Hacker Newsnew | past | comments | ask | show | jobs | submitlogin

AWS specifically have really dropped the ball on this.

I interact regularly with AWS to support our needs in MLOps and to some extent GenAI. 3 of the experts we talked to have all left for competitors in the last year.

re:Invent London this year presented nothing new of note on the GenAI front. The year before was full of promise on Bedrock.

Outside of AWS, I still can’t fathom how they haven’t integrated an AI assistant into Alexa yet either





There's Alexa+ [0] which uses generative AI but it's planned to be a paid option at $20/mo.

[0]: https://www.aboutamazon.com/news/devices/new-alexa-generativ...


> Alexa+ costs $19.99 per month, but all Amazon Prime members will get it for free.

I'm curious if non prime members make up a big market for Alexa. I rarely use my smart devices for anything beyond lights, music, and occasional Q&A, and certainly can't see myself paying 20$/month for it.


I'm curious why anyone would pay $19.99/month for Alexa+ rather than just buy a Prime membership (which is $14.99/month).

Unless of course this is going to be met with a price hike for Prime...


That's what happened with Prime TV, and I absolutely expect it for the AI, too. And it might finally mean I cancel my Prime membership.

Amazon Prime’s price hikes have a predictable cadence: * 2014: $79 to $99

* 2018: $99 to $119

* 2022: $119 to $139

We should expect a price hike from $139 to $159 in 2026, assuming the trend continues.


Meanwhile, Google Fiber has been the same price for 15 years. At least according to the billboard outside my window.

It works out for them because bandwidth gets cheaper over time but inflation eats away at that. $70 today is like $50 back in 2010 when GFiber first launched.

If ChatGPT's Advanced Voice Mode could be served through an always-on device like Alexa, I'd pay for it.

Hmmm... maybe I can install do this through a cheap tablet....


Model sizes have come down enough that it will be possible to run smart home control and simple Q&A entirely locally.

you can do this with home assistant already

Alexa consistently fails with the simplest of questions.

Only thing it can do is set a timer, turn off a light and play music.

It is still nice, but it’s so frustrating when a question pops into my mind, and I accidentally ask Alexa just to get reminded yet again how useless it is for everything but the most basic tasks.

And no, I won’t pay 240 dollars a year so that I can get a proper response to my random questions that I realistically have only about once a week.


> Only thing it can do is set a timer, turn off a light

And it can't even do that without an Internet connection. As someone who experiences annoyingly frequent outages, it never ceases to boggle my mind that I have a $200 computer, with an 8" monitor and everything, that can't even understand "set a timer for 10 minutes" on its own.


Why did you guys bought this in the first place?

Originally bought it for an elderly parent in assisted living; wasn't as useful as we'd hoped; repurposed it as a kitchen-timer/music-box, for which it works adequately as long as the Internet connection is up. I would not recommend anyone else buy one.

Fair question, playing music and timers are nice, but their AI is abysmal so you can imo not ask it to do anything else. I previously worked for a smart home company so I wanted to test out product’s integration with Alexa, so we have some at home. I’m planning to get rid of them, though, and only leave one in the kitchen.

and despite all this, I would pay 240$ a year so that Siri can reliably do what Alexa does today

oh the irony


Alexa has pretty much zero value for me.

Being able to just order something with zero shipping has a ton of value. I could drive down the street but it would still be an hour at the end of the day.

Video streaming has some value but there are a lot of options.


Just pay 0 USD and use Grok app for free?

By far the best thing currently available.


I'm predicting that Grok fails simply due to half (?) the software engineering populating not wanting to use anything Musk has developed.

Grok has to be more than n-times (2x?) as good as anything else on the market to attain any sort of lead. Falling short of that, people will simply choose alternatives out of brand preference.

This might be the first case of a company having difficulty selling its product, even if it's a superior product, due to its leader being disliked. I'm not aware of any other instances of this.

Maybe if Musk switches to selling B2B and to the US government...

If you piss off half of your possible user base, adoption becomes incredibly difficult. This is why tech and business leaders should stay out of politics.


> I'm predicting that Grok fails simply due to half (?) the software engineering populating not wanting to use anything Musk has developed.

I think that's a wildly optimistic figure on your part.

Lets assume that developers are split almost 50/50 on politics.

Of that 50% that follows the politics you approve off, lets err on the side of your argument and assume that 50% of those actually care enough to change their purchases because of it.

Of the 25% we have left, lets once again err on the side of your argument and assume 50% care enough about the politics to disregard any technology superiority in favour of sticking to their political leanings.

Of the 12.5% left, how many do you think are going to say "well, let me get beaten by my competitors because I am taking a stand!", especially when the "beaten" means a comparative drop in income?

After all, after nazi-salute, mecha-hitler, etc blew up, by just how much did the demand for Teslas fall?

The fraction of the population that cares enough about these (on both sides) things are, thankfully, single-digit percentages. Maybe even less.


