[BBobley]

Below is my idea for a future version of ReplayTV. My idea is to create what I call intelligent "ReplayTV Agents." My apologies about such a long post, but I've been thinking about this for a while and I wanted to be clear about what I was getting at.


I am a fairly new ReplayTV user. Like many of you, I have found that it changes the way I watch TV. There is no doubt that technology like this will soon become commonplace. I think all the Replay programmers deserve a lot of credit for what they've done. But I also feel that the current version of both ReplayTV and Tivo are only in the infant stages of what they will become in the future.


Recording the programs digitally is only part of what makes this technology so great. Both ReplayTV and Tivo do a fine job of this. But what is really going to change, I feel, is the way in which we are able to perform sophisticated searches against the program guide information to find programs we like. Both ReplayTV and Tivo do this to some extent, but they could do so much more. (Hopefully the upcoming MyReplayTV.com site will enhance these types of capabilities). As an IT person, I must regularly parse gigabytes of data to search for patterns. As such, I see the potential for using sophisticated search techniques for parsing the program guide which aren't being applied now. Let me give you a quick example. Using ReplayTV, my wife has set up theme channels to find and record many of her favorite movies. She might set up a theme to record "Days of Heaven," figuring that eventually the show will come on a channel somewhere and be recorded. She has set up many themes to record films she likes. But because you are limited to only one search criteria (that is, you can't use boolean logic like "title=foo AND director=bar AND actor=smith"), the system often records the wrong thing. The other day we ran into a different problem. One of my wife's themes was to record "The Haunting." Well, after several weeks we came home one Friday night to discover that the movie had recorded! The bad news was that several other shows we had asked to be recorded that Friday night did NOT record because they were on at the same time as "The Haunting." Since we had set up The Haunting theme weeks ago, we couldn't have anticipated this conflict.


But this message isn't about conflict resolution, per se. I'd like to see the "Themes" and "Zones" enhanced by what I'll call an intelligent "ReplayTV Agent", not unlike the intelligent web agents that can be used to search for things on the web. Here is what I envision a ReplayTV Agent could do. I could give the Replay Agent a list of my favorite movies, actors, directors, or themes. I could be specific (Movie Title=The Godfather AND Director=Coppola) or general (Movies with "shark" in the title, westerns, best picture winners, etc). The Replay Agent would constantly scan the program guide for these movies. It would then contact me to let me know what it found. I'd like to be able to set up multiple ways of alerting me. For example, the Agent could send me an SMTP e-mail or it could send me a pop-up or a message on the ReplayTV machine. That would be up to me. I could assign different conflict priorities (e.g. If movie "Godfather" is found, it should be set to record automatically and gets priority over anything else. If movie "Jaws" is found, record only if there is no conflict. If there is a conflict, send e-mail. etc). Ideally, this agent would run every night right after the program guide download. That way, if it found a movie on my list, it could contact me if there were any potential conflicts and give me plenty of time to react.


I see this idea of intelligent ReplayTV Agents as the next big step. Plus, less sophisticated users wouldn't have to use it, as the system would function fine without the use of the ReplayTV Agents. So its complexity would only be used by those who want it. What do you think?


BBobley

 

[FreezeFrame]

An interesting idea, I think it would be better to run the agents on the server side though. More horsepower, faster updates and the ability to crunch on more data. Also I would like to see the ContentBot results combined with human input(collaborative filtering style).


One idea I had was to allow anyone to become a critic. i.e. A user goes to the MyReplayTV.com web site and browses through the shows(Channel Guide, ReplayGuide, RawList...), as they do they rate the shows according to their own personal preferences. Then if they wish they could "publish" their personal picks for others to use as filters. As users rate shows, the software could seek out other "critics" who have similar likes and then allow you to add them to your ContentBot's filtering lists. Over time some "critics" would become popular and those could be advertised for brand new users to pick from! End users could pick handles, like "SciFiGuy" or "SoapJunkie" ... to make it a bit more fun!


I totally agree with your general point about PVR's. They are indeed already highly useful, and I am also looking forward to all the really cool things they are going to bring!



[This message has been edited by FreezeFrame (edited 10-29-2000).]

