Let's figure out something that only streaming games could do
Presently we are all watching Stadia, which is the latest attempt at streaming computer games to people and in this case it is Google's turn, be a complete disaster. This is to the surprise of nobody that even vaguely knows how computers and games work, seemingly save perhaps those people that funded the thing. Maybe it is a long play to be the one leading the dance if it all manages to actually kick off but right now... yeah.
However a comment on another thread got me curious
What it needs is a killer app : that one exclusive game that otherwise couldn't be possible. A game that fully tells the world 'screw your scepticism because I Am Here To Stay'.
What might this be? What could be done that would be extremely difficult, if not impossible, for conventional game hardware in consumer friendly packages to pull off that gamers either want, or don't necessarily right now realise they want?
The immediate thought for most is they can have the absolute screamingly fastest setup available and push the kind of pixels in real time that consoles and most PCs can only dream of, but make it available for the masses. We have seen stuff like this before, indeed anybody that ever heard the phrase "arcade perfect home port" experienced a version of that -- arcade machines by virtue of their continuous income could push the hardware boat out massively in a way that home consoles (PCs of the time were not quite games machines) could never compete with, or if they could then it was only after a few years and then it starts all over again as the state of the art has moved on further.
However that is boring so I want specifics if we are doing that, limited to what the real world or near future likely provides -- a 8 CPU and thus 64 core and 16 card SLI setup might be possible in theory but it still limits what can be done practically in the modern world.
I saw in another discussion about upcoming flight sims possibly storing terabytes remotely such that full recreations of real world locations to continent level is an option. Doing that locally is tricky for the time being if 90 gigs is pushing the envelope for a single game install... however storage costs are dropping all the time so might that catch up before long?
Similarly would that be all that interesting? As it stands procedural generation can give me an awful lot if I so desire, and do it in an absolutely tiny package (the original Elite from 1984 could give me a massive universe to explore, though they were told to limit it more than they could theoretically have done, and more directed stuff can do better still).
Some devs regard procedural generation, and the AI that follows from it, as a dirty word or a false promise* (I would probably consider them incompetent or uninspired but we can skip that one for now), there are however still practical limitations to the concept and a bit of human touch still appreciated for the time being. I don't know what the level size limit for the human mind actually is (number of friends has things like Dunbar's number, it is also generally assumed most humans don't think much beyond the horizon but we can probably recall more games maps than that, and characters from works too). This also says nothing about what benefits from more complexity there might be; "I don't have to visit it, but the option being there is enough". Similarly we can achieve that on a normal system now without needing some kind of remote processing or storage.
Maybe instead of graphics we could have AI worth noting. I don't have my own supercomputer but I can talk to one to play chess that will beat even the best humans going if I so desired. Chess is not really the end goal but such things could happily be expanded. Related at this point would be discussions of game complexity, recent efforts with machine learning to produce computers that can compete at Go and things like the old XBLA Magic the Gathering AI being reasonable jumping off points.
*Such things will eventually do far more than today. As it stands AI procedurally generated music, voice overs, news stories, behaviours, levels and more are all things doing very well. Now while story scripts still leave something to be desired it is generally a sooner rather than later anticipation here (most news stories being written by AI are usually simple here is a change in the market, here are some other changes, here are some things that happened leading up to this, which is all some news sites really need for a lot of stories) then song lyrics will probably be the first.
Instead of AI maybe we see some human actors be more viable than they are right now? Though again is that not already something of an option? Surely the likes of World of Warcraft or Eve Online could host a human controlled something but have them paid to run a shop or something slightly tedious.
Multiplayer is a thing here. As it stands for games like Civilisation we have seen some fairly creative ways to essentially do a play by email and not allow people the option to cheese the system taking the best turn they can (start of turn, send scout in every direction loading each time, pick best course of action as a result of info gleaned and send that save back to the server). Being remote then you could consider it essentially immune to this outside of someone hacking in which already is a rare problem, not to mention limiting options for sniffing data they are theoretically not supposed to have.
Somewhat amusingly then for lag (more on that later) if everybody is effectively in the same bit of system memory then it somewhat turns the ping negation and other aspects of "the internet sucks" on their head, obviously still got some problems with lag to work around (maybe some kind of AI takeover if ping drops too much) but it is a different take on the matter and solves several problems all at once. Might it be how we reasonably get 1000 vs 1000 player servers? One idea that has consistently floated around is the option for 1000 people to all join at a given time and recreate a large battle, even if one player ends up being little more than a gunner responsible for a small section of a ship in that particular instance.
It might also allow for some measure of control enough that gambling and other real money going in and out to users becomes a proper option. People have been saying if people can earn a living from games then that will change everything for many years now (far from the first but a good source might be those things for 2012's Diablo 3 with its "real money auction house") but it still remains true that if many people can then that will change things.
We have seen data from fighting game tournaments be fed back into games to improve AI there. Mega CPU plus it all being local to it could make for some serious tweaks here, and possibly some kind of push back/balancing/openings for an interesting story twist if it thinks things are going wrong (we already see stuff like the left4dead director). Theoretically such things could already be done (input + time of input is not lots of data) but I am sure the folks that spend time contemplating metrics are giddy with anticipation of what such services could allow them to do. If nothing else stats + human "invisible" moderation + instantaneous patching means exploits might not propagate far; if player times/damage/funds have achieved these results outside of expected variables then oh no look your glitch gun broke.
Combining with the memory above then maybe we see some more persistent world stuff rather than endless replayed instances. Though frankly that could be done now in some ways (see minecraft)... that said combined with the large level stuff from earlier then maybe not. I would not mind seeing a true survival game like this, especially if pinging out saw my character go AI and attempt to return home/finish the task within my capabilities as it were.
So starting with the assumption that streaming allows those doing streaming games a complete control of essentially a supercomputer, one somewhat immune to hackers, one with only the lag to remote players (which could be made low if certain expensive things are done) being the main issue then what could be achieved that presently is difficult, if not impossible, that would make for a far more compelling experience?
As part of this theorising I can allow some kind of lag negation; if they did it local to the town/city, or even better, then it would be more than acceptable, but that is actual investment and seemingly more than they care to do right here. We can also skip the "I don't own my games", "muh bandwidth caps" and "I want mods" type discussions as they are issues for some but not so very great when all is said and done (people already give their money to Steam for scarcely any better, such caps are a non issue in a lot of the world and *points at consoles*).