HN Reader

But actually every other company has been much more strategic, Microsoft is bullish because they partnered up with OpenAI and it pumps their share price to be bullish, Google is the natural home of a lot of this research.

But actually, Amazon, Apple etc aren't natural homes for this, they don't need to burn money to chase it.

So there we have it, the companies that have a good strategy for this are investing heavily, the others will pick up merges and key technological partners as the market matures, and presumably Zuck will go off and burn $XB on the next fad once AI has cooled down.

Truthfully, I don't think anyone would recommend their acquaintances to join Amazon right now.

That said, Amazon is actually winning the AI war. They're selling shovels (Bedrock) in the gold rush.

Don't need to train the models to make money hosting them.

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

AWS has Bedrock to use various AI providers and has bundled the licensing into the price, so they are getting the users without having to develop the actual AI.

They provide the compute, networking etc, and they provide the users to the AI vendors.

Why would they need to develop their own?

Also Amazon is in another capital intensive business. Retail. Spending billions on dubious AWS moonshots vs just buying more widgets and placing them across the houses of US customers for even faster deliveries does not make sense.

The blessing right now is the limit to contextual memory. Once those limits fall away and all of your previous conversations are made part of the context I suspect the game will change considerably, as will the players.

While they're protected now, https://news.ycombinator.com/item?id=20980557 quotes the one I recall...

      - Nobody has figured out how to make money from AI/ML other than by selling you a pile of compute and storage for your AI/ML misadventures.

AWS enables thousands of other companies to run their business. Amazon has designed their own Graviton ARM CPUS and their own Trainium AI chips. You can access these through AWS for your business.

I think Amazon sees AI being used in AWS as a bigger money generator than designing new AI algorithms.

To me, that's a pretty good explanation.

The world is crazy with AI right now, but when we see how DeepSeek became a major player at a fraction of the cost, and, according to Google researchers, without making theoretical breakthroughs. It looks foolish to be in this race, especially now that we are seeing diminishing returns. Waiting until things settle, learning from others attempts and designing your system not for top performance but for efficiency and profit seems like a sane strategy.

And it is not like Amazon is out of the AI game, they have what really matters: GPUs. This is a gold rush, and as the saying goes, they are more interested in selling pickaxes that finding gold.

"Search is broken. If I search for wwvb watch, I get shown tons of watches which are definitely NOT WWVB."

"What browser are you using? Could you try Chrome?"

I had an Amazon interview loop on the calendar during my recent job search, a couple of months back, but it was difficult to get excited; they think so very highly of themselves, for what they're offering - and I don't just mean the money, but the culture too. They treat you like an interchangeable wage slave, not like a respected professional; it's all hoops to jump through, and procedures to memorize - dance, monkey, dance!

The recruiter was shocked when I cancelled the rest of the interviews, like, aren't you even going to give us a chance? But no: I had received a good offer from an ambitious, well-organized, well-funded AI startup which was excited to have me on board. With that on the table, why would I put up with Amazon? They won't offer better pay, they can't offer a better culture, and they don't have more interesting problems to work on.

Of course, the AI talent war may end up being an expensive and misguided strategy, stoked by hype and investor over-exuberance.

And I say, good. We need new, smaller companies with different cultures in this space. We don't want these giant corporations to dominate and control everything.

1) High-quality training data is effectively exhausted. The next 10× scale model would need 10× more tokens than exist.

2) The Chinchilla rule. Hardware gets 2× cheaper every 18 mo, but model budgets rise 4× in that span. Every flagship LLM therefore costs 2× more than the last, while knock-off models appear years later for pennies. Benchmark gains shrink and regulation piles on. Net result: each new dollar on the next big LLM now buys far less payoff. The "wait-and-copy" option is getting cheaper every day.

https://www.cnbc.com/2023/11/17/amazon-cuts-several-hundred-...

Of course not being able to monetise Alexa has always been a problem, but these and the article's issues are all to do with poor planning and top tier business direction.

Meanwhile, the models are getting larger and more complex, with more users, putting the support infrastructure well beyond what individuals and even small companies can afford to outright buy. You can easily spend well over a million on even basic infrastructure to try to support some of the newer models and make it available to a few end users.

As a point of strategy for individuals and small entities, it really is cheaper in this case to spin up some AWS instances for a bit to do some LLM work and then spin them down when not in use.

So if you were AWS do you mine for gold? Or do you sell shovels?

Why wouldn't consumer AI be a natural home for Apple?

Apple is constantly under blast for being slow to AI but if you look at the current state of AI, it feels like something Apple would never release -- the quality just isn't there. I don't necessarily think Apple only dipping their toes into AI is that poor of a decision right now. They still have the ability to blow the roof off the market with agents and device integration whenever the tech is far enough along to be trustworthy to the average consumer.

Don't really agree here, yes they screw you financially on cross-AZ bandwidth, but all of their popular services are built to work well across availability zones.

Most people don't need access to a low level consensus service like Paxos, instead they will be using one of amazons managed database services, or s3, that provides higher level abstractions, and automatically manages consensus behind the scenes.

Amazon though, sells physical goods and access to physical servers. Whatever is going on with AI, Amazon will profit from without having to burn its own money in advancing SOTA.

Has anyone had a chance to use Kiro at all? At this point I'm not even interested in it anymore, even if I got an invite.

Why should they need to develop their own models?

If only the technology existed to do work remotely, what a shame.

Why buy the cow if you can get the milk for free?

Because Amazon will build services on top of the technologies that come out.

Just today on hn there was a guy that trained his tiny model and got better results than most of the big models. He wasn’t paid 200m.

The gold rush is here, but the results are still shaking out.

The rambling answer to the “why are you behind” question on the last earnings call indicates it’s a sore spot for leadership, but at this point it’s too little too late. The best talent has already settled elsewhere. The only real saving grace is that if/when the AI bubble pops being so far behind might not be a terrible thing.

No mention of reputation for harsh/ruthless/backstabby management practices towards employees (including for tech white collar, not just biz and blue collar)?

Is that not a major factor? Or are they not aware of it? Or is mentioning it politically off-limits? Or is putting it in writing a big PR risk? Or is putting it in writing a big legal risk?

I know Amazon's reputation for treating employees poorly came up in multiple discussions at one university's big-name AI lab, for example. Not only do some people read the news, but people talk, in groups and privately.

Metaverse will never be FaceBook.

Jassy’s long rambling answer on the last earnings call though does suggest that being way behind on AI is a sore spot for leadership.

Attracting top talent though is a challenge for Amazon beyond just AI. Their reputation has become a real issue and the top folks simply have better options.

My bet is on Apple's upcoming announcement.

It’s no surprise that AWS’s revenue growth is lagging behind GCP and Azure.

Beyond the AI talent gap, Amazon seems to be making serious missteps in its own core business.

It reminds me of Apple. At first, people thought Apple was being strategic by staying out of the AI race and waiting to pick the winner. But in reality, it turned out to be an inability to adapt to the new trend. I expect the same pattern from Amazon.

No! Really? With RTO? Unbelievable /s

I think they learned some hard lessons from Alexa.

Or more likely -- Amazon management knows just how hard writing actually is, how hard to produce something with clarity and signal instead of just common-knowledge cliches, and so they understand that this LLM wave is overhyped. They're letting the other big players do the hard work, and effectively selling LLMs short by abstaining from the race.