,
Why Python? The SEC actually asks for
comments on whether they should
mandate Perl, Java or something else
instead. I use Perl quite extensively,
but the idea that Perl is a suitable
language for implementing a
transparency requirement is laughable.
Perl is a model of powerful but
unreadable and cryptic code. As for
Java and C-Sharp, there is little
point in having open source code if
the interpreter is not also open
source. I do not use Python myself,
but it appears to be a good choice for
the task at hand.
This is what Allison meant by "default language", I think: not necessarily the one you'll choose to implement a given task (e.g. the above post's author would prefer using Perl), but a language everybody's supposed to be able to read in order to understand an algorithm that is published or otherwise presented -- as Bruce Eckel (deservedly-best-selling author of books on C++ and Java) puts it here,
Python is executable pseudocode.
You can look at the "executable" part as a bonus (it does guarantee lack of ambiguity, which non-executable pseudocode might lack;-) even though large systems such as reddit and youtube have been implemented in it.
At the other extreme, if you're not necessarily looking for immediately useful knowledge, but for mind-broadening, Scheme or Erlang might suit you best (but the syntax in each case is quite different from most other languages, be warned;-).
However, in that case, I'd suggest Mozart, to go with the masterpiece that is Van Roy's and Haridi's Concepts, Techniques, and Models of Computer Programming (that book is plenty motivation to learn Mozart, just like SICP is to learn Scheme -- indeed, I've described CTMCP as "SICP for the 21st century"!-).