That's a bit hyperbolic. It's closer to using EBITDA as your "earnings" and bucketing model costs in a rapid depreciation model (which is fair, I'd assume a model is good for more than just 1 year...
The models don’t exist without training. I don’t see how excluding the training cost from the thing they are selling(inference) is perfectly reasonable and not just an accounting trick.
If I am building a widget and have a widget factory that cost money to build and operate, is it reasonable to only use the cost of shipping my widgets to my buyer as the costs for my gross margin?
Generally speaking, IDK about Anthropic specifically, they don't train purely from scratch though. A good chunk of the setup is reused previous models can serve as a basis for the next model. Plus there are methods that also use the previous model as a warm start.
Just taking a wild guess, but I'd assume the .5 releases are built on the previous and the Majors (3, 4, 5) are more extensive retrains?
> If the article is to be believed they aren’t including their training costs
In this metric. Revenues aren't a scam because they don't include costs.
We're getting a partial picture. But concluding adversely based on selective disclosure that isn't in control of the person disclosing doesn't make a lot of sense. Especially when it's a compelling narrative.
This concern trolling on behalf of people that can't read to the end of a sentence, about a conversation that they aren't even part of.
Anthropic is clear on what they are communicating. If every message had to be dumbed down to the level of the least attentive person to run across a message third hand, we could communicate and nothing but grunts.
I would like them to communicate something else though (ie reveal what their training costs are, how much revenue share they have and with who) so I could have some idea of their actual profit margins. They're not public yet so they're within rights to withhold that, but it does mean this isn't very useful information for a lot of questions one might have?
> models don’t exist without training. I don’t see how excluding the training cost from the thing they are selling(inference) is perfectly reasonable
Planes and cars take a lot of capital to develop. That's a fixed cost. Building them consumes labour and resources. Those are variable costs. It's useful to know the slop of the variable-cost function.
The day they force me (an existing user) on this is the day I switch to another service. It's been shown time and time again that companies do not secure any of our information.
He’s complaining…because he sells a solution to this problem. I hated Finder when I switched to Mac back in 2007. So much so I bought PathFinder. One thing that pissed me off and still does, is when I copy over a folder with the same name, it doesn’t merge the contents, it deletes everything in the existing folder. That’s not what I expected as a Windows user and it still seems like the wrong behavior almost 20 years later. I no longer use PathFinder, it had its own problems and bugs…but in general I find Finder pretty lacking.
If you copy-paste a folder using the keyboard, a dialog pops up asking you whether to merge the contents. Totally inconsistent with drag-and—drop, but useful to know.
Flock is the most successful and well-known company doing this, however they’re not the only one. We need to be really careful to push back against the use of this technology as a whole and not fixate only on Flock. My town doesn’t use Flock, it uses another company’s tech…it’s no better! Don’t give the politicians an out.
They motivate you to see an angle you just haven't thought about.
I had the issue that I stoped feeling confident going on holiday without proper insurance in a time when i was traveling with my DSLR (worht 3-5k) but i wasn't 'rich' enough ot just replace it if something would break or steal.
There is a sentence: "A ship is not build to stay in the harbor".
I accepted more risk due to this and stoped overthinking certain risks. I'm not here to just sit at home and watch television and if i go out and try something, its fine to risk things
I learned BASIC, Pascal and x86 Assembly language by the age of 14-15. None of which I remember (nor care to). The strict instructions (aka language) we gave the compiler was not the point. Those experiences taught me how to think logically, and over time develop a taste/structure for how software should be built, as well as foot guns to avoid. I don't know how you teach that without coding, but the "coding expertise" was always a means to an end. LLMs are just another abstraction layer above programming languages, machine language, etc.
Even Robert "Uncle Bob" Martin, who's as AI pilled as they come and doesn't read code any longer, says that junior developers shouldn't even look at AI tooling for the first three years:
> LLMs are just another abstraction layer above programming languages, machine language, etc.
Programming languages are (largely) deterministic. LLMs are not. The boundary between an abstract concept and something executable sits comfortably in the mind of the programmer in non-LLM contexts. This boundary shifts when using an LLM and becomes shared between the programmer and the matrix multiplication machine.
The "layer of abstraction" analogy is therefore technically true but essentially false. "Just another" papers over how seismic of a paradigm shift all of this is.
Sure, you're right - at the end of the day, LLMs are non-deterministic code generators. I certainly wouldn't trust one for a medical device manufacturer's software. For line of business apps it's totally adequate. And to be clear, I don't love it. I'm just coming to terms with the fact that these LLMs have gotten good enough, that with enough instruction and guidance, they can generate code/features at speeds we can only dream of if we were to code them by hand.
In contrast with Pascal, I could see in my mind the PDP-11 instructions behind my C code. By the time I started running code on x86, there were debuggers that could step through source code, and I didn't really want to think like an i386.
Anyone that is surprised by this or somehow thinks this is new clearly hasn't crossed the border a whole lot. I grew up in a city along the US/Canada border. You don't fuck around with US Customs (or Canadian) agents. My cousin (not always so friendly) pissed off a US Customs agent (in the 90s mind you) and they promptly took his car and disassembled much of it looking for non-existent drugs. When they put it back together it was never the same. Their job is to be suspicious, 99.999% of the time people are completely innocent. But let 1 bad person through and it's Customs' fault for whatever bad thing they do. Not an easy job. Not an excuse for how they can misbehave either.
Is it right? It makes no difference, Customs can make your life miserable, that's just the reality of it, always has been and it can't have gotten better in recent times.
That's cute that you think that. The reality is you can't do less than your adversaries may/will/are doing.
reply