I don't understand how this article rose to the top page of HN... it's not telling us anything that's interesting, new, or news in any way. Artificial intelligence has been around since the dawn of computers (think Clippy), helping us do our shit. In fact, AutoCAD is a bunch of 'artificial intelligence' that's letting you draw vectors and write code in a more intuitive way (ok, not exactly and technically, but you know what I mean)
It seems like the author of this article has never in their life seen a computer do anything more than push numbers, but machine learning and AI have been around for at least a decade. This is just a very mainstream case of very well developed AI.
Saying that 'this is the start of a revolution' is not only ill-advised, but totally baseless. What's the innovation? What has changed in what was previously possible? For a revolution to occur, there needs to be an innovation that causes expansion of technological capabilities. Although Siri is (supposedly) great, it does not represent an expansion of technological capability. Just because the team behind Siri spent years building out a great voice activated AI system does not mean that we all can now do it. Computers have not become smarter... they are the same as they ever were. People have not realized a more efficient way to create AI... it's the same as before.
I'm truly surprised that such a shallow and misguided article made it to the popular page in a community like this.
Hi Author here. Thank your for the harsh words. I guess I gotto take it to the chin.
Please allow me explain my point in another way. I think Siri is to A.I. what Apple II is to computers. Does this analogy explain what I am trying to tell better?
I think that was clear from the article. What jenius, I think, was saying is that you don't offer much support for this claim. How, for example, can startups get on the AI-as-a-service bandwagon without having the resources and experience of a company like Apple?
Exactly right john_b... the author's follow-up comment addressed nothing that I said, just made another baseless claim - that "I think it's like the apple II of computers". Why? What makes it like the apple II? You have to back yourself up. I personally think they are nothing alike, and here's why:
The Apple II was a revolution of personal computing. It brought the personal computer to the masses, and made it infinitely more accessible in both form and function. The form and parts for the Apple II became the base for all other computers. Nobody had ever made anything like the Apple II before. The Apple II on its own sold like crazy, and brought the entire company to success. Siri is a small piece of software on an already successful device (which is significantly more 'revolutionary' that siri on its own). The idea for siri has been come up with and implemented many times before, in the exact same manner (see the many android comments). Apple just worked really hard on it and they allegedly did a good job.
Hey the author again. :-). It is true that in the article, I did not explain why I think that the era of A.I. started on October 4th. The article starts with the assertion that it has begun and moves on from there. I guess more than anything, the article is a call for action. I wanted the discussion to start and it did.
If I can present my chain of thoughts leading up to my decision in an interesting way I may write another article called "Why I think Siri is the Next Big Thing".
But here is my chain of thought for the curious mind in a very crude way:
What I did was to observe the moves of Apple and try to figure out their strategy. I put together the information that is available on the Internet. I connected the dots.
I knew Apple bought Siri for about $200 mil. I knew about Siri's history. You don't buy voice recognition software for this price. I watched the Mossberg-Jobs interview where Mossberg keeps insisting that Siri is in the search area and Jobs keep correcting him saying no Siri is in the A.I. area. I watched how the critical question "What are you going to do with it?" got lost during the conversation. When Mossberg said that the iPad is great for consumption but not for creation, it was very clear that Jobs did not see it like that at all. I also watched some Apple videos from late 80's which depicts a professor talking to his computer and getting his daily business done. It seemed to me that they had this vision all along and now they were implementing it.
Then I realized how iCloud fits in to this all. With the introduction of Siri it was clear to me that the next frontier the competition is moving to is A.I. Apple is going all in. IBM is pushing hard. Apparently, Google is baking some stuff.
While I was thinking about all these and trying to find its meaning, the interview between Fred Wilson and Carlota Perez helped me to put it all in context.
I think this is the time to invest in A.I. If a VC expects huge returns from his investment in 5-7 years, I think A.I. is the right way to go. If an entrepreneur wants to do something amazing, I think A.I. is the right way to go. My speculation is in 1-2 years this is going to be all we talk about.
I think A.I. is where the puck is going now. I might be wrong, I might be right. We will know soon enough.
I have thoughts on the particular question you asked too. But I cannot cram every bit Of thought I have to one blog post. It would make a terrible article. This article makes a prediction baldly. It asserts a threshold in A.I. is crossed and it's implications will be big. Diverging into startup funding in this area would just not fit in well. That deserves its own post. :-)
Also guys -- remember smarterchild? Did just about the same thing in terms of the artificial intelligence (not the voice recognition, but that wasn;t the author's point)... how long ago was that?
What a blindingly ignorant rant. A couple breathless prognostications that really annoyed me:
"If A.I. knows about insurance, will you ever google for it?"
Uh, all the "A.I." is doing is googling it. So basically Siri will use speech-to-text (which I've had on my Android phone for years) and put it in the google search box for me. Then it will read the first result. Zippity-doo-da.
