[HN Gopher] Dear AWS, please let me be a cloud engineer again
       ___________________________________________________________________
        
       Dear AWS, please let me be a cloud engineer again
        
       Author : gregdoesit
       Score  : 94 points
       Date   : 2024-07-13 18:41 UTC (4 hours ago)
        
 (HTM) web link (lucvandonkersgoed.com)
 (TXT) w3m dump (lucvandonkersgoed.com)
        
       | tootie wrote:
       | I have to say I find their GenAI offerings muddled at best. I
       | genuinely don't understand what Bedrock is trying to solve.
        
         | refulgentis wrote:
         | Recursive doom loop, we need to mirror everyone else's models
         | at the same price and add Enterprise(tm) verbiage*, lest Wall
         | Street say we don't have a story.
         | 
         | Ironically, trying to compete like that, and then _focusing_ on
         | that, causes the problem they 're trying to avoid IMHO. They're
         | always going to be third as long as MS gets to serve OpenAI and
         | Google has in-house AI talent building models that are top-tier
         | competitive. And if you set this _one farking thing_ aside,
         | they 're #1.
         | 
         | * ex. first time I've seen "FM" as an acronym is on the page
         | you link. they mean "foundation model", which itself is a term
         | I'm likely to see in the Economist, but never on HN.
         | Colloquially, it means "big AI like ChatGPT"*
        
         | atypeoferror wrote:
         | Not sure about Bedrock, but I used Q to write some Python code,
         | and I have to say - it was pretty good. Even ended up covering
         | an edge case I didn't think of. In the end it was all pretty
         | basic (walking a directory tree and compiling some analytics on
         | content) - but it saved me time and I never felt frustrated
         | with it. My anecdotal $0.02
        
       | refulgentis wrote:
       | Hear hear. I'm a founder of a GenAI startup, left Google to do
       | it. And yet, cannot believe how much debasement has occurred, and
       | trust has been lost, by every. single. product. and. company.
       | thinking it has to have an AI story, and it has to be _the_
       | story.
       | 
       | c.f. Google IO keynote this year. I couldn't tell you a single
       | thing Google is launching this year, beyond limited, rushed
       | features where Gemini chat is in a side pane.
       | 
       | And that's not on me: it's because Google literally didn't talk
       | about a single other thing.
       | 
       | And as usual, Google is out of touch and doesn't get the joke,
       | c.f. at the end, Sundar presenting their own count of how many
       | times they said AI.
       | 
       | I _sorely_ miss tech industry of the 00s, I simply cannot imagine
       | ex. 2000 Apple /Steve Jobs falling for this. There's this weird
       | vapid MBA brain drain in charge everywhere. But hey, stonk goes
       | up.
        
         | mattgreenrocks wrote:
         | > And that's not on me: it's because Google literally didn't
         | talk about a single other thing.
         | 
         | Most tech companies don't have a whole lot to show right now,
         | so AI sucks up all the oxygen in the room. This becomes a
         | feedback loop with the stock market, too.
         | 
         | > There's this weird vapid MBA brain drain in charge
         | everywhere.
         | 
         | Yep. And Apple's playing along as well. Their latest WWDC
         | presentation has the most weird tone I've ever seen in their
         | presentations ever: "we added some AI features and they're
         | pretty cool I guess...also it's super private! Here's all the
         | ways it's kept private."
         | 
         | So much hedging going on. So little excitement. Because they're
         | just playing to someone else's tune, and they're not good at
         | doing that.
        
           | refulgentis wrote:
           | 2004 Apple with AI: we need to get 100 people locked down and
           | working on Finder 2.0 with this, and keep execs focused on
           | reviewing it wholeheartedly. this year, let's lol at vista
           | security via confirmation dialogs, vs our best in class OS.
           | 
           | 2024 Apple with AI: this is our best year ever. Look at our
           | CTOs hair. We fixed Siri being eons behind, now if it thinks
           | ChatGPT can help you can tap a confirmation dialogue.
        
