When I saw Google Home for the first time back at Google I/O, I was excited at the prospect of having a brainier Alexa in my home. Like others, I waited and almost forgot all about it until it was reintroduced last month when I actually could go and pre-order it.
I got my Google Home at the end of last week and placed it in the same room Alexa has been calling home for almost a year now. The experience has been interesting, mainly because of the high expectations I had.
Making comparisons with Echo is natural. There are things that are somewhat unfair to compare because of the time the two devices have been on the market and therefore the different opportunity to have apps and devices that connect to them. There are others, though, that have to do with how the devices were designed and built. I do not want to do a full comparison as there are many reviews out there that have done a good job of that but I do want to highlight some things that, in my view, point to the different perspective Amazon and Google are coming from when it comes to digital assistants.
Too early to trust that “it just works”
Like Echo, Google Home has lights that show you when it is listening. Sadly, though, it is difficult to see those lights if you are not close to the device as they sit on top rather than on the side like the blue Echo lights that run in circles while you are talking to Alexa. This, and the lack of sound feedback, make you wonder if Google Home has heard you or not. You can correct that by turning on the accessibility feature in the settings which allow for a chime to alert you Google Home is engaged.
It is interesting to me that, while Amazon thought the feedback actually enhanced the experience of my exchange with Alexa, Google did not think it was necessary and, furthermore, something that had to do with accessibility vs. an uneasiness in just trusting I will be heard. This is especially puzzling given Echo has seven microphones that clearly help with picking up my voice from across the room far better than Google Home.
The blue lights on the Echo have helped me train my voice over time so I do not scream at Alexa but speak clearly enough for her to hear even over music or the TV. This indirect training has helped, not just with efficiency, but it has also made our exchanges more natural.
OK Google just doesn’t help bonding
I’ve discussed before whether there is an advantage in humanizing a digital assistant. After a few days with Google Home, my answer is a clear yes. My daughters and I are not a fan of the OK Google command but, more importantly, I think there is a disconnect between what comes across like a bubbly personality and a corporate name. Google Assistant – I am talking about the genie in the bottle as opposed to the bottle itself – comes across as a little more fun than Alexa from the way it sings Happy Birthday to the games it can play with you. Yet, it seems like it wants to keep its distance which does not help in building a relationship and, ultimately, could impact our trust. I realize I am talking about an object that reminds you of an air freshener but this bond is the key to success. Alexa has become part of the family from being our Pandora DJ in the morning to our trusted time keeper for homework to my daughter’s reading companion. And the bond was instant. Alexa was a ‘she’ five minutes out of the box. While Google Assistant performs most of the same roles, it feels more like hired help than a family member.
Google Assistant is not as smart as I hoped
The big selling point of Google Home has been, right from the get go, how all the goodness of Google search will help Google Assistant be smarter. This, coupled with what Google knows about me through my Gmail, Google docs, search history, Google Maps, etc., would all help deliver a more personalized experience.
Maybe my expectations were too high or maybe I finally understand being great at search might not, by default, make you great at AI. I asked my three assistants this question: “Can I feed cauliflower to my bearded dragon?” Here is what I got:
Alexa: um, I can’t find the answer to the question I heard
Siri: Here is what I found… (displayed the right set of results on my iPhone)
Google Assistant: According to the bearded dragon, dragons can eat green beans…
Just in case you are wondering, it is safe to feed bearded dragons cauliflower but just occasionally!
Clearly, Google Assistant was able to understand my question (I actually asked multiple times to make sure it understood what I had said) but pulled up a search result that was not correct. It gave me information about other vegetables and then told me to go and find more information on the bearded dragon website. The first time I asked who was running for president I received an answer that explained who can run vs who was running. Bottom line, while I appreciate the attempt to answer the questions and I also understand when Google Assistant says, “I do not know how to do that yet but I am learning every day”, the experience is disappointing.
Google is, of course, very good at machine learning as it has shown on several occasions. I could experience that first hand using the translation feature Google Home offers. I asked Google Assistant how to say, “You are the love of my life” in Italian. I got the right answer delivered by what was clearly a different voice with a pretty good Italian accent. Sadly, though, Google Home could not translate from Italian back into English which means my role as a translator for my mom’s next visit will not be fully outsourced.