 

[jasonl99]

I'm by no means trying to start a "mine is better than yours" thing here, but what you have articulated (quite well, I might add http://www.avsforum.com/ubb/wink.gif ) is very similar to the Wishlists which are part of TiVo's 2.0 software (which you get now if you have DirecTiVo, and Real Soon Now if you have a standalone).


You can read more about it in depth on the FAQ page , but I'll give you an idea how it works.


First, TiVo doesn't segment its space like the Replay does - everything goes into one pool (and a giant one at that if you've done the upgrade hack http://www.avsforum.com/ubb/smile.gif ). So conflict management becomes quite a bit easier.


You set up your season passes (which record every episode of a show on a single channel). You can then prioritize these -- for example, Friends always has priority over This Old House.


With the wishlists, you create a set of criteria, and it checks against all channels, looking for matches. You can create Actor, Director, and Keyword wishlists, and they can be limited by category.


The nice thing as far as conflict management -- and I agree that this is really importat -- is that the Wishlists are incorporated into the early-referenced prioritizer. So you might have a wishlist with "Copola" as a director which has priority over Friends.


The only thing, from my perspective, that seems to be missing from TiVo's implementation is full boolean logic. But I'm sure that will come eventually.

 

[FreezeFrame]



I truly hope ReplayTV does not go the route of the TiVo mechanism. There is only so much an algorithm can do!!! Keyword searching and boolean logic are rather primitive compared to the advance processing of humans(wetware). Picking personal entertainment is highly subjective and won't lend itself to being done on low end embedded systems, it's a job best left to people!! What the technology can do is leverage that human activity.

 

[Lee Thompson]

Quote:

Originally posted by FreezeFrame: I truly hope ReplayTV does not go the route of the TiVo mechanism. There is only so much an algorithm can do!!! Keyword searching and boolean logic are rather primitive compared to the advance processing of humans(wetware). Picking personal entertainment is highly subjective and won't lend itself to being done on low end embedded systems, it's a job best left to people!! What the technology can do is leverage that human activity.

I agree. I'm very picky and set in my ways as to what I like and I bet I'd make a TiVo core dump on how picky I can be. http://www.avsforum.com/ubb/smile.gif


NetFlix.com actually has a similar system and it's usually wrong http://www.avsforum.com/ubb/wink.gif

 

[FreezeFrame]



I'm talking about the general concept of having "purely" algorithmic suggestions. There is only so much that an algorithm can do. Let's say there is a program coming on shortly that is about "Current Managers of the NFL". In that show is a clip about my high school QB, who happens to be a Manager of an NFL team. It's very unlikely that any piece of software will ever have enough knowledge about me to record that show, or just the part of the show I am interested in. Perhaps one of my high school buddies knows about the show, and knows I would be interested in the show. If he makes a suggestion to my web based suggestion engine, then I can decide to record it based on that info.


I'm not saying to not allow searching and picking, nor even algorithmic suggestions. It's just there are limits to what it can do. It seems TiVo is headed down the path where the will reach that dead end. A web based "collaborative" filter is much more open ended. It's a case where I think ReplayTV should do things differently, i.e. don't follow TiVo's example.

 

[jasonl99]

FreezeFrame,


I'm not so sure TiVo is headed down the wrong path. To use your example, you could do this with the new Wishlists:


Create a new "keyword" wishlist that uses the word "Coach." Then select the Sports category and Football subcategory. You'd now have a list of all programs that meet that criteria, on all channels. Then you could decide to let TiVo automatically record it for you, or pick individual shows off the list and explicitly ask TiVo to record those.



[This message has been edited by jasonl99 (edited 10-31-2000).]

 

[FreezeFrame]

In my specific example there would be no indication that the particular content was in the show of interest. Also I wouldn't even have it in my "wishlist" as it's not "content" I'm looking for. It's also something very particular to me and has nothing to do with my viewing habits. In other words there is probably no way to automate it. What I want is a way for someone else to tell me about it, i.e. collaboration. This is something humans do well, but computers don't. I'm not dis'ing computers/software, it's just that they have limitations and it pays to know what they are. TiVo are building an infrastructure that is heading towards more and more automation of the recording engine. I think a better architecture would be to head towarards an one that supports *both* automated and collaborative(human) mechanisms. Another aspect of this is the ability to leverage the efforts of others in finding content of interest. I think a collaborative architecture makes that much easier and much more likely to be useful than "box filtering".