         | hnlmorg wrote:
         | > I simply cannot imagine ex. 2000 Apple/Steve Jobs falling for
         | this.
         | 
         | I wouldnt be so quick to assume that. Let's not forget that
         | Jobs bought Siri and then integrated it into every platform
         | they sold.
         | 
         | AI has been a buzzword for literally decades. It's just
         | exploding in popularity right now because the capabilities of
         | GenAI have recently exploded.
         | 
         | It's a little like how VR and AR has been around for decades.
         | It's just taken this long for the technology to make mixed
         | reality a possibility for the masses.
        
           | refulgentis wrote:
           | Yeah, def not arguing my paycheck is thin air -- its that
           | there's a surface level-only respect for it. "Eyewash", in
           | baseball parlance.
        
         | asciii wrote:
         | > I simply cannot imagine ex. 2000 Apple/Steve Jobs falling for
         | this.
         | 
         | No he would not entertain any of this nonsense. IIRC he had a
         | hard time with the Siri demo too. It appears Apple is happy to
         | take itself and shareholders on a side quest with all their
         | cash...good time to be earning interest on it too.
        
       | ko_pivot wrote:
       | I think the main issue here is brand and growth. AWS needs to
       | convince CIOs/CEOs to use them over Azure/GCP, not engineers. And
       | even if AWS cared about convincing engineers, we already prefer
       | AWS so introducing new services for container orchestration
       | wouldn't move the needle. What does move the needle is being
       | perceived by enterprise leadership teams as just as cutting edge
       | as the competitors. "Generative AI" is the only signal those
       | teams understand these days.
        
         | 8organicbits wrote:
         | It's strange because as a senior engineer I'm telling everyone
         | that genAI is not ready, has glaring quality, safety, and
         | security issues, is underpriced by VCs planning to crank prices
         | later, and even if the magic was real offers less promise than
         | tried and true conventional solutions we haven't tried yet.
         | 
         | AWS should be spending significant time explaining how their
         | giant portfolio of conventional tools are improving. If they
         | have stopped, they've lost focus. But hopefully it's just the
         | marketing team focused on magic beans.
        
           | roncesvalles wrote:
           | >AWS should be spending significant time explaining how their
           | giant portfolio of conventional tools are improving.
           | 
           | See but that won't work because leaders are dumb. The central
           | tension of civilization is that the smart ones are needed
           | down in the boiler/operating room to solve complex low-level
           | problems, and whoever's left must steer the ship.
        
         | layer8 wrote:
         | If this is true, software engineering leadership is highly
         | dysfunctional, and that should be an issue of major concern.
        
           | marcosdumay wrote:
           | Welcome to some professional circle somewhere on the world at
           | the 21th century.
        
       | pram wrote:
       | TBH AWS re:Invent is frequently like that. In 2017 or so there
       | was an absurd amount of "Alexa integration" events that were
       | completely useless. I get the sentiment but, you know, it's
       | really just a giant marketing and pre-sales exercise.
        
       | jfengel wrote:
       | I didn't get it. All of the new features they're adding are for
       | AI, but the old features still exist, right?
       | 
       | If nobody wants you to use them, is that because everyone already
       | has as much conventional architecture as they need? Perhaps the
       | new opportunities are all in AI because we've pushed conventional
       | stuff as far as it could go, and we were just rearranging deck
       | chairs.
       | 
       | I'll be honest that, if we've run out of ideas, I dunno if AI
       | really solves any problems I want solved. But even if not I don't
       | see how appealing to AWS fixes anything.
        
         | oaiey wrote:
         | Clouds/PaaSs are just at the beginning at their evolvement.
         | Writing apps is still far too complicated.
         | 
         | So, I do not belief that clouds are at the end of their
         | innovation.
        
         | 8organicbits wrote:
         | > Perhaps the new opportunities are all in AI because we've
         | pushed conventional stuff as far as it could go, and we were
         | just rearranging deck chairs.
         | 
         | There's a ton of low hanging fruit in all the cloud vendor
         | products. Look outside AWS at tailscale, vercel, and fly.io for
         | some obvious examples.
        
