Siri vs. Wildfire
It’s very interesting to see the resurgence of consumer-grade AI-enabled voice recognition in Siri, the new digital assistant component of the iPhone 4S. In reading the largely positive reviews of the cloud-assisted on-device service I’ve been surprised that there has been no mention of Wildfire – a service I was one of clearly very few people in the U.S. to have used back in the ’90′s. The parallels are many – the primary difference being that Apple seems to have brought the concept to the masses, where before it was an expensive, proprietary solution geared almost exclusively to professionals.
There’s not much information floating around any more about Wildfire. At the time I subscribed to the service – around the early – mid nineties – it was run in the U.S. by Virtuosity. The cost was prohibitive to most – I was paying around $280 a month if I remember correctly – mainly because I think they had to dedicate a server to each user of the system.
But I have to say that, even back then (almost 20 years ago) the system was pretty awesome. For one, you could use Wildfire with any phone – mobile or landline – everything was server-side and voice / tone driven.
Most reviewers of Siri have mentioned the Scully era “Knowledge Navigator” piece by Apple, which was certainly a pivotal vision piece (I’ve often referred back to it – a Vanavar Bush type reference point) – but you have to give whoever came up with Wildfire credit for actually delivering (cost aside) a service that was almost as good as Siri, a good 15-20 years earlier. I wonder how much the Apple team studied Wildfire in the development of Siri – I suspect a lot.
A few parallels, and contrasts
Siri is cloud-based, but (since it currently runs only on the iPhone 4S) may also be leveraging special device-side hardware to do it’s job also. Wildfire was purely a server side (cloud) implementation relying on a combination of voice rec (which also relied on signal quality), tones, and caller ID. The advantage with Wildfire is that you could use the service with theoretically any phone – mobile or landline (once you registered each number with the service, and all had caller ID – so Wildfire would recognize you).
Siri is clearly anthropomorphic. Most reviews I’ve read mention the fact that over time you start to think of Siri as a “person”, to the point even of asking Siri personal (in fact frequently existential) questions. Same with Wildfire. Wildfire had a pleasant voice, a sense of humor (“she” told jokes), and was clearly designed to be a model personal assistant. They both say / said things like “let me work on that for you …”, i.e. conversationally.
Siri doesn’t seem to interface (yet) with anyone other than the owner of the device. This is a very interesting point. What I mean is that I have not yet seen a usage scenario – except for typing out texts or emails from voice – that actually interface with others or (even more interestingly) Siri to Siri (like a machine to machine [M2M] interface). I bet it’s coming. Wildfire could not only interact with the owner, but also others, to whit …
While I have not seen a review yet explicitly describe working with Siri during a phone call, I would presume that it is possible. The question is how. With Wildfire, if you were on a Wildfire managed call, Wildfire was always “listening in” (a bit disconcerting at first). If you said “Wildfire” during a call, “she” would come on the line and say “I’m here”. You would hear the “I’m here”, but the other party(ies) on the line would not (they would hear the initial “Wildfire” from you, and other commands, which was funny to other people frequently – one associate of mine referred to Wildfire as my “digital bitch” – his words, not mine. This stand-by functionality takes a Siri-like service to another level, and if it is not a part of the current Apple implementation I would almost guarantee it is coming.
The thing with talking on a mobile device is you are almost always doing something else at the same time (unlike a land line) – running (as in the Siri promo video), driving, etc. So it’s awkward, even when you are just sitting doing nothing else, to change modes to do the myriad of other things that you almost always have to do during a call – look up a meeting, an address, take a note, etc. Wildfire was particularly amazing in her ability to fluidly manage calls – especially call filtering and ad-hoc conferences (a feature I used constantly).
A typical Wildfire driven conversation
ran something like this:
(I’m driving. Phone rings, I’ve set it for Wildfire to screen. I answer on the hands free system installed in the car. Remember, this is the early / mid nineties.)
The point here, with Wildfire, is that the assistance was persistent during calls, which is I think a clear use case and one that I presume Siri enables, or will enable.
John Gruber, in his early review of the 4S, brought up a very interesting point regarding whether or not Apple would offer third-party application access to Siri via an API. And he was a little stuck on the issue of how to manage request conflicts of Siri from potentially multiple apps. In other words, as basically a system service, iOS determines (as Gruber describes nicely) when to use the native Stock application vs. Wolfram Alpha stock data. But in fact there are many instances where the user may want to override – temporarily or permanently – the OS defaults. And there should be a mechanism to do this, much like you can override the default launch application for a given file type on Mac OS if you prefer one application over another.
But with regard to voice, I think the key here might ultimately be in simply how Wildfire was used in conjunction with another “application” – that being a voice call. To access Wildfire assistance, you simply called “Wildfire”. Yeah, it made for a lot of “digital bitch” jokes, but it also worked perfectly. So, in other words, I’m wondering if there won’t ultimately be a simple syntax that allows not only Siri to be addressed (system-wide, as now), but also to help Siri address specific apps (as either temporary or permanent override to the OS defaults). So, if a developer implements a Siri API into their app, they become “addressable” I would presume by a verbal unique key or equivalent. Thus, in the future, regarding Gruber’s example, a request could be:
“What was Apple’s stock price 10 years ago according to Wolfram?”
Which would then override the OS default to use the Stock application, using the developer / application keyword “Wolfram”.
I do believe that we are on the verge of a transition long anticipated by Steve Jobs that will ultimately rival the transition from mouse to finger (multitouch), which is finger to voice (Siri). And I’m certain that scenarios of voice control that go far, far beyond this current implementation are alive and well in the recesses of the Apple campus, and have been for years. This has Steve Jobs written all over it. I’m sure he took Wildfire apart back in the day – noted what was interesting, valuable, intuitive, personal, human – and also what was technically stupid, too early, inhuman, unintuitive, not leveragable. Just as he did with the portable music player, the iPhone, the touch interface, the tablet – thus voice.
I was always a little self-conscious walking or driving around talking to Wildfire. But it was in it’s time a brilliant implementation – way, way ahead of it’s time. And I think the Wildfire system (and probably others I don’t know about) deserve more credit than they’ve been given. I look forward to seeing how Siri does in the marketplace, and what plans Apple has for expanding the service – across the OS, and out to developers. Definitely a game-changer.
From the archives:
Here are two wholly inadequate and dated demos of the Wildfire system in practice:
Here is a copy of my old Wildfire crib sheet. Never thought I’d pull this out again: