The question is specifically about the marginal cost of the inferencing part exclusively. Not the R&D spend. AFAIK, there has never been a statement on that from neither OpenAI nor Anthropic. What it actually costs to serve one more token determines wether it is "subsidized" at a certain price point. That it is sold for more elsewhere is not evidence.
Many people still 'work-out' at the gym while their 'real' working lives involve close to zero physical effort. Many people 'go for walks' when a motorcycle could have gotten there faster and with less effort.
The utilitarian view of mental 'work' might shift in the same way. Deliberately opting for mental excersize not because of nescessity but for mental health and enjoyment.
Just like in the physical case, there will be widespread mental obesity because many will not.
Do you have actual data on the marginal costs of inferencing, or are you using the term 'subsedizing' loosely as in 'discount', meaning a lower price compared to another price for the same offer?
Amazed Claude didn't refuse this for "cybersecurity". Then I noticed this was this Opus 4.8? Did the author start with 4.8 or was this a fallback from the 5 series?
Apparently they followed an "alternative" reading of the DMA concluding that unless you prove the fees or other obstacles would make alternative distribution commercially unviable or discriminatory, everything is just fine. Isn't it "interesting" how do many of these EU regulations always have these side doors that allow the very practices they promoted to adress?
It has gotten so much worse over the last month. The default writing style of the Claude 5 model series in Claude Code is some sort of jiberish jargon.
Abstention is commercially a no go zone, eating 'different' is a goldmine. It is why 'going on a diet' in the 1970's still meant eating less, while starting in the 1980's the phrase was turned into just eating the latest industry fad miracle food.
Unfortunatly average 'leadership' never had the required skills to lead technically critical 'stuff', and, maybe counterintuitively, because, code is cheap now, right, AI amplified the incompetence.
I do all that, but an 'AS-BUILT' full review of my project still eats 3x my 5 hour budget on max 100€. Meanwhile, my 20€ GPT never hit a limit. Different, but just saying.
Not yet, but in the end high quality tokens are a commodity market. Every optimization to increase inference efficiency will be universally rolled out. The 'hyperspenders' will run into demishing returns unless regulatory capture succeeds.
The default vernacular has become absurd. I have a lifetime in software, decades in AI, but the jargon language the 5 series claude models are spitting out by default makes me go 'what?'.
Slightly off topic: The title hits home. When I was young, books were a luxury item. Less rich people just used the library. But the library had a strict 7 books rental policy. Not enough for the avid reader that also needed a small programming reference library. So I had 3 steady, occasionally 4th, library card pool from family members to keep my bookshelf operational.
Not sure. Do you have numbers? AFAIK having the eID reader was mostly an IT crowd thing. 'Normies' never had them, and even if they did they would have borked on the install.
Why would you release a model if you are the current frontrunner? Only when a competitor pulls ahead, or comes close enough to actually get traffic, you prepare a new release.
It was not an option. Kids need a regular schedule with school and things. It was rough though. I did not ever drink coffee before that time, but starting then I needed several cups just to avoid accidents in the morning.