         | crabmusket wrote:
         | FTA,
         | 
         | > The same goes for feature releases. If the vast bulk of all
         | new feature releases are geared towards GenAI, it implicitly
         | means AWS is rerouting investments from classic infrastructure
         | to shiny GenAI. It means that the products I love get smaller
         | budgets. It means that the products I use will not get the next
         | feature I want, or only at a slower pace.
         | 
         | I think the article does hyperbolize a bit, but this seems like
         | a hard truth. Unless AWS has hired an entirely new swathe of
         | AI-focused engineering talent, or if their public face at
         | events is significantly disconnected from where they're
         | spending their real money.
        
       | game_the0ry wrote:
       | I feel I am going to be downvoted for this, but...
       | 
       | I think the technical specialty that will be most at threat from
       | automation by AI would be the exact job that he authored has --
       | solutions engineers that build commodity cloud infra on AWS,
       | Azure, G cloud, etc.
       | 
       | Look at progressions and range of abstraction between standard
       | sys admin IT work to serverless deployments, especially with IaaC
       | tools.
       | 
       | You can describe your architecture to chatGPT and it can spit out
       | a CloudFormation YAML. It will be rudimentary and poor, but I
       | could see a Gen AI tool offered by cloud providers where al you
       | do is describe your app and then deployed infra on your behalf,
       | and optimize form there.
       | 
       | Not trying to talk down on folks who do this type of work, but
       | sharing my opinion on where I think the author is ultimately
       | coming from.
        
         | pylua wrote:
         | Serverless, architectless, and programmerless.
        
           | layer8 wrote:
           | AIless will be next.
        
             | pylua wrote:
             | "I'm sorry, I can't let you do that, Dave."
        
         | 8organicbits wrote:
         | > the technical specialty that will be most at threat from
         | automation by AI
         | 
         | Can you point to any actual AI product in this space that
         | functions? Everything I've seen is like, if you squint then it
         | kinda looks like it's doing something, but it's actually
         | producing something embarrassingly wrong, unsafe, or otherwise
         | unusable. And no, having a SME repeatedly prompt until it does
         | the right thing doesn't really make sense.
         | 
         | If we're just talking about hypothetical tools that someone
         | could make, but haven't, we're talking about magic.
        
           | game_the0ry wrote:
           | > Can you point to any actual AI product in this space that
           | functions?
           | 
           | Today? No. But I think we will get here sooner than
           | automating on any other type of engineering role.
        
             | cleandreams wrote:
             | Sure. It's coming with GPT-5.
        
           | fragmede wrote:
           | > embarrassingly wrong, unsafe, or otherwise unusable.
           | 
           | You might know the difference as an SME, but if you're not,
           | and it passes terraform apply, it's getting used.
        
             | 8organicbits wrote:
             | This mentality is why I'm confident that I'll have a job in
             | both software development and security.
        
       | lolinder wrote:
       | > Then there was AWS re:Inforce - the annual security conference
       | - which was themed "Security in the era of generative AI".
       | 
       | This tagline is representative of every part of the hype around
       | GenAI. It makes it sound like security has fundamentally changed
       | and we all need to re-learn what we know. Everything to do with
       | GenAI is treated like this: we need new security plans, we need
       | AI Engineers as a new job title, we need to completely reevaluate
       | our corporate strategies.
       | 
       | Security in the world of generative AI is not substantially
       | different than infosec has been for a while now: User prompts are
       | untrusted input. Model outputs are untrusted input. Treat
       | untrusted input appropriately, and you'll be fine.
       | 
       | The same goes for "AI engineers", who are in the business of
       | wiring up APIs to each other like any other backend engineer. We
       | take data from one black box and transfer it to another black
       | box. Sometimes a black box takes a very long time to respond.
       | It's what we've always done with many different kinds of black
       | boxes, and the engineering challenges are mostly solved problems.
       | The only thing that's really new is that the API of these new
       | black boxes is a prompt instead of a deterministic interface.
       | 
       | Don't get me wrong, there will be things that will be different
       | in the post-LLM world. But my goodness do the current crop of
       | companies overestimate how large that difference will be.
        