B.t.w. did you see the bit on SVB last weekend about the new company doing PVR for settop boxes. http://www.mbtv.com


What I want to be able to do is send everyone on my PVR mailing list a "ShowMark" that suggest their PVR record that show. Everyone on the list would get the "ShowMark", run it through their filtering engine and record it on their PVR(if they get the show and allow me to make suggestions). It's a somewhat different model than the one TiVo seems to be heading towards.

 

[jasonl99]

Got it. You mean like I suggested to TiVo back in June http://www.avsforum.com/ubb/biggrin.gif ?

Quote:

I've actually had a this idea for TiVo for awhile. Imagine if you had a screen where you could put in your friends names along with their serial #. You'd then add a menu option for the shows detailed information that would allow you to send a message about the show to any/all of your friends; the could then choose to record or not record. Sort of like TiVomatic Friends

 

[FreezeFrame]

The TiVo suggestion algorithm will never be that good? Why? Because it lacks the information it needs to do a good job at it! In other words it's fundamentally flawed. The amount of information needed would require significant data interaction with the end user. Computers and the Internet(Web) provide an avenue for such interaction. Thus using servers to do the content matching will yield better results....


Ok, so not everyone want's to share that kind of information. Fine, if you don't want to do it, then don't. I don't consider the information any more secure just because it's on your PVR. The information on a PVR can be retrieved just as easily. So the security of the information is only as good as the company that controls it. Either you trust them or you don't. Therefore I see no disadvantage to doing this kind of stuff on a server, in fact there are significant benefits(processing power, connectivity, more sophisticated tools, ...).


What I am talking about is a more advanced concept that includes the capabilities of TiVo suggestions and wishlists. It also does more. A lot more.


As for calling someone to let them know. That's great, no problem, except for the following:


You have to reach your friend, or leave a message.


Your friend has to then program his PVR.


You have to call the rest of your friends and tell them also.


I want suggestions from a lot of people/groups.



I suppose the concept is similar in some ways to email lists and the incoming mail filters you can use in a browser. Except I want to push the functionality onto the server so it can also be merged with a contentBot results.

 

[FreezeFrame]

A couple of other examples of what you could do with the infrastructure I was talking about:


Parental Content Control:


The new parental controls are rather coursed grained and still leave gapping holes. Here is a better solution based on ReplayTV and web based agents. First modify the ReplayTV software so that the there are parental lockout options. The first option would be to allow content recording to be defined ONLY from the web site. The second would be to disable normal TV watching. Make these password protected on the ReplayTV. Thus the system could be setup with a known selection of pre-recorded content.


Then a parent who wants to define what a child watches on TV can setup the ReplayTV so that the kids only see what's been recorded for them. The parents can use the web interface to select the programs...


The hard part is picking what programs to allow the kids to watch. This is where an agent based system comes into play. Instead of every parent having to make these decisions, why not allow them to pick a web based referal service that does it for them. Such a service could be set up by anyone on the web and parents would be free to choose the service that most closely aligns with their beliefs/desires. This provides excellent leverage for parents and it does not infringe on anyone in anyway.


Ad Clipper:


Shock, horror, sometimes I actually want to see an ad! Let's say it's time to replace that rusting 7 year old econo-car and I'm interested in a nice big new SUV. Wouldn't it be great to be able to go to a web site and sign up for a list of shows/ads that cover SUV's? Then have these sent via my recording agent to my PVR! I'm sure advertisers and companies would jump at the chance to actually help you record and watch this kind of content


That's just a couple of examples, there are hundreds if not thousands of such ContentList services that could be setup. All you need is the engine

to process/filter and forward them to your PVR. Using a web based agent coupled to MyReplayTV.com would be a very natural way to do this.


b.t.w. I'm NOT thinking about getting a SUV, it was just an example. So please don't yell at me for being a gas hog, I drive a tiny little Honda civic which performs perfectly for my needs and it get 40+ mpg! It's also 14 years old! Hmm, I just figured out that over it's 120,000 miles it's probably saved about 3000 gallons of fuel and/or 4.5k dollars.


[This message has been edited by FreezeFrame (edited 11-02-2000).]