Hacker News Reader

Comment by axg11 | Hacker News Reader

Context: I recently left Amazon

Some orgs within the company are fairly lean, some are hugely bloated. Amazon isn't homogeneous. For example, Alexa is very over-resourced and it's common to hear about 4+ teams working on slight variations of the same project. Lots of redundancy there. I believe Alexa as an org is losing $5b+ per year. Contrast that to all the criticism Facebook is receiving for their investment into VR/AR. The investment in Alexa was a platform play. Amazon thought that voice would be the next big platform. I'm not sure results and trajectory justify the investment anymore.

View on HN

Alexa is crazy. Close to 10K employees for a product that they don't even know how to monetize. I understand that it's part of the Amazon flywheel but there's a limit of how much you can rely on AWS to fund you.

Alexa is full of horrible managers that only care about growing their teams so that they can get a promo.

Alexa is funded by retail’s profits.

Retail has no profits. AWS is the profit center.

and advertising

Retail has no profit after all expenses. One of those expenses is Alexa

What needs to be distinguished here is retail as in Amazon procuring and selling goods, or retail as in Amazon being a platform and another version of Aliexpress.

The former is a low single digit profit margin business, the latter probably has a decent profit margin.

Alexa does seem like it has fantastic potential. To throw so many resources at it with so little result seems like an incredibly massive failure of management.

Alexa was Jeff’s pet project. Management failure, indeed.

Failure for whom? The shareholders, sure. But Alexa is pretty cool. It is the first voice interaction that I have used that actually works. So good for Jeff.

I think Siri and Alexa are simply at the end of their roads. To move Siri and Alexa ahead, we'd need a general AI so powerful, it would change the whole world.

I wonder if, within my lifetime, it will be possible to tell a digital assistant to "order all the ingredients I need to bake the cinnamon buns as detailed in the cinnamon bun baking video I watched last month with the Chinese lady that doesn't have an accent"?

Based on the capability of state of the art language models, we may get that soon. The technology largely exists, but the compute is still too expensive. Rumor has it that GTP-4 will be able to pass for human-like.

Right, the missing step to make these assistants actually-useful beyond very simple tasks like "play [album name]" or "set a timer for 5 minutes"—which are basically just a voice-driven command line—has almost nothing to do with voice recognition per se and is more-or-less identical to the problem of "invent AGI".

It'll be useful when you can have it create, say, a menu and shopping list for your week, interact conversationally to tweak it including understanding that something like "I'm allergic to that ingredient" means never ever try to include it again while other statements of preference may be much softer, learn your preferences so those conversations get shorter every week, and order the ingredients for you, too. It barely matters if that happens over keyboard input or voice, because the hard parts of that problem have nothing to do with how the input is done.

even a "voice-driven command line" would still be pretty useful if it actually worked reliably. Google's implementation misinterprets my parents' address about 50% of the time, choosing some town I've never visited in another state as the most likely alternative. I navigate to that address at least once a month, and I even have it as a saved place in maps.

The voice recognition (given decent hardware and a fairly vanilla accent) is quite good. The problem is being able to integrate, present options, learn preferences, etc. even to the degree of that subpar assistant who you only keep paying because they're the daughter of your mother's best friend.

You want to be able to give a few parameters for a trip you're taking and get some choices back. But, as you say, that's far more about "intelligence" than the type of input.

There's a The Smile song called "You'll Never Work in Television Again." Last week I said "play <name of song>" and Alexa responded that it could not find a playback method called "television." I tried to annunciate more clearly and it replied "OK, playing <song name> on repeat," because the song title ended with the word "again."

Nothing reminds you about the difficulty in cleanly parsing human language more than trying to work through a voice interface. I'm not sure what is missing precisely, but I'd say "an ability to use semantic understanding" might be helpful. All of our AI works explicitly with no understanding of semantics, but it's becoming clear to me that the ceiling that places on their "intelligence" puts them still leagues away from the world of true AGI.

If there was also a screen, so many more things could be done. For example it could show that list of ingredients and their prices, for your approval. Then let’s say you could interact with that screen by voice (your fingers are sticky because you’re cooking now) or touch.

There is one in Amazon Echo Show. I've never used one though so I don't know how useful it is in practice. The thing is though that I'm never far from a computer in my house so I'm going to one of those for anything remotely complex.

> Close to 10K employees for a product that they don't even know how to monetize.

Or work consistently.

"Alexa play music".. Nothing... Do it a few times and it finally does play music? I could have gotten up and done it myself easier/faster at times.

the links and offering is solid. I have spotify connected to my amazon account and this is NICE..but as i said, the inconsistency is frustrating.

> don't even know how to monetize

Amazon music subscriptions are $15.99/month. Doesn't that count? Sure, you can buy that without an Alexa, but that is the whole reason people do.

I have an Echo and a Homepod and I rarely use the Echo anymore because of their "by the way, did you know..." responses. They know there are a lot of users who want to disable that but unfortunately their goals aren't aligned with their users' goals.

Everything I use Alexa for:

"Alexa, set a timer for x mins / wake me up at x time."

"Alexa, what's the weather today?"

"Alexa, can dogs eat <type of food>?"

"Alexa, play <the song I just referenced so my wife knows the reference even though she doesn't care>"

Mostly just the first two. I don't care about anything else they try to push on me.