         | sylens wrote:
         | While that was indeed billed as the theme of reInforce, there
         | were plenty of sessions and workshops that did not involve
         | GenAI at all. There was a great chalk talk about the
         | underpinnings of how the AWS IAM service works across services
         | and regions, for example.
        
         | throwaway8481 wrote:
         | > Model outputs are untrusted input.
         | 
         | I think the problem is they're trying to introduce nuance and a
         | narrow path to allow this. They want an acceptable level of
         | risk to using untrusted model output for the
         | efficiency/productivity gains it will bring, notwithstanding
         | hallucinations.
         | 
         | Generative AI would not have flown in the security theater of
         | Yesteryear, but CTOs see productivity multipliers.
        
           | usea wrote:
           | > CTOs see productivity multipliers
           | 
           | The CTOs are hallucinating as much as the LLMs are.
        
             | marcosdumay wrote:
             | The GP didn't state the multiplier's value. Those things
             | absolutely are productivity multipliers...
        
           | lolinder wrote:
           | Right, but that's not a new problem either. We want to allow
           | people to send emails with some acceptably-low level of risk
           | that spam will get through. We want an acceptably-low risk
           | that our image upload feature won't be hosting CSAM. And we
           | want it while still getting the benefits of allowing our real
           | customers to pay us for the services we offer. Businesses
           | have been figuring out the balance of risk:reward for as long
           | as infosec has been a concept.
        
         | q7xvh97o2pDhNrh wrote:
         | As a person replying to your comment in the era of generative
         | AI, I'm inclined to agree the hype is a bit much, _even_
         | considering how impressive the technology can (sometimes) be.
         | 
         | Another big area of hype is "prompt engineering." That one
         | seems to have calmed down slightly, but for a while, there were
         | _large_ swaths of the Internet who were amazed that the set
         | intersection of  "talk like a decent human being" and "be
         | precise in your communication" could generally lead to good
         | results.
         | 
         | In many ways, "AI" right now is magic marketing sprinkles that
         | you can put on anything to make it more delicious. (Or, if
         | you're inside a big company, it's magic prioritization
         | sprinkles.)
        
           | bongodongobob wrote:
           | Maybe the prompt engineering should have caught on more. I'm
           | convinced that the large swaths of people commenting here and
           | elsewhere "I don't get AI, it's just a parrot and it's always
           | wrong and hallucinates, it's not useful" just don't
           | understand that the prompt matters and the idea isn't to one
           | shot everything. It writes good code for me every day, so I
           | can only assume they're asking "Write me an OS from scratch"
           | and then throwing their hands up when it obviously fails.
        
             | lolinder wrote:
             | I think that calling it "prompt engineering" is what made
             | it fail to catch on. We didn't call it "Google engineering"
             | back in the day when you could actually craft a Google
             | search to turn up useful results, we called it "Google-fu"
             | [0].
             | 
             | "Google-fu" sounds like a fun skill to learn and acquire,
             | where "prompt engineering" sounds either like something
             | well out of reach or like pretentious nonsense depending on
             | the audience.
             | 
             | [0] https://blog.codinghorror.com/google-fu/
        
             | refulgentis wrote:
             | Been quietly wondering something similar to you for a year:
             | I've ended up 95% confident that phenomena is due to people
             | evaluating it in terms of "does it _replace_ me? "
             | 
             | Cosign prompt engineering. Startup is tl;dr "what if i made
             | a on-every-platform app that can sync and let you choose
             | whatever ai provider, and you pay at cost. and then give
             | you a simple UI for piecing together steps like dialogue /
             | ai chat / search / retrieve / use files"
             | 
             | Seems to me the bigs are completely off the mark, lets cede
             | the idea there's an omniscient AI available now. Cool. It
             | still has no idea how you work.
             | 
             | you could see the hitchhiker's guide the galaxy "42" thing
             | as a deep parody of this category error
        
