You make a solid point about the growing importance of owning the means of production (i.e. data centers), especially for companies like Microsoft, Amazon, Google, and (to a growing extent) Xai. As AI models get more complex and costly, the ones with direct control over the infrastructure, data, and training resources are in an advantageous position.
Regarding AWS, it’s interesting to consider that they might be playing a longer game. Their focus on providing cloud services without rushing (to quickly) into LLM based AI products could indeed be a strategy to wait until the space matures. When they eventually roll out their own models, they’ll have a massive customer base already using their infrastructure, which could make the transition smoother and allow them to leverage their existing brand power.
The comparison to Amazon’s strategy in product development is also apt. They’ve built a history of entering established markets with a different approach—using their platform’s data and reach to outcompete incumbents. If they can apply the same model to AI, offering better integration with their cloud services and established ecosystems, they could present serious competition to existing players even if they’re a bit late to the party.
Yeah, and that seems especially so for these sorts of models, because the open source stuff isn't that far behind the proprietary things. So you can kind of just wait around and then jump to the front of the line whenever you want, without spending anywhere near the money that the companies that are fighting on the frontier are having to spend. So if you're amazon, it seems like you're in a great spot - just make tons of money from selling compute to AI companies, and if you eventually want to make your own LLM, you can almost choose whenever to do it.
Yeah, and a couple things were already released earlier in the fall. Though I will say, having done this exact same stunt at Mode, you do have to stretch some things. 12 things is a lot of things.
That seems like a marketing/PR axiom; I doubt its veracity.
Marketing is full of accepted wisdom that doesn't appear to be all that wise.
And stunts - unless it's a guy dropping to earth from outer space, or something really funny - can just annoy. If one is invested in the promise and future of a product and communication from that product only wastes your time - well, you've just lost some love and increased skepticism.
Shorter version - companies should stop all the bullshit unless it's deliberate funny bullshit.
I've got mixed feelings about all of it. On one hand, for sure, I very much agree that it seems annoying and like a waste of time. And if it feels low effort or completely soulless, I guess I mostly agree. On the other hand, even the soulless stuff kinda works? Maybe not in the super long term, but I'd guess people have gotten so immune to so much of it that there's not that much penalty for doing it anymore, and there's always a reward for people being reminded you exist.
Yeah, though, that seems a little more psychologically manageable? Nobody blames the gas stations for the price of gas going up. If every AI application is prohibitively expensive, it becomes a kind of collective risk that lots of people wants to fix, rather than one where OpenAI can bury your business and enrich your competitor's at the same time (unless you've got weird pricing structures or bad tech, in which case, that sorta just seems like the natural dangers of market).
Brilliant, funny, and too true! I kept that “companies who might acquire us” spreadsheet, and Amazon was by God on it!! And you are further correct, they didn’t buy us!
You make a solid point about the growing importance of owning the means of production (i.e. data centers), especially for companies like Microsoft, Amazon, Google, and (to a growing extent) Xai. As AI models get more complex and costly, the ones with direct control over the infrastructure, data, and training resources are in an advantageous position.
Regarding AWS, it’s interesting to consider that they might be playing a longer game. Their focus on providing cloud services without rushing (to quickly) into LLM based AI products could indeed be a strategy to wait until the space matures. When they eventually roll out their own models, they’ll have a massive customer base already using their infrastructure, which could make the transition smoother and allow them to leverage their existing brand power.
The comparison to Amazon’s strategy in product development is also apt. They’ve built a history of entering established markets with a different approach—using their platform’s data and reach to outcompete incumbents. If they can apply the same model to AI, offering better integration with their cloud services and established ecosystems, they could present serious competition to existing players even if they’re a bit late to the party.
Yeah, and that seems especially so for these sorts of models, because the open source stuff isn't that far behind the proprietary things. So you can kind of just wait around and then jump to the front of the line whenever you want, without spending anywhere near the money that the companies that are fighting on the frontier are having to spend. So if you're amazon, it seems like you're in a great spot - just make tons of money from selling compute to AI companies, and if you eventually want to make your own LLM, you can almost choose whenever to do it.
"there was Projects, which are folders for your Canvases."
This was one of the daily new feature releases? Folders‽ Oooooooooh.
Yeah, and a couple things were already released earlier in the fall. Though I will say, having done this exact same stunt at Mode, you do have to stretch some things. 12 things is a lot of things.
One could just release the new stuff. People see through this. Do the marketers know we know? We are not clapping; we are annoyed.
Eh, I imagine most people see it as all attention is good attention. It's a stunt, but if it's a stunt that people talk about, it did its job.
That seems like a marketing/PR axiom; I doubt its veracity.
Marketing is full of accepted wisdom that doesn't appear to be all that wise.
And stunts - unless it's a guy dropping to earth from outer space, or something really funny - can just annoy. If one is invested in the promise and future of a product and communication from that product only wastes your time - well, you've just lost some love and increased skepticism.
Shorter version - companies should stop all the bullshit unless it's deliberate funny bullshit.
There is a special place in my heart, however, for radio stations that drop turkeys from a helicopter.
I've got mixed feelings about all of it. On one hand, for sure, I very much agree that it seems annoying and like a waste of time. And if it feels low effort or completely soulless, I guess I mostly agree. On the other hand, even the soulless stuff kinda works? Maybe not in the super long term, but I'd guess people have gotten so immune to so much of it that there's not that much penalty for doing it anymore, and there's always a reward for people being reminded you exist.
Yet all startups who are actually just wrappers will always have a huge risk to manage : their dependency to OpenAI pricing.
Yeah, though, that seems a little more psychologically manageable? Nobody blames the gas stations for the price of gas going up. If every AI application is prohibitively expensive, it becomes a kind of collective risk that lots of people wants to fix, rather than one where OpenAI can bury your business and enrich your competitor's at the same time (unless you've got weird pricing structures or bad tech, in which case, that sorta just seems like the natural dangers of market).
Good point. And they can still rely on other AI providers ..
you have tickled me
I like how you think Benn.
Brilliant, funny, and too true! I kept that “companies who might acquire us” spreadsheet, and Amazon was by God on it!! And you are further correct, they didn’t buy us!
They love to flirt but they'll never date you.