"Eclipse" music video

The release of the trail­er for Mor­gan pro­vides fur­ther insight that Arti­fi­cial Intel­li­gence (A.I.) is creep­ing slow­ly onto the cre­ative stage. After ini­tial progress was made in 1996 when IBM’s Deep Blue beat World Chess Cham­pi­on Gary Kas­parov, IBM’s Wat­son then beat human cham­pi­ons in Jeop­ardy, and more recent­ly Google’s Deep­Mind con­quered the ancient Chi­nese game of Go. Google Pho­tos gen­er­ates videos (with music) auto­mat­i­cal­ly from images on my phone. It’s becom­ing obvi­ous that deep strate­gic think­ing is at least pos­si­ble using machines.

So, can a machine cre­ate a video nar­ra­tive? Could we tell the dif­fer­ence?

The unfor­tu­nate fact is, of course, that the Mor­gan trail­er is hol­low and poor­ly-paced (even with the help of an “IBM film­mak­er”), and the musi­cians behind the AI-direct­ed “Eclipse” music video have dis­tanced them­selves from the end prod­uct.

Look­ing deep­er into each of these projects, they still required a human hand to direct/collate/guide the machine — it’s a ground-up approach to AI, so there’s no “Wat­son Video Edit­ing Soft­ware” on the mar­ket.

How­ev­er, the build­ing blocks have already been cre­at­ed — the Google search engine uses nat­ur­al-lan­guage pro­cess­ing, and Wolf­gram Alpha accepts com­mands in basic Eng­lish. We now have (pret­ty good) auto­mat­ic web sum­maries and head­line analy­sers. There’s a rea­son why Google’s Prin­ci­pal Film­mak­er Jes­si­ca Brill­hart thinks Zork’s lan­guage pro­cess­ing will heav­i­ly influ­ence the future of VR.

It seems that although we prob­a­bly have a while to go before cre­ativ­i­ty is real­is­ti­cal­ly threat­ened in any way, most peo­ple won’t care if some­thing has been cre­at­ed by com­put­er. For exam­ple, much of the in-house pro­mo­tions we cur­rent­ly see on TV chan­nels are pack­aged in a way that wouldn’t require human inter­ven­tion — so per­haps it might not be that long after all (for spe­cif­ic sit­u­a­tions).

So a machine can assem­ble a video (I even hes­i­tate to use the words ‘edit’ or ‘direct’). But not very well — at least not yet. How­ev­er, as Lin­guis­tics expert Noam Chom­sky said, per­haps we are even ask­ing the wrong ques­tion:

“Think­ing is a human fea­ture. Will AI some­day real­ly think? That’s like ask­ing if sub­marines swim. If you call it swim­ming then robots will think, yes.”

via ibm.com and bbc.com