Contradictions...
[apologies in advance for the length of this post!]
A lot of people here are asking for dumb TVs. With Sky decoding. Or Freeview. Or NetFlick. Or DVD playback. Or Media Streaming.
A truly dumb TV has a single video (and maybe audio) input. Probably an of/off switch as well. And maybe volume, brightness and contrast controls as well. Uh oh, feature creep is setting in already.
A slightly less dumb TV as a set of inputs, allowing you to either have multiple devices connected simultaneously or to give a choice of how you would like to to connect. No Scart? use HDMI instead? As soon as you get multiple inputs the TV has to start having input source selection. This is the first point that starts confusing the hell out of the average user... average as in Joe Public, not an El Reg reader who by definition is usually capable of mashing a keyboard semi-coherently.
A less dumb TV also has a TV tuner built into it. Except that this depends on the connectivity to your video source... as in over the air (generally FreeView), Satellite or Cable. These use channel numbers as a matter of historical and remote control convenience, not for the ease of use of the end user. These channel numbers are stupidly not consistent between FreeView, Satellite and Cable services and don't integrate with the Input Selection which is also a channel selector in its own way. When a user turns on a TV they often have a TV channel in mind, for example Sky One or BBC 2, they don't have "150" or "102" in mind and instead they have to either learn the specific device's numbering scheme or use what is far too often an extremely clunky or slow program or channel guide to select a channel. A program guide? Now that's yet another source of confusion as they are usually far too hard to use, inconsistent in operation and rely on short cuts on the remote control. Why does the remote control not have a "next day" button and instead the user has to learn to press the Blue button instead. Many Joe Public users just don't use the program guide or the channel listing, can't quite remember the channel numbers and instead repeatedly hit the channel up and down buttons on their remote until their desired channel eventually appears. Eventually? Yep, with digital decoding we're gone from the 1/25th of a second channel change to waiting for the first I-Frame in the MPEG stream, which on heavily compressed streams can often take a second or two.
We haven't even got to anything that would be considered a "smart" TV yet, and already we have layer upon layer of User Interface (or User eXperience) failures. The Smart TVs barely stand a chance.
As for where I'd like it to be heading...
Most users frequent no more than six channels (there's research on this somewhere). Give them a remote control with six channel buttons clearly marked with their channel preferences. This means a smart, two way remote. If the remote controls are cheap enough, supply more than one - one for the kids, one for mum, one for dad and so on. The alternative is one what can be switched between user profiles very quickly and simply. If information is stored on the remote control, back it up to the TV itself this way when a remote is broken or lost or multiple remotes are used they can be configured very quickly. Little things like this are really appreciated by users.
Kill Infra Red remote controls. They may be quite power efficient but they're one of the reasons behind appalling user interface response times on TVs and similar devices. Two way wireless communications, e-ink display and quite possibly wireless charging would be the way to go. Stick with buttons and not touch screens. The older the user, the more they appreciate and can handle button interfaces - they are far superior for limited functionality applications than the more adaptable but harder to use (and hugely more expensive when done well) touch screen interfaces.
If "smart" TV functionality must be implemented in a TV, do it using an established platform and not recreating the wheel with poor quality often underpowered components and awkward operating environments. Android is the ideal platform for this and removing the telecomms stack and associated chipset would save a chunk of the cost. It even comes with wifi management which saves another problem... In theory, iOS or even WinRT would do the job, however Android, if the components are correct, the applications are coded efficiently and unnecessary baked in applications are removed it can be quite efficient and would not have trouble rending at full TV screen resolution as similar resolutions are in place on mobiles and tablets already. Android, iOS and even WinRT would have to be tweaked to be usable without using touch input and without having a horrible pointer overlay alternative. Once tweaked though, there would be a huge market for applications that either just work on this kind of device or have been specially created for them.
Integrate the channel and input selectors into one seamless list and ditch the entire concept of channel numbers. Channels should be sorted by separate sorting text behind the scenes which will sort the channels into the order the user expects. i.e. BBC One, BBC Two, BBC Three, BBC Four and so on. Alphabetically these would be "BBC Four, BBC One, BBC Three and BBC Two" which isn't quite so user friendly. Regional variations on the channels are an added complication, in theory they should be listed together with their parent channel, such as "BBC One North, BBC One Scotland, BBC One South" this makes for an appalling use case for the users that regardless of the interface will still frantically press Channel Up and Down buttons to select their desired channels. We'd doubtless see a proliferation of channel named "Arrdvark 1" and so on but there is no perfect solution.
If the broadcast content is the same, always select the HD option rather than the SD when presented with two otherwise channels. This should be togglable of course, but given ease of use it should be automatic and not needlessly presenting a user with two versions of BBC One is a good thing.