Hacker News Reader

Comment by deepnotderp | Hacker News Reader

This article mentions that AI as a service (aka machine learning, OK, let's be honest, Deep Learning) will be a major pillar of their business in the future. However, the chances that they will be able to beat Google seems vanishingly small.

EDIT: to justify why, let me make these points: Google has more machine learning relevant data.

It has a strong brand image as "THE company for AI "

It has (at this point in time at least) better machine learning hardware. This will change in the near future, but Google is far more likely to take a competitor in this area seriously and acquire them.

Google has not one, but two (Deep mind and brain) world class research labs.

Google's ML APIs are objectively better.

View on HN

> However, the chances that they will be able to beat Google seems vanishingly small.

I'm not sure about that, for one main reason: Google's acquisitiveness regarding data is even more notorious than its habit of losing interest in products and dropping them suddenly. Both are major turnoffs for business customers, to put it mildly.

Amazon, while it does have a habit of vigorously competing with its own customers (in the retail "marketplace" space, and in others - I find the whole Amazon vs. Netflix dynamic absolutely fascinating) is at least reliable in provision of services and transparent about the exchange of cash for value.

In other words, Amazon is, at heart, a ruthless retailer, where Google is at heart an intelligence agency with occult motives. Google usually has the better tech, but Amazon's the company people are comfortable doing business with.

Edit: I should also say, this listicle of an article is mostly bunk. Lab126 is not a great hardware design shop. The Echo/Alexa "ecosystem" is mostly a bunch of (to be fair, well-considered) first-party integration deals to support popular home automation products - "skills" are a joke for discoverability and usability reasons, even if there were a market for them, which there isn't.

I do agree that logistics (and operations more generally) is an enormous strength. Probably the key one, because Amazon doesn't usually come with the firstest or the mostest - it just executes relentlessly at complex, labor intensive tasks. It is at its strongest in places where it is not primarily competing on tech or features but on consistent execution and the customer reputation that comes from that.

> where Google is at heart an intelligence agency with occult motives

That seems off to me. Google is, at heart, an advertiser. This leads to Amazon being more customer centric because most Google customers are actually the product Google is selling and most Google "products" are just the lures they use to attract their actual product. Google never loses interest in their products, but they frequently choose to switch their method of attracting the eyeballs they sell. No cloak and dagger explanations needed.

Google BECAME an advertiser. Some of us remember the pre-advertising, and later pre-Doubleclick-acquisition Google.

I'm not implying Google's motives are "cloak and dagger," just that they are not transparent beyond "index all the world's knowledge" (with all that implies.) This is the mission it started with, which isn't in and of itself a complete business model. It's the model of an intelligence agency, and that's what I meant by my statement. Right now, advertising brings in the money, but it'd be a breathtaking failure of imagination and intellect for that to be the "best and highest use" of Google's corpus.

An optimist would say "advertising subsidized google's mission which is essentially one of public service." A pessimist would say "who knows WHAT google will do (or may already be doing) with what it knows - but it's probably evil." I'm neither, valuing the services Google provides while distrusting the power it has based on its data and reach.

Ads is the business.

You need the tech to build it. And Gooogle is not an AI company. Google is definitely more of a computing company than anything else.

DeepMind will not reach its current stage without Google's computing capability, and Google's computing capability is literally no where else to be found.

Disclaimer: I work at Google.

What makes you say that? Google is 'better' at AI? OK, even if you believe that to be true, like every other market AI will be decided by lots of factors other than who has the 'best' tech. Think VHS vs Beta or Mac vs PC.

They include:

  1. Cost.
  2. Platform/Ecosystem development.
  3. Regulatory compliance (lots of rules if you want to handle health/gov data).
  4. Sales/Marketing (in this case, probably, marketing to enterprises).

I would argue that Amazon is a clear leader in 2-4, and has the history as a low-margin retailer to be competitive in 1.

Google had the best data centers, yet Amazon won the data-center-as-a-service race.