"We will teach it how to use Excel."
Can you possibly imagine trying to tell a speech recognition system to input into a spreadsheet? Quite a step backwards, if you ask me.
Look, the interpretation of language into meaning and then into commands has been around for a while, and it's been constantly improving. Siri is another product in that crowded market. It's really swell that they included it in iOS; like I said, I've had it on my Droid for two-plus years and it's pretty handy. This is the worst kind of blind Apple fanboy-ism. They are a good company that makes some great products. Stop it with the hyperbole simply because you want to justify your new phone purchase to your wife before you're off-contract.
No, no and god no. Siri is not voice recognition. It is not even in the same universe as the voice control in Android.
And yes of course A.I. will use one or more search engines to look information up. But Google might not necessarily be one of them and consumers will not care two cents as long A.I. gets the job done. My point it that A.I. will fly over Google's moat unless it is Google's A.I.
So does Google, I am pretty sure it remembers your previous searches and knows therefore that if you are searching for "Paris" you mean the city and not "Paris Hilton" (or vice versa).
Kind of like how I just typed "When is thanksgiving day?" and Google told me the date. It understood my question and gave me the answer. How does Siri differ?
Try "will it rain tomorrow", or "should I wear a coat tomorrow"? Siri is an answer engine that tries to determine the meaning of your query and answer it. Google in recent years has started to add those capabilities to it's search interface, but it is not there yet.
Siri is much more like Wolfram Alpha or IBM's Watson then a standard search engine.
Bullshit - Siri has some special loops for recognizing questions about date, location and the weather. That's it. It really is not more advanced than "call xyz", except it not only uses the address book, but also the calendar and the weather service.
... Siri is a front-end to Wolfram Alpha and other data providers. It roughly identifies what you're looking for and forwards it along to the best data provider. Google has the advantage of being one of those data providers.
I love how I'm being downvoted to oblivion when nobody is able to produce a single example of what Siri does that other products don't. That's my entire argument, that the technology already exists and is not some major revolution. I guess no place is impervious to Apple fanboyism, even HN.
Yaaa....I think this flew over your head. You are saying that, for example, Wolfram Alpha is like Google. They are completely different: One gives you links that you still have to go through to reach what you need and one gives you a direct answer. Even Woz said in his interview on TC...we don't need search engines, we need answer engines. To say Siri is a search engine with voice to text is a great mistake. Voice recognition is just the outer layer that makes the experience a human experience.
I don't take the words of the author as a prophecy but I see where he is going; and isn't that what science fiction authors kinda did 20-30 years ago? They dreamt; and I wish more people can dream of greater possibilities.
And I think that my post flew over your head. Wolfram Alpha and Google are different, sure, but at the end of the day they're simply algorithms that take requests, process the data to find the best answers they know how, and return results. That's it. To call those, Siri, or anything else more than that is intellectually dishonest.
Siri might be a great interface, and I bet it's really handy to be integrated into applications. I'm not hating on the concept or the product; it just annoys the hell out of me when people pretend it's something it's not. It shows a lack of understanding about how these things work.
Remember the hype about the Segway? "Walking is obsolete!" some people proclaimed. It turns out people didn't end up liking it because it made them look like lazy idiots. Do you really envision having a conversation with your phone, saying things like "I feel like Italian food today"? I don't. If I feel like Italian food, I'm going to ask my handy device where I can get some. And my phone does that without the "magic" of Siri.
Yes but they are not generally seen as mere transportation. If you use it as that you kinda look like you are a overgrown teenager. If you use it as a platform for tricks and jumps it is again fairly dangerous. This time not in lives lost but in bones broken.
When I see Tony Hawks Segway Pro in the shelves because people want to do those tricks but are afraid to break their bones, then maybe then Segways will be cool. All I see on the horizon is "Ride your Segway like Woz" the iPhone App that really only plays 3 videos (But you know how hard it is to get a video into iTunes).
You use a reasonable amount of energy skateboarding. I would guess 300-500 cal/hr depending on what you are doing. You'll notice there are no fat skateboarders. The Segway didn't take off for a combination of factors. It looks lame, there's no place you can ride it legally, it costs the same as a motorcycle but you can't finance it, etc.
While I agree with you in principle, I feel like the conceptual use cases you're imagining implement current practices with basically a speech to text. What the author is suggesting is that it would be beyond this, skipping out the step where you determine how you do something, but instead telling Siri what you want to do. For example, with excel I'm not sure he envisaged something like "A5 is 20, A6 is 30" and so on, but more like, "Take my taxes from this pdf and put them into a spreadsheet".
That said, irrespective of any kind of AI revolution Siri may (or as I suspect) may not induce, I feel like predicting it's impact before we've even had the chance to test it out in any kind of capacity (OK the 4S has been on sale for two hours, but you know what I mean) is somewhat short sighted.
