By Brandon Engel: Important touchstones in technological history come about when something jumps the gap from experimental testing to widespread consumer use. Nowhere else today is this more apparent than within the market for “digital personal assistants”, voice-user interfaces which facilitate an understanding between us and our machines.
Within the broader field of artificial intelligence, machine learning algorithms are rapidly improving to better extract knowledge from our natural, spoken language patterns. The profit motives of large corporations like Apple and Amazon indicate the companies behind these increasingly-powerful softwares are attempting not only to perfect human-to-machine communication, but to also to insinuate themselves and their services in a position that will expand their authority. That has led to something of an arm’s race, as firms jostle for control of the most accurate means of determining consumer’s spoken – and unspoken – desires.
We’ve come a long way from the days when Microsoft Word users wanted to slap Clippy, the first widely known “PDA”, for providing inaccurate editing assistance. Users of Apple’s iOS-based products have now been happily making friends with the company’s Siri system for some time. Amazon, however, is perhaps the best positioned for dominance in the speech automation space. The Echo home speaker is grounded in its Alexa AI service. Even Microsoft has made huge strides with Cortana, another hands-free helper at the intersection of computer science, linguistics, and human psychology.
A New Frontier
These intelligent virtual assistants signal a new frontier in computing. They go much further than just recognizing your requests and attempting to fulfill them: gathering information from the Cloud, their “minds” are informed by millions of pieces of data input by users in real-time all around the world. For the biggest internet players – Apple, Google – access to a massive pool of voice data has help software coders built systems that act as flexible roadmaps for how certain questions should be answered. When consumers speak a search query aloud or ask a question, the software that supports Alexa and her cohorts is able to act efficiently on a wide range of natural language tasks.
A.I. That Listens and Learns
At their worst, these products can seem a bit creepy. At their best, they offer a highly customized experience. Amazon’s push into consumer electronics is not easily separated from its interest in many smaller Internet-connected devices and its broader role as one of the world’s biggest retailers. Now that the day has come when customers no longer have to visit a screen to shop online, voice assistants are encouraging purchasing decisions and facilitating new modes of customer interaction.
According to this website, smartphone users make the most use of voice assistants in their home – at 43%. Using voice assistants while driving ranked second at 36%. Assaf Ronen, Amazon’s vice president of voice shopping, recently applauded the company’s voice assistant skills saying, “Our Alexa speech science engineers have made something so incredibly challenging appear effortless to customers.” On “Prime Day”, Amazon’s yearly discount sale, users ordering products via voice assistant were offered special deals – resulting in the company’s most profitable shopping day ever.
The Invisible Hand
The privacy concerns attached to such systems seem so obvious as to hardly be worth mentioning. Carrying around a smartphone means carrying around a GPS-connected device that ties you to the “Cloud”, where all information is accessible should you know how to look. The fact that we’re all online, all the time has not been lost on advertisers and sellers, and neither by crooks, hackers, and the identity thieves.
Like many ethical concerns in the consumer world, the final verdict is likely to be rendered by the invisible hand of the market. Will end users prize the newfound intelligence of their digital personal assistants? Will they value them enough to trade a sense of privacy for the promise of a better experience while shopping, reading the news or just searching the web? In the meantime, the major software companies are more focused on figuring out the how rather than dealing with the ethical and legal implications. How it all shakes out remains to be seen, but it will be exciting to watch as A.I. grows as an everyday presence in our lives.
Discover more from File 770
Subscribe to get the latest posts to your email.
AAAAAAUUUUUUUGGGGHHHHHHH!
Mike, you need to put a “Mr. Clippy” trigger warning in the title of this post! 😯
“I see you are trying for Fifth, can I suggest you wait for another three posts?”
Cadbury Moose: “I see you are trying for Fifth, can I suggest you wait for another three posts?”
JJ: (opens drawer, furtively pulls out hammer)
I’ll address the general concept by asking this question: can we please have a study to determine how many person-hours are wasted by correcting auto-correct?
“Siri, enslave the human race.”
Not mentioned in the article is the fact that the speech recognition doesn’t happen “on device” but is recorded and sent to a server farm somewhere with more CPU power to throw at it–so every little thing you say that is picked up by a voice assistant like Siri is shipped off to “the cloud.”
Big Clippy is watching you…
Darren says Not mentioned in the article is the fact that the speech recognition doesn’t happen “on device” but is recorded and sent to a server farm somewhere with more CPU power to throw at it–so every little thing you say that is picked up by a voice assistant like Siri is shipped off to “the cloud.”
Apple keeps the data but strips all the metadata away, so the individual user cannot be recognised.so Apple knows how many users asked who the host of File770 was but not who asked that question. This aggregated data makes for more accurate answers.
Modern machine-learning demands mountains of data for even small increments in accuracy. It’s a truism that the team that gets the best results is usually the one with the most data, not the one with the most algorithms. But it has to be labeled data–raw data is almost worthless. That means it needs at least a clue as to whether the result generated was actually what the user wanted. (At Microsoft, we used to pay people in India to examine inputs and outputs and then guess whether the customer was satisfied.)
Anyway, future progress pretty much depends on companies amassing huge quantities of user inputs like this. Barring a major breakthrough of some sort.
@Tasha:
I’d like to nominate this for a Bram Stoker award for short fiction.
Also, double fifth.
What I find unfathomable is, given all this advancement in machine learning, why does Google’s web search keep getting worse and worse? Sure, if you’re vague, inarticulate, confused and uninformed, the mighty Google might help you find what you think you’re looking for. But if you have a clear idea of what you need, and it’s really specific, Google suddenly turns into Valley of the Linkspammers.
Probably it’s a side effect of monetizing the search results, but only Google can tell us, and they’re not talking. :/
@GiantPanda
*bows*
Donald Kingsbury’s 2001 novel, “Psychohistorical Crisis” predicted the development of smartphones to the point where the resulting “Familiar” or “fam” was literally indispensable. Instead of jailing a malefactor (or femalefactor) they just took away his/her fam. Without it they literally couldn’t go to the corner grocery store…
asfi235: A question I’ve often asked. Google needs a not shopping option on its search bar.
bookworm 1398: Google needs a not shopping option on its search bar.
Try adding
-“price”
(make sure that there is not a space between the minus sign and the search term)
to your search terms list. I’ve found that helpful.
I am…85% certain…that I once saw Clippy flip me off, but it was a long evening at work on the streetlight outage hotline, and I consider myself an unreliable narrator.
bookworm 1398, JJ: Getting “shopping” links isn’t the problem. If the only way to get ahold of the document I’m seeking is to buy it, the eBay results are useful.
The problem is a more general one. If I’m seeking the specs on [x] made by [y], google will go all helpy and tell me about user manuals for [p] made by [q]. And yet going “verbatin” overcompensates, as it then misses results I know are out there. Bleh.
Pingback: AMAZING NEWS OF FANDOM 9-11-2016 - Amazing Stories
Pingback: Top 10 Posts For September 2016 | File 770