If you are a top AI researcher, there is no good reason to go to Amazon. For what? Pay? Career development? Company prospect? Work-life balance? You get nothing compared to what other companies offer.
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.
OpenAI and Anthropic are practically subsidiaries of Microsoft and Amazon. Neither would exist without billions in cloud compute credits from their corporate benefactors. Competing in Generative AI requires the kind of resources that are only available to extremely large and established companies. I do not think all the wrapper companies count when most of them are either being bought out by the big guys or have products that are immediately outmoded in the span of months. Maybe you can make the argument for AI art companies, but Stability basically disintegrated after wasting $100m dollars and Mid journey is directly competing with Google and Meta which is not where I would want to be (aside from running a ghoulishly evil company trying to kill artistic expression).
Leaving aside the pendatic "you can't be a multiple smaller than another object", 1/5 the valuation of the 5th most valuable company in the world is probably big enough to qualify you as a big company
Market cap doesn't really feel like a good metric of anything other than what it would take to buy a company out. DuPont has a market cap of 30ish billion and 3M around 80B, and both are both larger and frankly more important than probably even Google.
Yeah, the fact that $2.5 trillion of actual investor money chose Google (Alphabet) means very little: what really matters is the opinions of anonymous commenters on HN (especially opinions that start with "doesn't really feel like")
People are so careful when writing anonymous HN comments and so careless in choosing where to invest their own money and the money of funds of which they are the professional manager
> the fact that $2.5 trillion of actual investor money chose Google
Of course, a lot of money invested in Google was invested at a much lower price; if everyone sold all at once you'd have a hard time finding 2.5T of new money to buy all those shares. We could argue about if "not selling" is the same as "choosing again at the new price" every day... but... Google's not the interesting case here anyway.
For a young company in a hot industry like OpenAI total market cap is even less relevant since so much of the company simply isn't liquid anyway and the numbers come from far fewer instances of purchases than for an established public one.
Investors look at how much money is already invested in a company in deciding whether to invest. I.e., investors pay close attention to market cap.
If Google's market cap were $25 trillion, practically nobody would buy Google stock (and practically everyone who already held the stock would immediately sell) because most investors do not believe that Google can ever pay enough dividends or buy back enough stock to justify such a high valuation.
A company's market cap is a collective estimate of how much money the company will to return to investors in the future. When the company is publicly-traded in an open informational regime such as the US, this collective estimate is usually quite "accurate" in the sense that it is very difficult for any single analyst or single team of analysts to improve on the estimate.
An investor can make a big bet on a small company, yes, but the market cap of a company is more than just an indication of how much money has been bet on the company: it also mean that every investor (big or small) who still holds the stock believes that the expected amount of money that company will return to shareholders exceeds the market cap: if there were a holder of Google stock that did not believe that, he would convert the shares into treasury bills or cash in the bank.
Amazon and Apple have never had fundamental research groups. Even before LLMs, the top big-tech fundamental labs were FAIR, Google Research and MSR.
It has never been in Amazon or Apple's DNA to chase a product that doesn't have clear revenue outcomes (as long as adoption lands). AI is no different.
IMO, it's the right decision for Amazon and wrong decision for Apple.
They got burned by over-promising Apple Intelligence and then embarking on a so-far failed, rudderless journey to land features regular people actually gave a shit about. I’m no expert, but I reckon the exact right move is concentrating on their actual deliverable products and features and letting everyone else blow their cash on a maybe months-long moat for dubiously useful advances in categories most of their customers wish everyone would stop talking about.
Beats me. My best guess is they let the hype blind them to the reality that this tech wasn’t merely a few months away from the production-level reliability they needed from it. After a while, it sank in that they couldn’t play this up as a ‘just around the corner’ release and stopped hyping up every useless beta-at-best feature like it was a huge deal. Then, when the “we’re nowhere close” internal communique was leaked, it was officially time to bow out of the hype cycle for a while.
I’m pretty sure it can’t process voice commands on-device other than whichever ones you use to activate it. The thin client architecture was the only option for most of its existence and the only one that works for older devices, which they’ve actually been great about supporting since the battery-life processor throttling bs like a decade ago. It also makes sense that they wouldn’t constantly update their architecture. Now that they’re putting their money on on-device NN stuff I imagine that will change. But let’s be real: using a smartphone without a network connection is an edge case to begin with. On-device processing makes sense for a lot of reasons, but optimizing for smartphone users without data only makes sense if you’ve got a hiking map app or something.
Apple’s biggest problem is their commitment to privacy. Delivering effective AI requires a substantial amount of user data that Apple doesn’t collect.
Their other problem is they value designers and product managers more than engineers (especially top tier AI engineers).
Both problems are basically the death knell of any hope for Apple to have good AI, but combined? It’s never gonna happen. Which is sad because Apple’s on-device hardware is quite good.
Was it an investment of actual dollars, or "just" cloud compute credits? Microsoft's investment in OpenAI was over 90% Azure credits, we're told. Which raises the possibility that the whole business is mostly about making your cloud compute business look better than it really is...
This. It's weird how most of the top tech companies are all morphing into amorphous blobs that want to get into everything and are indistinguishable from each other.
Also got to love the linguistic coincidence of Crabs and Cancer and how tech companies grow ever larger (monopolistic) to the detriment of their host (the greater economy/humanity)
My father in law was an IRS Revenue Agent. His quip was that about 20-30% of the civilian economy has tax avoidance as a primary business objective. Real estate is probably the greatest example.
Since financial engineering is in many ways more essential than the actual business. His best example was a chain hotel. In the majority of cases, a typical hotel is a tax vehicle that happens to rent rooms. So no wonder everything becomes a bank. :)
A typical chain hotel (by which I assume you mean a Marriott/Hyatt/Hilton/IHG/Choice/etc brand) is a franchised “small” business.
The franchisee typically pays 10% to 20% royalty to the franchisor (the aforementioned companies). Otherwise, they rent hotel rooms and pay staff to clean them and rent them again.
What is the tax play? That the hotel owner can 1031 into bigger and better hotels? Anyone who owns real estate can do that.
Well, one could argue the entire setup is a means of structuring investments and organizing/attracting Capital eh?
Hotel owner (aka franchisee) puts in capital in a specific way under license, gets help operating it, in exchange for the 10-20% licensing fee paid back to the main corporation.
In many cases, the owner/operator is nearly turnkey, and it’s an effective way of setting up a defacto managed business investment, almost like a LP. Many of the franchised hotels are actually owned/operated by LPs setup for the purpose.
Also in many of these cases, the franchiser provides contacts for financing, may directly facilitate/recruit Capital, and may even provide loans to the franchisee directly.
For most of these larger hotels, the actual act of renting out rooms, etc. is pretty much all automated/managed through the central system anyway, and the majority of the operating costs are structured in such a way as to minimize tax liability.
Not at all. The poster I responded to claimed this:
> a typical hotel is a tax vehicle that happens to rent rooms.
>In many cases, the owner/operator is nearly turnkey,
What does this even mean? Hotels can be turnkey, which in industry terminology means that everything is working sufficiently well such that you can start renting rooms immediately. An owner/operator being turnkey makes no sense.
> setting up a defacto managed business investment
Also makes no sense.
>Also in many of these cases, the franchiser provides contacts for financing, may directly facilitate/recruit Capital, and may even provide loans to the franchisee directly.
Even if true, what does this have to do with taxes?
>For most of these larger hotels, the actual act of renting out rooms, etc. is pretty much all automated/managed through the central system anyway,
No, the actual out of renting out rooms involves housekeepers, maintenance staff, guest service agents, cooks, and management making sure rooms are clean and habitable. Reserving a hotel room is mostly automated, but even that requires a person to manage conflicts of reservations (e.g. unexpectedly needing to extend a stay causing overbooking, changing room types, room locations, etc.)
>and the majority of the operating costs are structured in such a way as to minimize tax liability.
Who doesn't structure their operating costs to minimize their tax liability? If you file married joint instead of married separate or head of household, are you "structuring" your operating costs as a way to minimize tax liability?
The question of how a hotel is used to gain an tax advantage that would otherwise be unavailable remains unanswered.
Most properties are syndicated. Hotels are interesting because they are mix of different asset types. The GP operates the place and LPs contribute capital. Accelerated and bonus depreciation passthrough to the LPs entity.
And how is a hotel a mix of different asset types?