Except Siri doesn't do anything of the kind. It would be trivial to build "take my taxes from this PDF and put them into a spreadsheet" into Siri (or indeed any speech recognition engine). But that would only be one more special case. The code is literally "if (recognizedText == "take my taxes from this PDF and put them into a spreadsheet"){writeExcelSheet(parsePDFTable("taxes.pdf"))}
Sure, and I'm not for one second Siri could do that, but
a) I was under the impression Siri was doing some kind of semantic analysis rather than straight text recognition?
b) I just meant if we suspended disbelief and Siri was this magical AI tool it would do something like that, my second point is just that because we don't know what Siri can/will do it seems ridiculous to say it's going to change the world.
I totally agree with you, I'm just saying there are almost two separate discussions here, one being Siri and it's implementation, the other being the impact of AI on our lives. If Siri was the new AI, then these would be the same discussion, but as I don't think it is, it's not.
Ahh, but the devil is in the details. In the given example, "Take my taxes from this pdf and put them into a spreadsheet", it sounds really nice. However, consider this: Anything that you can't tell a person to do, you won't be able to tell this system to do. If you task a person with "taking your taxes and putting them into a spreadsheet", they're going to ask, "how?" "What's the purpose?" "What do you want the spreadsheet to look like?" "Do you want everything from your taxes on one sheet?" The list goes on and on.
So, while the idea seems really nice, the idea that this or any A.I. implementation will be the kind of insane revolution the OP is predicting is laughable.
Or is Siri simply the practical application of machine learning and NLP which we've manage to derive from off-shoots of AI research?
I think it will be some time yet when I can make a natural language query and have an AI agent fetch relevant data, sort it, and present it contextually. Or sketch an idea for a building and have an AI do all the hard work of creating the blueprints.
In the near future though it's obvious that we'll see more practical applications of AI in every-day products and I think that's pretty neat.
>>> a natural language query and have an AI agent fetch relevant data, sort it, and present it contextually.
What about IBM's watson : you asks it a question: am i sick? it gathers symptoms from you , looks in the medical literature and concludes your possible illnesses and in what likelihood.
I think it can also refer you to the relevant literature.
>>> Or sketch an idea for a building and have an AI do all the hard work of creating the blueprints.
bluethink house designer[1] does that.You enter a very high level drawing , and it fills in all the details according to regulations, cost efficency and best practices.
I won't judge siri, as I do not know the details of technology or the machine learning algorithms used by it, but I would say this much: your one line pitch might just impress some businessweek readers but I don't think a really cool technology needs punchlines like that (..apple II..) - your article sounds like its coming from a non-technical observer for a non-technical audience - and that would be okay if the product had passed the stage of hype into early stages of common usage.
At this stage though I am more interested in how is it different from 100s of other AI applications that failed to take over the world or become the next big thing.
My need to criticize primarily came from your last paragraph where you are appealing to developers to jump in without realizing that that segment of your audience looks not very respectfully at phrases like 'buckle up' and 'amazing ride'
It's interesting that the Printing Press didn't make the list of previous revolutions, because I think this highlights two different ways of thinking about intelligence (and thus AI). We can either build AI systems that give us a correct answer 95% of the time and we don't care how they got it, or we can build systems that have some sort of designed thinking (ie we understand our own thoughts and try and replicate their process to some degree).
I think at some point the former became dominant, because its more achieveable, but its also the one that gets most interesting when it fails. Because of the way this AI learns, when its wrong it can be brutally wrong.
I think there will be a revolution in AI getting into more spaces, but only in areas where its ok to screw up (ie ultimately a human is filtering the AI response at some degree, whether its the user knowing when to ignore the GPS or Siri, or an overseer looking over the AI output and checking it before it goes out). I'd be really surprised to see AI doing anything important by itself.
Also, I thought Google already had a fairly solid AI research effort well underway.
Hi, Author here. I think you make an excellent point. I think the revolution will start from small things where it is OK to make an error. Stuff like Siri does. Then it will move on to productivity tools or the like...where it is still OK to make a mistake. And A.I. will improve really fast, like the PC has improved. I think in 5-10 years we won't believe what we have accomplished. I think Siri crosses an important threshold.
Didn't know about Google. They might be working on something but I don't know of any production use of their A.I. software.
How do you judge a new medicine? do you ask if does the correct things 100% of the time? or do you just want it to be better than the other drugs ?
I think the same logic applies to other technologies: products generally have to pass the standard safety tests/level common in their industry, not be perfect.
Will this too happen in AI ? maybe , but maybe due to political pressure , the standards require from AI will be much higher.
Well medicine's come with error bars, typically in the form of a doctor who determines whether the promised benefit plus risk is worthwhile for you. Often this is in conjunction with a consultation with the patient. So the general point that we don't give control to other entities that can make mistakes, I thnk still holds.
But you're right in the general sense: there are other areas where we leave the interpretation of the AI (as it were) to experts and have no say in it ourselves. Road bridges would be the simplest example I can think of (given that the safety simulations are akin to a simple form of AI).
No matter how innovative company Apple is, I wouldn't give them credit just because they bought Siri and integrated it into iOS. I somehow find it odd that no one was talking about Siri before Apple's blessing, although it has existed for many years. It has been constantly evolving, and would have without Apple. There's lots of information about Siri available and the vision and goal of the project has been clear all along.
Taking an impressive but not all that practical technology demo and integrating it into devices that are going to be in millions of people's hands deserves some credit. The major part of the credit goes to those who developed the technology, of course, but identifying the technology as practical for the masses and getting it out to them is important too.
I'd like to credit all those researchers in the original CALO project, and all the hardworking AI scientist all over the world, who have made the evolution of AI to a concrete product possible. If we begin to talk about artificial intelligence revolution, Apple has nothing to do with it in my opinion.
And we mustn't forget Maxwell, Ada Lovelace, Euclid, Plato, Roger Bacon, and the medieval monks. It's always important to trace back credit for any product from Apple to someone who isn't Apple.
Can anyone point me to write ups of Siri's implementation and history? I know Siri was previously a product on the iPhone, so I'm wondering what company created the the technology, and if they perhaps had a more open attitude in discussing their technology than we might hear from Apple and whether they might have some blog posts to look through, etc.
Is Siri really more than a little bit of sugarcoating on top of stuff others have been doing for years (Google, Wolfram Alpha)? I suspect it is doing ok for some tasks (calendar), but that is about it. For example on Android people have been calling contacts by saying their names for a quite a while already.
Android user here who frequently uses the voice commands. The primary differences I see between what we have and what Siri offers are:
- More built-in app integration (like GPS). "Remind me to call Jennifer when I leave work" wouldn't compute. The Wolfram Alpha integration is also a nice touch.
- Improved contextual awareness. "Is it going to rain today?" or "I'm in the mood for Italian" don't work for me. I have to explicitly use keywords like asking for a "map of italian restaurants".
- 2-way conversations. I only speak a command and the voice command opens another application based on that command. Siri can talk back to you and resolve conflicts, such as calendar schedule overlaps.
Have we had voice commands on our phones for what seems like forever now? Yep. But that doesn't make me any less jealous of the neat innovations that Apple's done in the speech recognition AI field. Again, a great example of the fierce competition driving innovation, and the consumer wins.
Still, I think those are only some special cases (weather, calendar). Admittedly they are nice, but I doubt they are an AI revolution.
Also I suppose it is kind of a "desktop search" for the mobile phone. Google can not search your contacts and calendar (it could, but the Google search is not really tied into it yet, I suppose).
It is. What's impressive about Siri is not the speech recognition (although it's pretty good) but rather two new elements that I haven't seen done as well on another consumer device:
1. Interpreting the meaning of your request
2. Remembering and using context
Siri has been a standout in those areas since before the Apple acquisition. Most people are surprised at just how naturally you can talk to your phone. It's doing more complex work than just passing the request on to Wolfram Alpha.
However, Google has a lot of impressive work in this space, too. I don't think it's long before android phones are this good--they are practically there already.
It's classic Apple: they didn't invent it, but they look like they'll get credit for being the first to popularize it and bring it to the mass market.
Siri has been in development for quite some time. Prior to spinning out, it was a DARPA funded research project at SRI. Give the guys credit. They didn't just slap down a bunch of glue to tie together Google and voice recognition.
I've never been able to get squat out of Wolframe Alpha. Even on things that seem like it should be right up its alley ("compare health care costs in Texas to Wisconsin" -- "no data available" -- Oh really?)
So if I said to my phone "please book me a flight to Kansas tomorrow" and it did you would point out that Travelocity does this and this is merely an improved method of integration into the phone?
The only way we're going to achieve AI in the short term is by loosening the definition of the phrase, for example to include a new generation of complex automatons. I look forward to AI meaning about as much as "cloud" does today.
It seems like the author of this article has never in their life seen a computer do anything more than push numbers, but machine learning and AI have been around for at least a decade. This is just a very mainstream case of very well developed AI.
Saying that 'this is the start of a revolution' is not only ill-advised, but totally baseless. What's the innovation? What has changed in what was previously possible? For a revolution to occur, there needs to be an innovation that causes expansion of technological capabilities. Although Siri is (supposedly) great, it does not represent an expansion of technological capability. Just because the team behind Siri spent years building out a great voice activated AI system does not mean that we all can now do it. Computers have not become smarter... they are the same as they ever were. People have not realized a more efficient way to create AI... it's the same as before.
I'm truly surprised that such a shallow and misguided article made it to the popular page in a community like this.