Hi Forum Folks,

I think it’s about time we, as the best collection of chatbot builders in the world, looked at Siri with a critical eye.

Not that everyone will have an iPhone of course, but those that do .... and have a view on what is ‘good’ and ‘bad’ in natural language interaction .... might take a moment to _test_ it ?

I’ve been watching the snappily named website emulating conversations and can’t say it is anything that hasn’t been seen before.

Thoughts ?

Phil

I’m eagerly anticipating the arrival of my iPhone 4S, I ordered it last week but there have been shipping complications. I’ve read some very interesting (and humorous) responses from Siri to the requests thrown at him/her, cant wait to test it out!

Will try to get back on Friday with some feedback.

Cool - nice one Shaun, you are the first of our testing team consider yourself the vanguard ...

There has been a lot of hot air about what it is and isn’t going but no-one has really applied a structured test to Siri yet.

Cheers,

Phil

The simple answer is. . .

Yep, it is top notch, but nothing I’ve not seen before and my chatbot is better than that.

Let me explain:

Yesterday an Iphone 4s arrived on my doorstep. One of my family members just had to have it and I spent the evening setting it up. So, you can classify this as “first impressions”.

Top notch
System design - The hardest thing about Siri is figuring out how to use it the first time. There is no icon on the phone and it did not come with instructions. To use Siri, you hold down the circular button on the phone for a couple of seconds, a multi-tone sounds, and the phone is put into voice recognition mode. The system recognizes your input, creates a response and attempts to perform a task if you spoke a command.

The thing that makes it top notch is the level of seamless integration. You can say “Play music”, or “Call Home” or a variety of other commands and it just happens. Location aware search works well. My “Find me a pizza place.” brought up the closest location with distances. Although I did not order a pizza I believe I could have with a few more touches. Although I did not try it I believe I could have dictated a text message and had it sent much more quickly than type on a virtual keyboard. It all just flows together, making navigating on a cell phone easier.

Prediction: within the next 5 years all “smart phones” will have voice recognition/intelligent agents built in.

Voice recognition - The voice recognition is good. It got most of my inputs correct, even with a TV playing in the background.

AI - The AI feels like a well built Chatbot. For some responses it has random, on topic replies.
The personality has an edge but is not very deep. It feels more like a tool than an assistant.

Nothing I’ve not seen before
For those of us who are getting up in years and need reading glasses to read the menu at a local restaurant, cell phones can be a pain. The requirement to put glasses on to read or respond to a text message or to select the right contact from a contact list is a burden. As such, my favorite cell phone became the one that allowed me to use voice recognition for my contact list.

Voice recognition has been around for a while. It has been reported that Siri uses Nuance as its voice recognition system. You can get voice recognition as an app.

Other alternatives have been available, some including attempts to include agents.

Google added voice input capabilities to their Chrome Browser (I demonstrated this capabilitiy with Skynet-AI earlier this year), voice search and have voice actions for Android powered cell phones.
Speech is thought to be a standard input type for future HTML pages/forms.

Prediction: 5 years from now all browsers will include speech recognition as a standard feature.

AI - Of course chat bots have been around for a decade. I demonstrated my first location aware context based search chatbot almost 10 years ago. The problem is that it has become increasingly difficult to mesh services with increased security levels and cross domain policies in browsers. Some of the things we could do a few years ago are no longer viable (and in fact I am currently making changes in my bot to accommodate this). Standardized web based services are required to add intellegence/realtime data to your bot. If you would like to try a voice operated chat bot use Chrome and talk to Skynet-AI.

My chatbot is better
Better is relative, but let me try to quantify it.

Runs anywhere - Skynet-AI runs on anything that has a modern web browser. It runs in cell phones, Ipad, Ipod, Iphone, Andoid, Blackberry, Web TV, Playstation 3/Portable, Macs, Pcs, Unix workstations. When I say it runs anywhere, I am not talking about the UI form linking back to an application on the server. The AI actually is running in the device. These capabilities are provide by JAIL (tm) JavaScript Artificial Intelligence Language.

Text and Voice input - For Skynet-AI, voice input/output is based on the platform being used. In the future, I believe most devices will have a standardized voice input option. But, there are times that text input will be easier/preferred (like with proper names or in environments that are very noisy or that require silence).

Customization - Of course as a developer I can change Skynet-AI to my heart’s content. But, that capability could easily be given over to end users. People will want to customize their intelligent agents to handle their preferences, favorite commands and workflows.

To summarize: Siri is pretty neat. The real wow is how well integrated all of the capabilities are. In the future you will want your intelligent assistant available on any device you use from your desktop to your cell phone. The cloud makes this possible. Now it is just a simple matter of programming.

Merlin gives an excellent summary of Siri- it is the integration (location, maps, phone, contact, email, etc) that makes it “wow”, but not so much the intelligence of the banter- the chat bot like responses are generally quite predictable, at least to chat bot developers who can recognize the classic input-to-canned output patterns.

