There are a couple interesting points to be made here, I think.
The first is that, of course, current consoles are old. It's not 2006 anymore. We haven't called the Wii/PS3/360 "next gen" for many years. We're in just a few months from the 6 year anniversary of the 360, and 5 year anniversaries for the PS3 and Wii, at time when consoles are traditionally replaced with something later and greater. The Wii will have a successor within a year or so, but that isn't supposed to be much better than the 360 or PS3. Those two systems are very much overdue for refresh announcements. So should we be surprised that devices which sell for $600+ without subsidies will be able to match hardware which will be 8 or more years old? No. It's just the timing, and the timing is poor for consoles. When a new generation of consoles comes out, it will reset the relative performance gap between console and mobile. Which brings me to the next interesting point.
Consoles currently use the ARM architecture. This should put them on the same playing field as mobile devices when it comes to power efficiency and cost-effectiveness. If you were to graph performance versus time for both consoles and mobile devices, both should be going up at the same rate, just with consoles making huge jumps every ~5 years and smartphones making very minor jumps several times a year. Let's pretend that connecting a mobile device to a TV becomes very easy, and controls or connected controllers are heavily improved to compete with dedicated gaming devices. For people who get high-end mobile devices already (smartphone users, people with tablets and high end media players, etc), the question will become "Is it worth $300-500 every ~5 years to have better graphics and a media streaming device?"
A final point is that we already have a platform which has the high performance range of consoles and the steady improvement curve of smartphones. It's called the PC. If there's one thing I've learned in knowing PC gamers, it's that very few of them have stuff from the latest generation. I know people who are still rocking the Pentium 4. I myself am using graphics cards from three years ago. Indeed, if you talk to many console gamers about why they don't want PCs, up there with the "It's too complicated" complaint you'll often find people feeling that they have to spend a fortune to stay up-to-date with the latest hardware. Of course, I argue to them that they don't NEED to get the latest hardware, but the notion is still there. This translates to smartphones. Not everyone wants to renew their contract and shell out big bucks every year, or drop $600+ off-contract, just to have something which is twice as fast.
So if everyone went the mobile route, what would that translate to? Fragmented performance levels. As a developer you can't JUST target the latest and greatest because then you're missing out on a huge part of the market. You need to make something which degrades gracefully for several generations at least, which could mean developing for something which is 1/5th the speed of your flagship. If you toss in tablets, which have larger batteries and can have faster processors, that can increase the performance discrepency even further. So what do you do? Develop something which can scale back to run on 10% of the flagship's resources, or not develop things that take full advantage of the flagship? The former is a challenge, and the latter just makes it not very appealing to upgrade to new hardware until a generation or two later... which starts getting back to the console mentality of strategic hardware releases that consumers and developers can all buy into.
I really don't understand what Carmack and Will Wright are getting at, in light of that. Pretending that smartphones are the only devices out there, there's nothing stopping you from having a 3-4 year development cycle. In fact, you can work it to your benefit. When you start making a game, you develop for the fastest thing out there. You get the core mechanics in place, basic graphics, sound, controls. You consider the fastest device, which is your standard dev platform for that game, to be the slowest thing it will target on release. A year rolls by and new hardware comes out. Great! While most of the team continues to work on a base code set and game assets which are optimized for your baseline hardware, you get part of the team to improve the game assets to take advantage of the newer, faster hardware. Continue to do this until it's release time, at which point you have assets optimized for all major performance levels. Alternately, develop game assets which are much higher quality than current platforms can utilize, targeting what you anticipate will be the performance level by the time you're ready to release, and then use tools to effectively downsample the assets... lower-quality shaders, smaller textures, simpler game models, fewer particles, etc. See also, PC gaming. Benefits: Assuming your computer is from within the past few years, you can take advantage of as much performance as it has to offer.