       | fragmede wrote:
       | > Your first leadership principle is customer obsession: "Leaders
       | start with the customer and work backwards".
       | 
       | > I'm your customer, and I'm begging you: please let me be a
       | cloud engineer again.
       | 
       | Only AWS knows how many H100 GPUs they have, how busy they are.
       | How many people are paying for them, how many people want them
       | and can't get them, and how many people just don't care at all.
       | 
       | It's possible that the focus on GenAI for Re:Invent 2023 wasn't
       | based on any hard data like that, and is really just up to the
       | whims of Adam Selipsky since Jassy moved over, but maybe someone
       | who better knows their planning process can comment.
        
         | dtech wrote:
         | Businesses don't always act like perfectly rational actors,
         | bubbles are a thing.
        
         | LarsDu88 wrote:
         | You make it out as if all the executives at AWS have some
         | master plan surrounding the probably absurd number of GPUs they
         | bought, but the likely answer is its just a bunch of fallible
         | people bandwagoning on the latest trend.
        
         | hnlmorg wrote:
         | I'd wager AWS makes more money from GenAI than any other
         | domain. So it makes financial sense for them to sell that part
         | of the business hard at the moment.
         | 
         | This opinion is based on admittedly anecdotal experience, but
         | I've worked in a large range of domains on AWS over the years
         | and by far the biggest AWS bills were for startups specialising
         | in GenAI.
        
           | AgentOrange1234 wrote:
           | Interesting. I figured all their AI efforts were motivated by
           | FOMO rather than actual returns. Why is AI stuff making so
           | much money? Wouldn't a new area like AI be a loss leader as
           | they try to get market share?
        
       | pylua wrote:
       | This is just in general true of generative ai. In many ways it
       | commoditizes skilled labor. They do not care about people in the
       | posters situation. It is meant to lower the bar and make the
       | labor cheaper.
        
       | skrebbel wrote:
       | I don't know, if you get existential worries about your job
       | because one supplier jumps on a hype bandwagon at a few
       | conferences, then aren't you a bit too married to that supplier?
       | I don't mean from a company perspective, I mean personally. I
       | don't understand how AWS conference topics prevent anyone from
       | being a cloud engineer.
        
         | tuyguntn wrote:
         | when a company focus and strategy shifts to a single
         | domain/product, everything else gets impacted.
         | 
         | Enhancements expected by the community will be delayed,
         | engineers need to come up with temporary solutions which needs
         | to be rewritten next year, new ways of solving problems will
         | not be shared, new content will be created only for a couple of
         | products.
        
       | lijok wrote:
       | An "AWS Serverless Hero" is upset AWS are acting vain around
       | their AI posturing.
        
         | kayo_20211030 wrote:
         | touche
        
       | cperciva wrote:
       | AWS is way too bandwagonny these days. Back when it was all
       | engineers they built things on the basis of "this is cool
       | technology". These days marketing runs large parts of AWS and
       | plans are decided more on the basis of "this will look cool on a
       | PowerPoint slide".
       | 
       | I keep going back to the basics: Serverless is servers. Machine
       | learning is servers. GenAI is servers. And, from what I've heard,
       | most of AWS revenue is servers and storage.
       | 
       | (For the record: I am also an AWS Hero, and an AWS customer since
       | 2006.)
        
         | zdragnar wrote:
         | To be fair, most of that criticism was also applied to "the
         | cloud" and most AWS services that weren't reserved EC2
         | instances back in the day.
        
         | fragmede wrote:
         | How is Lamda, servers? Obviously it is underneath but because
         | there's all sorts of weird tuning shit to get them to work
         | right, it's is own separate thing to deal with and so deserves
         | some sort of way to differentiate it. Are we just unable to
         | handle new definitions for new, different things?
        