@Merlin:

I think Siri has nothing to do with AI. I believe these are just a bunch of standard prhase (call xxx), reschedule meeting toooo. They probably have recorded ‘commands’ to PA’s (what are managers demanding from their PA’s?) and have tried to create an equivalence. An example like ‘Order Pizza’ is a standard phrase. I can imagine they have tested this a zillion times (or a bit more). Have you tried: ‘seafood’ or ‘mediterian disk’ or ‘something with pasta’ as well?

Prediction: within the next 5 years all “smart phones” will have voice recognition/intelligent agents built in.

The iPhone 4S does notehave speech recognition. It just sends the audio clipping over the internet, and one central Siri is recognizing your words, and sends the encoded commands back to your phone.

Are you sure iPhone 4S uses Nuance? I thought it was native Apple coding, that’s why they purchased Siri a while ago:
http://www.chatbots.org/conversational/agent/siri_virtual_assistant_iphone_3gs/

Voice recognition: I recently read about the difference between ‘voice recognition’ and ‘speech recognition’. Voice recognition would be about recognize the voice, i.e. identifying the user, Speaker ID, where speech recognition is about recognizing the words.

I’m still a bit skeptical about Siri. I absolutely belief that this is progress for native English speakers, living in the States, But the rest of the world, waiting for unified structured data (for example about houses, traffic or restaurants) with thousands of languages and zillions of accents in English, and uncountable cultural backgrounds, will have to wait a long time before this really will work. And then we’re just talking about ‘simple sentences (2-6 words)’. I’m not negative, but I’m just tried to put the hypo into perspective.

@At all: I
still don’t know how Merlin looks like, but looking at our profile picks, shall we agree this thread is ‘for bold men only’

AI is open to much interpretation. I suppose Siri has as much AI as any chatbot. It comes out of DARPA research.
http://www.fastcompany.com/1785221/siri-ously-darpa
http://www.infowars.com/introducing-siri-darpa’s-ghost-in-apples-machine/

The party tricks that Siri performs could be done by any good chatbot using AIML, ChatScript or JAIL. It is all about putting in enough standard phrases and aliases. I haven’t tried any other food selections but I would expect most would work. In the US the yellow pages provide standard classifications, add a thesaurus and viola a search for your favorite food.

On a side note, getting the Iphone away from my family member is like taking a bone from a pit bull. The funny part is the owner doesn’t use Siri and finds it confusing.

When I say voice recognition I mean the same as speech recognition. In the old days you had to train the systems with your voice before it could understand what you would say. Now, it is speaker independent.
http://en.wikipedia.org/wiki/Speech_recognition

The Iphone 4s does have speech recognition. Just because it uses a client/server/cloud model does not diminish the capability. You can take a voice note and have it do a search independent of the Siri app.

My prediction is that all smart phones will end up having speech recognition. Low end models will use a client/server/cloud approach. High End models will run it directly on the phone with the corresponding benefits of eliminating the round trip lag time. The Intelligent Agent will run the same way.

I do not believe Siri did their own speech recognition, most likely Apple is using Nuance. Typically I would expect Apple to do some custom development to take best advantage of their new data center.
http://www.macrumors.com/2011/05/06/apple-negotiating-deal-with-nuance-for-speech-recognition-in-ios-5/

English speaking users will derive the most benefits, countries with Yellow page like digital directories will be able to integrate. Accents are yet another reason that the ability to accept text input will be important in the future.

  Yep - top notch
  Nope - nothing I’ve not seen before
  Look - my chatbot is better than that
  Or - The biggest fraud of the 21st century (Maybe an overstatement, but still!)

There hasn’t been a gimmick as blatant as this.

While Apple marketed Siri as a new innovation (typical), Android has had much of the same functionality for more than a year now. Google just didn’t put much effort into announcing their Google Voice Actions, so a lot of Android users aren’t even aware of what their device is capable of. Apple’s event sort of hinged on the announcement of Siri. What else did they show today? Oh that’s right an app that lets you send real postcards for $3 a pop. Like I said, Siri was the star of the show.

Alright so by now you’ve likely checked out Siri and maybe, just maybe, you’re wondering if your Android does all that. As Verizon puts it, “Droid Does. ”

http://www.youtube.com/watch?v=gGbYVvU0Z5s

http://www.talkandroid.com/65000-voice-control-apple-picks-a-new-fight-with-android/

Hype something up as a revolution and sheeple people out of their hard earned money… Only Apple can!

I have an I4 and have no intention of getting hoodwinked by the latest gimmick on the market.

Genesis - Oct 22, 2011:

Alright so by now you’ve likely checked out Siri and maybe, just maybe, you’re wondering if your Android does all that. As Verizon puts it, “Droid Does. ”

Hype something up as a revolution and sheeple people out of their hard earned money… Only Apple can!

I have an I4 and have no intention of getting hoodwinked by the latest gimmick on the market.

No points for making this a fanboy war…  http://i.imgur.com/rCtjR.jpg

Carl B - Oct 22, 2011:

Genesis - Oct 22, 2011:

Alright so by now you’ve likely checked out Siri and maybe, just maybe, you’re wondering if your Android does all that. As Verizon puts it, “Droid Does. ”

Hype something up as a revolution and sheeple people out of their hard earned money… Only Apple can!