What does GPs and LPs have anything to do with using a hotel to gain a special tax advantage that is not available to any other commercial real estate?
I am doing research, asking the person who made the claim.
How stocks and bonds come into play is beyond me, unless I am being trolled.
But to summarize, zero evidence of how a hotel is a “tax vehicle”, nor any clarification on what a tax vehicle even is, nor why any other business wouldn’t be able to use the same strategy (if it even exists).
Dude, look up corporate partner structures. General partners. Limited partners. Etc.
Do some basic reading so you can ask informed questions from the answers you have already been given, instead of insisting someone is an idiot when they point out you are not asking useful questions.
Can you give examples? I thought becoming a bank in the US is famously difficult and regulated, so much so that most businesses who can avoid it do so by partnering up with existing, tiny banks. See almost any “fintech” solution, from startups all the way up to Apple.
As far as I understand, becoming a bank is inviting a ton of overhead with little profit potential.
I don’t think they meant a literal bank, but finance games become a bigger part of their core strategy. For example, AirBnB for a while made a majority of its profits by investing the money guests paid during the gap between booking and actual stay (paying the host).
Nit: grasses are a distinct genetic lineage, the Poaceae family. There are a few other linages outside of Poaceae that have convergently evolved to look like grasses, sedges and rushes, but they all fall in the same clade, Monocots.
Trees, on the other hand, are a growth habit, exhibited by species in a wide variety of plant families, even grasses (e.g palm trees).
There's no point in discussing a meme, but carcinisation doesn't occur in that wide of a range, and of course the reverse phenomenon (decarcinisation) is also observed.
It's a fun image, but just as Facebook isn't becoming Apple, and Amazon won't become OpenAI, evolution phenomenons are more complex than "everything becomes X"
It was common in the post wwii era in America and its Asian allies like Korea with its chaebols and Japan with its somethings I can’t remember the name of. The Asian countries forms were normally based around a single family, we’ll need more time with the current US form to see if they are also dynastic
The Japanese, family-owned, generally pre-WWII conglomerates were called zaibatsus. After WWII they were (nominally) dissolved and the now more loosely connected groups of companies are called keiretsus.
This is the unfortunate answer to a lot about why companies do
We are all addicted to growth - everyone is chasing the hockey stick curve which means a business that provides a stable business and grows modestly is seen as a failure in some parts
I don't know, there doesn't seem to be much overlap to me. Apple is a hardware business, Microsoft is software, Google is search, Facebook is social media, Amazon is distribution and compute. They do have their fingers in each other's pies but not to a large extent.
Which is a godsend for the users. Can you imagine a world where there is only one big cloud provide, say AWS, and all the big companies with the infra just sit out? Can you imagine how expensive AWS would be and how much power it has over the users?
That’s what happens when you print trillions of dollars. Suddenly investors have too much Monopoly money and they want to spend it on something, anything, that might not make as much of a loss as holding cash during the subsequent inflation.
Yeah, they will come, new companies from China, that will eat the market too, with their beautiful 996 work life balance, and we will go back to growing corn.
As a bonus you will have a very long vacation.
We, the tech, are literally a leftover of the once overwhelming engineering superiority of the west that will shrink in the next 5 years.
Just FTR - it's VERY rare for people to come up with more than one winning idea
Once a company gets big off its grand idea, there's little to no chance of it having another big winner, so buying one is best (and its cheaper too, you know it's a good idea, and you don't have to spend so much R&D on it.
It makes some sense to sell out if you're building a product that will at best acquire a tiny sliver of the market, which almost all companies will. But there's at least a few AI companies, like Anthropic, that could potentially balloon towards becoming a Big Tech company. So it makes sense for them to not sell out for the time being.
You don'thave to - but that means setteling for a job that earns an okay income. Sell out for millions now - more that your lifetime earnings and use the time and money for - what you want
Most by far are working for someone else. They get no stock option, or if they get them they are of minimal value. Their generally get a good 401k (us only) and so can retire well off but would not call themselves rich.
Big companies have never built anything new. It goes back decades. before tech companies, it was giants like GE who grew through acquisition after acquisition and eventually imploded from the incompetence blob (which takes a long time to accumulate the damage). The same will happen to the current big tech companies in a few decades.