>>After all, after nazi-salute, mecha-hitler, etc blew up, by just how much did the demand for Teslas fall?

I had been saving up for a Tesla but now I am looking elsewhere. I think a lot of people are doing the same here in Canada. You can grok the actual numbers if you want.


Yeah, a simple example is to just look at how many companies/universities have ChatGPT vs Grok subscriptions internally. I can imagine that many people would have a problem with subscribing to Grok, even if its performance is comparable.

Hmmm, thinking aloud, Oracle?

> This is why tech and business leaders should stay out of politics.

Yeah but they don't stay out of politics, do they? Gemini painting black Nazis was a deliberate choice to troll the vast majority of the population who isn't woke extremists.

My family subscribes to Grok and it's because of politics, not in spite of it. The answer gap isn't large today but I support Musk's goal of building a truth seeking AI, and he is right about a lot of things in politics too. Grok might well fail financially, the current AI market is too competitive and the world probably doesn't need so many LLM companies. But it's good someone wants AI to say what's true and not merely what's popular in its training set.


[flagged]


I think their point was that becoming very involved in politics in a way that alienates half of the population has tarnished Musk’s brand (although, I’d personally adjust that down to more like 1/3). If the point of your whataboutism is that previously it alienated the other 1/3… that doesn’t seem to improve their odds, right?

If anything they’ve now pissed off 2/3 of the population at some point or another.


None of that actually happened.

Literally documented proof.

Literally not. Elon Musk even published the infamous algorithm which he had claimed silenced and censored right wing voices. The only thing the algorithm did that was odd was that it had a special case written into it to boost Elon Musk. You can go look it up.

Mechahitler, the South African genocide debacle, explicitly checking Elons Twitter feed, “You get your news from infowars” system prompts, etc have basically made Grok not a real option for me. I do not want to use a product that is specifically being engineered to be a right wing disinformation machine.

And no, generic brand safety mishaps are not the same; everyone is not doing this.


I enabled Alexa+ few days ago on my devices. Everyone in our home immediately disliked the new Alexa. There were some fairly basic things that Alexa+ cannot do, and Alexa was able to do. Some fairly simple question/answering tasks, and questoins about status of an order.

Yes have seen about that. It’s crazy to me that they still haven’t released it. Really think it could save a dying product

It will be free if you have a Prime subscription (which means nobody will ever pay for it given Prime is cheaper and you get much more included).

But the project is pretty much dead, it was supposed to launch in February or March and is still not anywhere close to being out.


They basically have with Alexa+. It's slightly more limited than ChatGPT, but it sounds much more realistic than stock Alexa and blows it out of the water in terms of smarts. The old model was basically a Siri-like "set timers and check the weather with specific commands," plus some hit-or-miss skills you had to install separately. But the new one gives much more of a sense of understanding your question and can carry on conversations with contextual responses. I've been pretty impressed with it, and the nature of the Echo device makes it much easier to query at will than having to open the ChatGPT app and switch to voice mode.

I agree. I think the Echo devices are good for certain kinds of voice-driven LLM experience. Although it's not that useful for detailed responses and serious questions, since you can't go back and read its response again.

Having briefly interacted with AWS Q out of curiosity, I can see why they haven’t pushed much out publicly. Aside from giving someone a chuckle when they decided to call its suggestions “Q Tips”, it’s functionally useless.

They all but abandoned Astro, their home robot. My suspicion (and information I've heard internally) all but points at them only using Astro as a testbed for self-navigating warehouse robotics, and now that they got what they wanted out of it, the Vesta team basically got thrown to the wolves.

They bought a company that did warehouse robots called kiva before that

https://en.m.wikipedia.org/wiki/Amazon_Robotics


Lingo question: is MLOps like devops for ML, or like flops for ML? I wonder because… actually, either case seems like somewhere Amazon might be losing experts to hot startups.

As the other response said. It’s DevOps for ML. They have Amazon SageMaker which is the managed ML/MLOps offering that we use extensively because we’re a small team. The documentation is awful

All AWS documentation is awful.

The former. Basically: build, train, test, deploy, monitor, repeat for ML algos.

Isn't "Alexa+" doing this? (I have not signed up)

Didnt they basically can most of Alexa a few years ago? I think they realized asking a device questions doesn’t generate profit.

They thought Alexa will enable users to buy more from Amazon just by voice. But most users turned out like me. I would not spend a single dollar on Amazon without actually seeing the item on my mobile or desktop. I wouldn’t even add to cart via Alexa. That’s not an ideal user for device and service that requires hundreds of millions to run.

You saw this with Amazon Dash buttons too. This idea that users would just go "Hey order me some more Tide" and Amazon would just do the right thing at the right price like some sort of intelligent personal assistant. Which it by no means is.

They still have no serverless inference.

SageMaker have serverless inference endpoints

Only if the pipe is defined if you bring your own pipe sage maker offers nothing.

Bedrock is not usable for that?

Not for custom models.



Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: