The demise of digital voice assistants

Updated: December 12, 2023

I'm not a spiteful person. I hath no pride. But sometimes, I do enjoy a good moment of "I told ya so". After all, when, in the course of your life as a techie at a tech firm, you're occasionally forced to listen to lectures on the future, with words like DevOps and AI/ML thrown into the grinder, by "inspired" managers who just read about the new and cool thing on their Linkedin digest in between important meetings, it's only natural that one would feel vindicated when things out turn out, inevitably, as they should. Case in point, the impractical reality of digital assistants.

In the past year or so, reports and stories have come about un-profitability of digital assistants, and as a result, the culling of teams and technology involved. Once upon a time, Siri and Cortana and Alexa were all the rage, and now, they are sort of not. To me, this outcome was obvious from the start. I just had to wait a few years to be, once again, inevitably, proven right. Let me tell why.

Technology without purpose

Why, why, why, why, why. That's the question one should ask, ad nauseum, whenever someone brings up a new technology or proposes a new change that will "revolutionize" life. First, revolutions are all about massive bloodsheds and tragedy, so no. Second, life optimizes itself as it is, it needs no radical disruptors (except Cardassians). Three, if the person who proposes the change or technology can answer all of your why questions without stuttering, they might be on to something.

Which is not what I got whenever the digital or AI assistant story was brought up. All I heard were almost cult-level "can improve your life", "simplify tasks", "order milk for you", and similar IoT buzzwords, without an actual problem. In fact, the only scenario that I've ever seen is as follows:

A corny 3-min, slightly saturated video showing a busy, high-life marketing person-cum-techie (in the US, of course), with a Bluetooth dongle in their ear, jetting about their busy, high-life life. They are young, single, and cannot cook if their busy, high-life depended on it. They have way too many meetings to go about ordinary life. So they forget to buy milk! Duh! So now, as they are driving home, they "call" their digital assistant, and the assistant takes care of everything. It remotely turns on the AC and orders milk for you, and by the time they're done navigating a 2.5-hour traffic jam (because public transit is for poor people) to their overpriced studio, the milk is there, waiting for them, chillin'.

And ... that's it. This is pretty much the only thing I've ever been shown. Or perhaps slight variations like, alarm clock, music life, trivial nonsense stuff that makes no difference whatsoever other than being a cool and fancy gimmick like some cliche sci-fi movie from the 1960s.

Traffic jam

Credit: Photo by Jacek Dylag on Unsplash.

I've never seen digital assistant proposed for meaningful things, like helping disabled people or the elderly, helping people who are lonely, helping workers in high-intensity environments and/or production line. But I guess that's not as glamorous as someone with a Bluetooth ear thingie.

As a total cynical aside, it's also inherently a cruel reminder that ... you're not rich enough to have a real assistant, a human, who can take care of your everyday stuff. Which is indeed what the real power makers do have. Human assistants, who help with mundane affairs. Take a look at any serious VP, SVP, CXO, or whatever, they will have one or two people in tow. And at home, there might be a cook, a maid, a pool cleaner, who knows.

Designed by nerds for the normies

Now, the second and possibly an even bigger problem is that digital assistants don't speak the language of their users. Plain and simple. Supposedly yes, you can use "natural language" words and phrases to tell your digital box to do this or that, but it only works for basic things. Like milk.

If you want to have your assistant really work for you, you need a programmer's mind, and you need to instruct it with a set of pseudo-code if-then conditions, otherwise it will fail its task. For example, I wanted to buy a laptop. But I didn't have anything definite in my mind, until after I've spent a good few hours reading, browsing, comparing devices, reading human reviews, checking very specific details.

Now, how do you translate that to the machine? Find me a laptop that suits my taste? Nope. You need to say something like: I'd like a 14-inch laptop, with 400nits screen, preferably with 16:10 ratio and resolution of 2-4K, the processor can be either AMD or Intel, mostly likely mid-range, removable battery, and so forth. This is exhausting, especially if you have to rattle it off the top of your head. Or you could write all of your requirements down, which means you don't need the assistant to begin with. You save no time. And you might as well do that yourself, once, without extra homework making the machine do something that you will need to check and verify anyway. Rule hack no.1: there are no shortcuts. You cannot save time. That's bs.

Robot assistant

Your "AI" assistant after hearing your demands. Credits: Photo by Rock'n Roll Monkey on Unsplash.

Unreliable market

Next, assuming you're able to compose a super-complex query, and assuming the machine can reliably understand and interpret your request (which it often cannot, even in various regional English accents, let alone foreignese), then ... then what. Where does the machine go to buy the required hardware?