I got rid of my Echo mostly because of those types of responses. And also for how easy they make it for you to buy something accidentally off Amazon, or accidentally upgrade to Music Unlimited.

You can say "Alexa, stop by the way" and it will disable it for a time, but it will come back. If you really want to disable it for good, you can set up a routine to effectively issue this "snooze" on by the way at some interval.

The fact that organizations have multiple teams working on overlapping or even similar things baffles me. I can't imagine how that wouldn't be immensely dissatisfying for the individual.

Internally it was sold as: working together with other teams is too inefficient due to dependencies and differing priorities. It's faster to build it again independently than integrate with the existing service.

In environments where missing deadlines can have huge cascading effects it isn't unheard of to do redundant work so that you have a better shot of meeting your deadline and in a good case you have multiple successful options to pick from. When your deadlines are potentially expensive enough (think capital investment that need to be bought in time to be built out before old ones become uneconomical to operate) this approach can be well worth it.

Not all teams are created equal. Having multiple teams with an extremely similar charter promotes competition, harder work, and deeper search for synergies elsewhere. A single team with a guarded remit is not going to try as hard and they may outright fail on their mission even if they do have internal competition.

It makes a lot of sense if you can afford it.

This approach was made popular by Steve Jobs at Apple. He would have multiple teams competing to who's solution would survive, while the other teams products were thrown away

This goes all the way back to his early days (I think I remember a scene about this in The Pirates of Silicon Valley), but also reportedly multiple teams were building very different original iPhones

I think this approach makes sense for hardware but not pure software.

I think its easier to adjust pure software iteratively than hardware. So with hardware, it makes sense to have different teams working on similar but different things because variations of ideas need to be fleshed (and thoughts that spawn off those ideas) and it can be hard for one team to do this iteratively.

If you're right and this is where it came from, I think largely pure software teams doing this are wasting resources, but if its hardware I don't agree its wasteful

It was way before Jobs, IBM was using it ages ago.

Isn't this part of the premise of the book Hackers? Haven't read it in like 25 years, but that's what I vaguely remember.

I don't recall this coming up in Hackers: Heroes of the Computer Revolution, but it was definitely in The Soul of A New Machine, the story about Data General trying to build their first 32-bit computer.

Thank you... it was Soul of a New Machine. I don't think I even ever read Hackers!

Jobs gets credit for everything in computing the way Churchill gets credit for all the good quotes.

I think it was Mark Twain who said, “Hey, Churchill didn’t say that; that was me.”

Well played, Abraham Lincoln!

You spawn multiple teams working on very similar projects without them knowing about each other, let them grow for a while, then choose the most promising one (or with the highest traction) as your default solution and kill off the rest. Pretty normal exploration strategy.

Amazon has about 75,000 engineers. That's the equivalent of a large town. It's completely reasonable that many people would end up duplicating the work of other teams simply because they never communicate with them.

The philosophy is "1 > 2 > 0". This means ideally you'd prefer a single thing but single things often bottleneck delivery from different orgs as they have differing requirements. In that case you'll accept overlap/duplication if it keeps those orgs decoupled and both delivering faster.

I ran into this a couple years ago with partially duplicating a service that existed under another SVP org. It didn't feel good at the time but in hindsight it was probably less effort and higher chance of success than trying to align priorities across SVP orgs and making their thing work for our slightly different use cases.

Speculative execution .... which ever team moves fast wins ... heck sometimes its not even the team with the best features that wins ... teams with just enough features but who can move fast win sometimes and its bad for the long run. Most big companies have been doing this for a long time now.

Unlike Facebook Alexa was somewhat successful. See Alexa microwaves, cars, TVs, etc.. It looks like Amazon also knows when to move on vs. Meta.

Success is relative. Voice didn't turn out to be "the next big computing platform". It's still mostly a command interface. Perhaps it still could become something bigger. Alexa isn't widely used on the iPhone/iOS/CarPlay or Android/Android Auto, the two most important places for voice. Alexa works well but doesn't have a clear technology advantage vs Google Assistant (admittedly, Siri is trash).

Contrast that to VR, which is in a much earlier stage of its technology and adoption cycle: Meta has a strong technology advantage, currently has the best overall platform and _if_ VR becomes the next big computing platform, Meta is in a strong position to capture a significant proportion of market share. It's a big if! but at least there's still a chance.

Voice still has a lot of potential, especially with the advances in AI that we've been seeing recently. A robust voice-controlled interface would be immensely useful to consumers. I think the investment in Alexa voice is well worth it compared to whatever meta is doing atm.

Voice as an interface sucks for almost everything where choices exist. As a purely serial communications method you are forced to wait for enumerated choices.

As a primary interface voice was very widely used in the 80s and 90s in the form of IVRs and the industry was immediately destroyed by web browsers as fast as was possible. Nothing makes people angrier than having to go through a phone menu system and that’s just an organized version of your more general voice interface.

“Alexa set a timer” defines the kinds of interactions that are sensible to do with voice and it is very, very narrow.

It was never, ever going to be big.

From what I heard, and this was years ago, it is a very resource intensive service to actually run. Given how little I have personally used my Alexa to make or inform actual purchases, I suspect it unfortunately doesnt actually earn much.

> Amazon thought that voice would be the next big platform

It turns out that there are many interaction models where Voice is clearly inferior. Visual discovery (contrast with music) such as for videos, as well as navigation in a UI, are poorly done via voice.

Reply on news.ycombinator.com