           | NicoJuicy wrote:
           | Lambda is servers like kubernetes is servers
        
         | catlifeonmars wrote:
         | Having worked on both (servers and serverless), I personally
         | find the serverless programming model to be fundamentally
         | different. I don't have time to write, operate or troubleshoot
         | servers. Serverless is a huge productivity boost. I can set and
         | forget some glue piece of functionality and it will operate
         | without the issue. Then I can focus coding up the interesting
         | and differentiated bits.
        
       | whirlwin wrote:
       | Good reflections. It kind of remind me of the big data era where
       | everything needed to revolve around big data.
       | 
       | However, what happened is that it became apparent that not
       | everything needs to be big data. Business needs will shine
       | through as they always have and dictate what is truly important.
       | 
       | I'm not afraid of the wave of gen AI. Think of it as the new
       | power tool that just came out that everyone's currently talking
       | about. You'll add it to your toolbox because you don't want to be
       | obsolete. It'll blend into everything else once the hype wave is
       | over.
        
       | roncesvalles wrote:
       | I'm just going to say it: Gen AI is a complete and total
       | nothingburger.
        
         | shermantanktop wrote:
         | Current SOTA genAI tech is more than capable of generating a
         | comment like that - no insight, just a "+1" comment dressed up
         | as an emperor-has-no-clothes hot take for an audience that
         | already agrees with it.
         | 
         | So that's not nothing.
        
       | moltar wrote:
       | AWS really feels they were left behind so they are overcorrecting
       | a bit.
       | 
       | In private they are truly thirsty for AI applications they can
       | write uses cases on that they even offer upwards of 100K credits
       | for Gen AI purposes only.
        
       | hetpatel572669 wrote:
       | AWS mainly earns from new modern managed services because ppl do
       | not want to manage anything and AWS does it for them. All those
       | services are too expensive and AWS markets those and keep talking
       | about those as conventional managed services have less margin.
        
       | Eduard wrote:
       | it's not only cloud services. The AI hype has hit self-hosted
       | services and "normal" / "offline-first" applications as well.
       | 
       | For example, the team / leadership / foundation behind Home
       | Assistant has been pushing AI features hard in the past 18 months
       | or so. This coincides with my feeling that there hasn't been any
       | relevant improvement in Home Assistant's core features and
       | usability -- it's in stagnation for over a year now.
       | 
       | This is of course my own opinion, but it makes sense: if a
       | significant share of resources is spent on AI stuff, that share
       | is not available anymore for other needs.
        
         | crabmusket wrote:
         | This is why we need small company and community owned and run
         | software, preferably open source. Zulip's post about their
         | policies regarding AI training rubbed some the wrong way due to
         | its perceived anti-Slack positioning, but it was at least
         | sensible and measured. Venture backed businesses are all
         | vulnerable to investor desire to chase fads in the name of
         | speculative returns.
        
       | Androider wrote:
       | AWS is desperate to climb up the value stack. Compute and
       | networking is a commodity (with fat margins at retail prices to
       | be sure), and the second and third place providers are willing to
       | make deep discounts to land big deals. That's not going to
       | justify those future lofty valuations.
       | 
       | The problem is, for all it's talk over the last few years, AWS
       | remains a complete non-player in the GenAI space, much less so
       | than Azure. In my opinion the problem is exactly the same as for
       | every other high-level service they've tried to launch.
       | QuickSight, Lex, Polly, Cognito, CodeGuru, SageMaker, etc:
       | they're not good. Nobody ever said "I really like QuickSight, I
       | sure wish it had GenAI capabilities". So when the hastily-
       | expanded QuickSight team(s) then goes on to release 42 different
       | Q enabled SKUs, nobody cares. For various reasons, AWS is
       | organizationally incapable of launching a non-infrastructure
       | product that is simply great, as doing so would take attention to
       | detail and deeply caring about things like UX which are anathema
       | to Amazon.
       | 
       | On the positive side, GenAI model access will be commoditized and
       | part of the basic undifferentiated cloud infra, and AWS will do
       | fine there.
        
       ___________________________________________________________________
       (page generated 2024-07-13 23:00 UTC)