We all understand today’s assistants are not the real deal but rather, they are a promise of what we will have down the line. Assistant providers should also understand that, with all the things the assistants are helping us with today, there is an old fashioned way to do it which, more likely than not, will be correct. So, when I ask a question I know I can get an answer to by reaching for my phone or a computer or when I want to turn the lights off when I know I can get up and reach for the switch. This is why a non-experience at this stage is better than the wrong experience. In other words, I accept Google Assistant might not yet know how to interpret my question and answer it but I am less tolerant of a wrong answer.
Google Assistant is clearly better at knowing things about me than Alexa and it was not scared to use that knowledge. This, once again, seems to underline a difference in practices between Amazon and Google. When I asked if there was a Starbucks close to me, Google Assistant used my address to deliver the right answer. Alexa gave me the address of a Starbucks in San Jose based on a zip code. Yet, Alexa knows where I live because Amazon knows where I live and my account is linked to my Echo. Why did I have to go into the Alexa app to add my home address?
Greater Expectation
Amazon is doing a great job adding features and keeping users up to speed with what Alexa can do and I expect Google to start growing the number of devices and apps that can feed into Google Home. While the price difference between Google Home and Echo might help those consumers who have been waiting to dip their toes with a smart speaker, I feel consumers who are really eager to experience a smart assistant might want to make the extra investment to have the more complete experience available today.
We are still at the very beginning of this market but Google is running the risk of disappointing more than delighting at the moment. Rightly or wrongly, we do expect more from Google especially when we are already invested in the ecosystem. We assume Google Assistant could add appointments to my calendar, read an email or remind me of upcoming event and, when it does not, we feel let down. The big risk, as assistants are going to be something we will start to engage more with, is consumers might come to question their ecosystem loyalty if they see no return in it.
AI proponents are either misleading themselves or the rest of the world when they make it sound as if solving the ‘context’ problem of machine-to-human conversation and investing computers with Star Trek-like conversational fluency is just around the corner.
When you talk to another person, you are engaged in a constant process of guessing or evaluating **what is in the mind of the person you’re speaking with**. You make your best guess based on more than just the words you are hearing: inflection matters and if available, visual cues like facial expression and body language. Does the person sound worried, enthusiastic, happy, sad, etc. How fluent does the person speak, is there an accent? Etc. Aside from information directly conveyed through the conversation itself, what do you know about the person? Where is the conversation being conducted? Any significant information, news, or experience, that you two might have in common both just recently or further back? What are the person’s motive behind the conversation? Making a workably accurate guess of all that information, and judging the relative significance of all those pieces of information is not an easy computing task. It might not even be attainable for a computer.
It’s like that car insurance commercial (State Farm?) where the young lady is getting her first car and the older gentleman has had his car destroyed. Both using the same words. Each saying something completely opposite in meaning.
Joe
I agree that “Star Trek-like conversational fluency” is a long, long way off, but that’s not what I want a voice assistant for. I want it for boring functional tasks and advanced enough that I don’t have to learn much in the way of how to talk to it. I do stuff like dictating, adding reminders, getting directions, conversions, weather, and some general questions. Siri does need a lot of improvement but I already find it useful, and that’s all I want. I could be wrong, but I don’t think I want a voice assistant that tries to be too intelligent, I just want a good voice UI.
But even your modest requirements are harder than most people think. Humans can sense when a misunderstanding has arisen in the course of a conversation and efficiently make corrections on the fly. I don’t see how a machine that doesn’t have more than an extremely crude and rudimentary ability to guess what’s in your mind can make those quick corrections, much less infer or detect what precisely is the point of misunderstanding or miscomprehension. Even with the ‘simple’ tasks that you want a digital assistant to do, without the capability I described, communication bogs down as soon as you stray beyond the limits of the machine’s knowledge.
I’m not looking for it to be 100 percent perfect, Siri is already useful enough for me, and it should get better. So I have to try a couple times here and there, or rephrase a question, it is still more useful than not having the capability at all. As I said, I’m more interested in a voice UI, and that is already working today.
Well, good for you. Seriously. What you are looking for though is certainly far less than what most people want and what most AI proponents are promising.