What do you mean by best? In a car race, why would you want to build the best car? I just want my car to cross the finishing line first.

Besides, there is no race. This is continuous evolution. The race has barely started, let alone ended

If you're talking about the "best" datacenters...

Google hit upon the strategy of scale-out-commodity hardware a few years before Amazon did. Also, Google talks more about their infrastructure than Amazon does, so people often think it's "better" from the outside. Whether it is actually better or not is probably an unknowable question from the outside and even if you had all the data it might depend on how you trade off speed, power efficiency, cost etc.

Google is very secretive about its DCs... They appear open, just because they are much more advanced than you can imagine...

My impression is that GCE is relatively young and growing very fast. To say Amazon won seems premature to say the least. This is an active battle.

You could have made the same argument against AWS in the first place, how could a retailer possibly succeed against the likes of Google? Technical excellency isn't the only factor deciding success.

Yes, obsolescence (all aspects) has always been the bane of any leader and entrenched entity.

There are many factors that are in favor of AMZN and GOOG and the likes. Factors that are for you can also work against you.

Generational shift of competency and drive. People get older, a bit rounder around the waist, wealthier and are also getting more choices. All companies have to deal with this factor. And some will do that better than other.

You change the world and the world will change and change you. Are you able to adapt and account for all changes?

Why?

I imagine that data for ML is sort of a logistic curve of benefit, and that Amazon's assorted sources have them "on the plateau", even if not as high as Google.

There's good reason to expect that we'll see several major players in ML/AI, just like we do in pretty much every other market.

Why? Isn't the thing about AI that everyone in the field wants to publish? Doesn't that mean all of their competitors will benefit from their research?

It seems to me that Amazon will likely win because they're better at serving customers and willing to engage in price wars.

> However, the chances that they will be able to beat Google seems vanishingly small.

The respective track records of the two companies suggests otherwise.

Google doesn't know how to market and sell when there's competition. Not much sales expertise is required to make people buy into AdWords.

I've been in the room when they try to sell GCE. The sales pitch is roughly "you are not smart enough to figure this out on your own, so you need to buy this". That may be the case, but it's not the best pitch.

For the one where I was in the room, Amazon won out, even though they were clearly more expensive. They knew what buttons to push, and what buttons not to. And they knew how to at least fake a bit of humility when it mattered.

And how do you know that? The field is huge and mind boggling. I would think, considering the breadth of AMZN, they stand a better chance of being able to apply AI to for a wider impact

Amazon probably has more shopping data. More importantly, more credit card fraud data to train an AI fraud detection system.

They don't need to be doing cutting edge research to commoditize machine learning as a product.

I couldn't disagree more. The entirety of the internet runs on AWS. The rest is an inconsequential rounding error.

> Google has more machine learning relevant data

Subjective. Relevant to what?

______________

> It has a strong brand image as "THE company for AI "

AI is so early there's isn't a recognized brand yet. Maybe to the HN crowd, but that's astronomically small in the grand scheme of things.

______________

> It has (at this point in time at least) better machine learning hardware. This will change in the near future, but Google is far more likely to take a competitor in this area seriously and acquire them.

Competing on commodities? Doesn't seem likely.

______________

> Google has not one, but two (Deep mind and brain) world class research labs.

Again. Doesn't seem like much of an upper hand.

______________

> Google's ML APIs are objectively better.

AWS doesn't need better APIs or Developer UX. That's not what this is about.

> The entirety of the internet runs on AWS. The rest is an inconsequential rounding error.

None of the top 5 Alexa websites [1] (Google, YouTube, Facebook, Baidu and Wikipedia) run on AWS as far as I know. I'm not sure what rounding mode you're using.

[1] https://en.wikipedia.org/wiki/List_of_most_popular_websites

Ha. DL hardware != a commodity

Pretty sure having two of the premier DL labs is an upper hand in deep learning...

Well then what on earth does Amazon offer that's different from Google?

Reply on news.ycombinator.com