The big problem is that there's no truly 100% trustworthy marketplace. Wherever you go, online, real life, you will need to check and double-check everything. Even if there's no malice involved, people can make mistakes, mislabel products, you may miss a little detail, or change your mind. There are thousands of variations and little self-paradoxes in our everyday life that they cannot be reliably coded down to handed to a machine and expect any sort of normal results. Humans are inherently unpredictable, especially on the individual level.

There's no chain of trust, no chain of accuracy. Would you blindly let anyone - and I mean anyone, human or machine, do any sort of large-sum purchase for you blindly, without any verification. Chewing gum, sure, but furniture, electronics? No way.

I find it surprising when the executives get surprised their technology didn't yield amazing results. People would use these assistants for trivial things or trivial purchases, and they never really established any sort of predictable, repeatable pattern (the kind of thing sales people love - recurring purchases). On top of that, add the language-culture problems, the vast differences in the supply chain infrastructure readiness in various countries, the maturity of online shops, the available of local delivery, operation costs, and finally, the raw human nature, and you get a model that does not work.

Humans like to shop

For whatever biological reason, humans actually like to invest time in obtaining goods. Be it food, clothes, electronics, cars, or furniture, people like to add their own touch to these objects through minute personal decisions, because that's what differentiates us from machines. If you want to kit out your living room, sure you can buy off-the-shelf items from somewhere like IKEA, but every single home will be different, and tons of time and energy will have gone into adding the personal details to the otherwise generic, soulless stuff.

When you buy food, sure, you will buy the same basics, but one day, you will go for bread A and the other for bread B, and maybe one day you'll try pasta, and who knows. That's the chaotic beauty of human nature, and it cannot be algorithmicized. Even busy people, when they shop online, they will STILL invest time in figuring out what they need. Perhaps they will save time on going to the shop and the actual delivery, but the middle part remains, and it must be human. But that's exactly what the digital assistant promise to do.


Credits: Photo by Atharva Tulsi on Unsplash.

Thus, we circle back to the original solution without a problem really. Digital assistants have been advertised at some soulless corporate borg, who is too busy to enjoy life, and who can be sustained through a Utopian network of predictive purchases with 100% success and guarantee. What happened was, both ordinary people and geeks got to try the goods, realized they have to invest the same amount of time they would ordinarily require to fulfill their needs, just in a slightly different manner, they figured they love their money too much to entrust it to a speaking box, they realized the purchases of expensive items are way more complicated and unpredictable than they seem, the online shops cannot really be blindly trusted, and ... it's all nonsense.

Perhaps the scenario works in a sterile environment of a sci-fi movie, not in the real world with angry people, traffic jams, whimsical decision making, tons of fake products everywhere, and then some. Finally, let's not forget the HUGE privacy implications of telling the voice box pretty much anything about yourself, including very precise daily schedule, purchasing habits, and who knows what else.


I am sort of pleased with the outcome. Not happy that people working on these technologies get to see their ideas and products go down the drain, but hey, blame the buzzword-happy management. Thinking of specific examples, I spent so much time and anger neutering Google Assistant on my different Android phones that I'm happy to see it become sidelined, gone, whatever. It doesn't serve any purpose for me other than to annoy me and waste my time disabling it.

The digital assistants are basically a less efficient you. They can't actually physically do any of the stuff you need, so they're merely an interpreter of your thoughts, which you need to phrase out loud, in a language designed for the machine. Like an external consultant project manager on a project that already has a company-assigned project manager. If anything, it's a pointless ritual where you're the one serving the box, not the other way around. Highly ironic if you think about it. But I guess there's some hope for humanity after all. Even the normies didn't fall for it, so perhaps we ain't as stupid as we seem.

LOL, the joke is on you. There's the new generation of AI assistants. Bye Google Assistant, Cortana, hello ChatGPT, Bard, Copilot. Buzz, buzz, buzz the word goes. If the first round didn't work out, try again until the user succumbs. So now, we have fresh AI stuff everywhere, same same, but different. Will this new experiment with "all natural" AI work out? Unlikely. The modern incarnations, whatever they are called, may create lovely images, spew text like a pro, and do wondrously magical things, but at the end of the day, they still require precise input to be useful, and useful data to be precise. And with humans being the epitome of randomness and self-paradoxes, there ain't no amount of computing power that can solve that. But we shall see. You thought I forgot about these new things, didn't ya? Take care.