Yes, if it's small enough and/or there is enough temperature control over the finished piece.
Glass and steel in contact can be fatal as the steel expands more than glass with heat and causes internal strains which can crack or even shatter the glass.
Low expansion glass like pyrex borosilicate glass will be the safest to use.
Low expansion ceramics, like Zerodur which has almost zero expansion and is used for making telescope mirrors instead of using pyrex glass, can be coated with aluminium or other metals like gold, chromium etc, and then hard silica coated, both done in a vaccuum sputtering process which will make an object that looks all metal.
Zerodur is normally supplied in blocks requiring machining, hence casting isn't possible since it is already fired by the manufacturer.
It might be possible though to get the un-fired product if you can find a supplier, which would make extraordinary sculptures possible as you'll see by the constructional detail of some of the objects in the links.
Here is a large ceramic mirror for a telescope, coated wth gold the same way as above, but with a much higher surface accuracy than is needed for visual smoothness of a sculpture or similar art piece. This one cost megabucks.
http://www.google.com/imgres?imgurl=http://www.schott.com/magazine/images/si104_01_zerodur.jpg&imgrefurl=http://www.schott.com/advanced_optics/english/projects/exploring_space.html&usg=___WKfc3oiuYdq8xHT4I_3h3uKp7Y=&h=160&w=270&sz=14&hl=en&start=0&zoom=1&tbnid=h5ZaU0PMb4TwjM:&tbnh=127&tbnw=215&ei=s7f9TevGGIyLswavzcXyDQ&prev=/search%3Fq%3Dceramic%2Bzerodur%2Bmirror%26hl%3Den%26sa%3DX%26biw%3D1004%26bih%3D616%26tbm%3Disch%26prmd%3Divnsb&itbs=1&iact=hc&vpx=746&vpy=288&dur=1813&hovh=128&hovw=216&tx=119&ty=94&page=1&ndsp=12&ved=1t:429,r:7,s:0&biw=1004&bih=616 . . . . .
http://www.pulleshanique.com/02_zerodur.htm . . .
Small coated Zerodur object for a science application...who would know by looking at it?
http://www.google.com/imgres?imgurl=http://www.lambdaphoto.co.uk/images/optics/asf.jpg&imgrefurl=http://www.lambdaphoto.co.uk/products/110.600&usg=__BFbKM8XZHofVOH33EDbb2zmKwfY=&h=153&w=230&sz=9&hl=en&start=52&zoom=1&tbnid=6nYGLjHjhbVZDM:&tbnh=122&tbnw=184&ei=s7f9TevGGIyLswavzcXyDQ&prev=/search%3Fq%3Dceramic%2Bzerodur%2Bmirror%26hl%3Den%26sa%3DX%26biw%3D1004%26bih%3D616%26tbm%3Disch&itbs=1&iact=rc&dur=230&page=5&ndsp=12&ved=1t:429,r:8,s:52&tx=139&ty=87 . . . .
The large glass block on your other Q (thanks for the response, what a world, making things eh?) could be made with blocks surrounding a central one cast to enclose the steel.
It will require making flat enough surfaces for the joins which takes a lot (a l_o_t) of work to avoid optical distortions.
You won't need anything like the accuracy required for an optical flat for a telescope or coelostat ( a mirror system that follows the Sun's path to provide a reflected beam into a fixed telescope or other instrument).
The 10" flat in the second link would give most folks nightmares. 4" is bad enough.
To fix the blocks together a transparent cement or resin can be used as in compound lenses, and with care a lot of work could be avoided by having the cement or resin, which is the same refractive index as glass, fill the gaps where flatness isn't great and making them just disappear.
The joys of matching refractive indices....
Be very careful about small bubbles though, which takes care to avoid when making microscope slides the same way with a thin glass cover slip half an inch across to protect the specimen.
You won't need this but you get the idea. Flattening glass is very hard to do well and takes a long time
http://www.astrosurf.com/jwisn/diagonal.htm . . . .
http://www.loptics.com/ATM/mirror_making/four_flats/four_flats.html . . . .
An alternative is a glass tank filled with water containing lead salts to increase the refractive index and give the appearance of looking through solid glass.
It has the advantage of thick flat glass panels being availble up to a few inches thick including armoured glasses ( at a high cost) and with optically good surfaces.
Silicon and other bonding agents are easy to find and to use for those, but the joins will need sorting to make the object look like a block of solid glass instead of joined panels.
Whichever way it's done it's an expensive project.
Have fun....