I have an I4 and have no intention of getting hoodwinked by the latest gimmick on the market.

No points for making this a fanboy war…  http://i.imgur.com/rCtjR.jpg

Not a fanboy. I dislike any company that runs a sheep-herd.

Speaktoit Assistant for Android- http://www.youtube.com/watch?v=myE498nyfGw

Vlingo Personal Assistant for Android - http://www.youtube.com/watch?feature=player_embedded&v=t1X34uNymhE#!

Refuse their wire brush and blowtorch!

@At all: I
still don’t know how Merlin looks like, but looking at our profile picks, shall we agree this thread is ‘for bold men only’

Well, let me put some hair into the conversation

My prediction is that all smart phones will end up having speech recognition. Low end models will use a client/server/cloud approach. High End models will run it directly on the phone with the corresponding benefits of eliminating the round trip lag time. The Intelligent Agent will run the same way.

I’d tend to agree with you on that one. Though I have my reservations on the ‘client/server approach’, not that it wont be the ‘preferred model’ but mostly with it’s privacy. I think I’d prefer locally-run agent. I mean, think about it, what a treasure trove if you’d have an central AI recording and analyzing everything you say to your phone (in combination with where you are,....) Mmm, too easy pickings

The trade-offs are similar to any Cloud vs client discussion. Until the client hardware has the processing power it can’t run the same app. Usually then, the UI runs in the client and the processing takes place in the cloud. If you have watched the ebb and flow of technology you have seen the constant migration between mainframe/server/cloud and terminal/client/app.

Assuming the client has the hardware capabilities (which is ever only a matter of time and Moore’s Law), then the choice will be left to the end user.

Cloud - Do you want to run your agent on multiple devices and not worry about if you loose your device all you data will be compromised and you will need to start from scratch training a new agent.

Client - Run on a primary device without the need to be on-line all the time. No client server lag.

There will also be mixed models. Take a look at how people currently use email and you can extrapolate with an intelligent agent.

Erwin Van Lun - Oct 21, 2011:

@Merlin:

I think Siri has nothing to do with AI. I believe these are just a bunch of standard prhase (call xxx), reschedule meeting toooo. They probably have recorded ‘commands’ to PA’s (what are managers demanding from their PA’s?) and have tried to create an equivalence. An example like ‘Order Pizza’ is a standard phrase. I can imagine they have tested this a zillion times (or a bit more). Have you tried: ‘seafood’ or ‘mediterian disk’ or ‘something with pasta’ as well?

Prediction: within the next 5 years all “smart phones” will have voice recognition/intelligent agents built in.

The iPhone 4S does notehave speech recognition. It just sends the audio clipping over the internet, and one central Siri is recognizing your words, and sends the encoded commands back to your phone.

Are you sure iPhone 4S uses Nuance? I thought it was native Apple coding, that’s why they purchased Siri a while ago:
http://www.chatbots.org/conversational/agent/siri_virtual_assistant_iphone_3gs/

Voice recognition: I recently read about the difference between ‘voice recognition’ and ‘speech recognition’. Voice recognition would be about recognize the voice, i.e. identifying the user, Speaker ID, where speech recognition is about recognizing the words.

I’m still a bit skeptical about Siri. I absolutely belief that this is progress for native English speakers, living in the States, But the rest of the world, waiting for unified structured data (for example about houses, traffic or restaurants) with thousands of languages and zillions of accents in English, and uncountable cultural backgrounds, will have to wait a long time before this really will work. And then we’re just talking about ‘simple sentences (2-6 words)’. I’m not negative, but I’m just tried to put the hypo into perspective.

@At all: I
still don’t know how Merlin looks like, but looking at our profile picks, shall we agree this thread is ‘for bold men only’

Although they don’t reference Apple directly, Nuance is winning from the deal:
How Siri boosted Nuance earnings

Nuance alluded to the gains from the iPhone 4S halo in its statement and prepared remarks. Non-GAAP revenue for Nuance’s mobile and consumer business was $118.7 million, up 31.6 percent from a year ago. The company said it landing new bookings and design wins from Amazon, CIBC, Comcast, Daimler, Diebold, Ford, Kyocera, LGE, Nokia, NTT DoCoMo, On-Star, Renault, Samsung, Sony Ericsson, State of Maryland, Time Warner Cable, Tom Tom, and ZTE.

If you look at the list of companies I would say my cell phone prediction is pretty safe. I should probably add that you will soon see high end voice enabled TVs, Cars, and appliances.

iPhone 4 Explodes Midflight on Australian Airline

http://news.yahoo.com/iphone-4-explodes-midflight-australian-airline-224014583.html

“While on Australian flight Regional Express ZL319 Friday, a passenger’s iPhone 4 (not the iPhone 4S, which is Apple’s latest model) suddenly started “emitting a significant amount of dense smoke, accompanied by a red glow,” according to a Regional Express statement.”