I think you’re right on “what most AI proponents are promising” but wrong on “what most people want”.
Mmmmm… a sample of ONE query ? Seriously ?
I asked other questions some of which I got an answer for the same I did from Alexa and some I did not. Straight forward questions like how many inches in a foot or how far is the hearth from the moon were not problem. asking if you need a permit in California to own a wolf did not get an answer yet search on google gives you the proper answer. as you can imagine I could not list all the things i asked and i was making a point about the difference between search and assistants
That’s weird, the Assistant in Allo gives me seemingly correct answers to both questions (which it understands correctly), with more in the sliding suggestions. Maybe the voice-only answer is limiting it.
https://uploads.disquscdn.com/images/c7330ca5b9763734f3281a173831bddd98d666f84e0b96b7559c17384a343aaf.png https://uploads.disquscdn.com/images/b4bd40937ab459adafbe71f1bb09010b0724932b6736d724e56206f7578102f0.png
this is exactly the point I am making search vs assistant. i get the correct results when i search so why not when i ask over voice vs over text. see the point I made in the article about understanding search is not AI
Carolina, As others have pointed out your experience is seriously suspect. I have both an Echo and a Google Home and Google is vastly better at understanding what you want.
Saturday morning my kid comes up stairs and tells me that his game is away and NOT home like he had told me. He mentions some high school I never heard of. I had not yet showered and needed to know quickly how fast I need to get going.
Ask Google Home what does traffic look like to high school. Google comes back 25 minutes with some road closures. Turns out there was a local Fun Run that morning.
I was in a rush and did NOT have time to ask Echo as in kitchen. When got home with some time asked Echo and it did NOT know the high school. Tried a couple of times even trying to word differently but the hangup is it did not know the high school.
This is a real life experience not a made up use case. We received the Google Home on Friday and this was Saturday. So in one day the Google Home helped in a way the Echo was unable.
This personal experience causes me to be very suspect of this article.
Be honest do you work at Amazon?
Mine was a real life experience i was actually preparing food for our bearded dragon and I did not know if cauliflower would be ok.
I am also not surprised at all that on directions Alexa is pretty useless compared to Google.
I certainly don’t work at Amazon, do you work at Google? LOL
Seriously, I am not alone in expecting more from Goole Home just cause it is Google, I am simply pointing out the higher risk that Google faces due to those higher expectations.
Very interesting review. Thank you.
Attempting a joke while not being sure if you asked a serious question is a kind of a big deal,no? Had it knew you had a real bearded dragon (although you said my…) , it might have answered differently.
I attempted the same question with Google assistant on my iPhone and it did give me
a correct answer to a slightly modified question and a set of search results to the exactly same question. https://uploads.disquscdn.com/images/8980db3545a3b1b412b37d88c21336a1f510be691c9fd294d5cb505f16de1f96.jpg https://uploads.disquscdn.com/images/bc608ab80553762c8826f5e41d84a6c05b6901bd50717a2e79cc0943feb209be.jpg
I wonder if Goggle Assistant learned already a correct answer or Google Home uses a different assistant engine?
it is a legitimate question though that someone could ask, no? it was interesting to me that it understood what i meant and pulled something close to the right question but it does not seem to process the info. I had tried to ask the question differently and I got the I cannot help with that question. your search result is similar to what Siri pulled up for me.
I tot a lot in a question depends on an intonation, a context, an accent etc. How Home can know wheter you are serious or not when you ask a question? Try asking Home the same question wit a different intonation? Tat would be an awesome ting to try!
Purchased an echo on launch and family loves it. We received our Google Home (GH) last Friday and love it.
The biggest difference is understanding and doing what you want. So I have a terrible memory for song, artist, etc. At times with the echo I would have to do a quick Google search on my phone using a lyric or similar to get song name so I could tell the Echo to play.
The GH just gets it. We will keep the Echo in the kitchen for ordering and then get the kids a GH for Christmas. I hope Google does a three pack like the are doing for their mesh network stuff.
In the end it all comes down to working on what you most want it able to do and even thought the Google Home just came out the Echo honestly is just not competitive in understanding english at the same level of the Home and everything else really does NOT matter.