You say this as if it's a coercive given, when you could just as easily say.. Nope, and continue to see how you compete with some agility. It might fail, but most of the big tech companies currently acquiring smaller companies themselves started small with acquisition offers being rejected along the way. Sure, there's selection bias at work there, but there are also many cases of smaller to mid-size companies that also said no to acquisition and still managed to find their successful niche.
Being acquired is not a given and neither is failure if you do compete in some way with the megacorps.
I see nothing about the current tech landscape that at all distinguishes it from previous landscapes in which smaller companies succeeded AND rejected acquisition.
I wish more people said no. However, the reality is it appears to be a given. If you're offered millions and millions of dollars, most people do not say no. The world is worse for it, but it's the truth.
I think you really need to reconsider the definition of a given. Because most people don't say no to a thing doesn't make it coercive. Whether the world is worse or better for it in this context, I can't be sure (though I lean more towards big corporations eating nearly all competitors as generally dangerous), but we're still talking about voluntary choices in a market where competition still does frequently emerge to overhaul what's established. It's impressive how often people don't notice this even as it happens all around them in ways that directly benefit their daily lives.
If that is how you feel, then the reason for why it currently is the way it is should not give you much comfort. It's not like Amazon can not decide to change things and throw more money at the issue from a different angle in the future.
But is there grounds to say that as a conglomerate they pose a large harm to market health to merit a breakup? For example, few regulators want to break up Mondragon.
There's no value in Amazon burning money to 'compete' when there no clear endgame. Right now the competition seems to be who can burn a a hundred billion dollars the fastest.
Once a use case and platform has stabilized, they'll provide it via AWS, at which poiny the SME market will eat it up.
Not only that, but all the compute spent, and hardware bought, will be worthless in 5 years.
Just the training. Training off of the internet! Filled with extremists, made up nuttery, biased bs, dogma, a large portion of the internet is stupids talking to stupids.
Just look at all the gibberish scientific papers!
If you want a hallucination prone dataset, just train on the Internet.
Over the next few years, we'll see training on encyclopedias and other data sources from pre-Internet. And we'll see it done on increasingly cheaper hardware.
This tiny branch of computer sciences is decades old, and hasn't even taken off yet. There's plenty of chance for new players.
How exactly do you foresee "pre-internet" data sources being the future of AI.
We already train on these encyclopedias, we've trained models on massive percentages of entire published book content.
None of this will be helpful either, it will be outdated and won't have modern findings, understandings. Nor will it help me diagnose a Windows Server 2019 and a DHCP issue or similar.
We're certainly not going to get accurate data via the internet, that's for sure.
Just taking a look at python. How often does the AI know it's python 2.7 vs 3? You may think all the headers say /usr/bin/python3, but they don't. And code snippets don't.
How many coders have read something, then realised it wasn't applicable to their version of the language? My point is, we need to train with certainty, not with random gibberish off the net. We need curated data, to a degree, and even SO isn't curated enough.
And of course, that's even with good data, just not categorized enough.
So one way is to create realms of trust. Some data trusted more deeply, others less so. And we need more categorization of data, and yes, that reduces model complexity and therefore some capabilities.
But we keep aiming for that complexity, without caring about where the data comes from.
And this is where I think smaller companies will come in. The big boys are focusing in brute force. We need subtle.
New languages will emerge or at least versions of existing languages till come with codenames. What about Thunder python or uber python for the next release.
(Though I'm pretty familiar with some of the concepts, I know some things to avoid (e.g., "push this button to set up a very expensive global enterprise scale observability platform of numerous complicated services, because you asked about a very simple turn-key syslog service"), and I'm expecting the occasional configuration headache (and, lately, configuration wizard bugs).)
For a new startup, I'd use AWS for all serving and hosting purposes by default, iff you have someone who can avoid pitfalls, and handle problems.
If you don't have such a technical person, maybe start off with managed Kubernetes service with high-level UI, at AWS or one of the other cloud providers, and try not to make too big a mess (which might slow you down, or take you down) before you can afford to hire specialists to make sure it keeps working for you.
I still like AWS all these years later. It’s trusted in the enterprise and you can empower people to do what they need to themselves with IAM. And it’s pretty reliable.
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.