https://www.anthropic.com/claude-haiku-5-5#further-updates
This section makes the reader think: why would I not pick Sonnet 5.5 instead of Haiku 5.5?
11,629 karma · joined April 27, 2012
This section makes the reader think: why would I not pick Sonnet 5.5 instead of Haiku 5.5?
740M total (270M text, 170M vision, 300M audio)
> Claude also verified this paper’s main results using the Lean 4 proof assistant with the Mathlib library.
They plan to pay for most of it through excess cash flow, aka profits.
Follow on question: "they make $100s of billions per year? what did they do with all that previously?"
Answer: share buybacks, nothing (aka put it in the bank), acquisitions.
These companies make a ton of money and have a ton of margin over expenses.
I was told to point at someone and say, "YOU. Call 911."
"Thinking" and "intelligence" and "conscious" have no testable definition or specification, therefore it's a complete waste of time to suppose that AI is thinking or intelligent or conscious.
I fear we'll be suffering through these useless semantic merry-go-rounds forever.
One previous example: https://news.ycombinator.com/item?id=45802029
https://www.federalregister.gov/documents/2026/10/02/2026-20...
1. The US economic regulation regime allows one or a few companies to hold an amount/quality of power, beyond which the FTC is supposed to get involved and break up the companies so that power isn't concentrated. This exists because pure capitalism quickly results in one winner, which makes for zero incentive for it to compete/improve/lower prices/increase quality. The fear of anti-monopolistic action by the FTC is low enough and actual action is rare enough that lots of power is unchecked.
2. Humans like bright lights/shiny objects (gambling), feeling tribal/us-vs-them (rage bait wins, negativity drives engagement), are easily exhausted by or ignore options (overwhelming acceptance of defaults), etc.
The above, either in combination or separately, combine to drive low quality computer experiences. Marketing/advertising alone drives an incredible multiple profit compared to actual quality of product/service.
If you're someone at OpenAI or Anthropic and you truly believe what you're making could destroy the world, this is the kind of thing that isn't doing you any favors when it comes to convincing the public. The dissonance here is stark:
- widely proclaiming that your new model is so dangerous it needs to be released only to select people, for safety
- widely proclaiming the model easily found 79 bugs in linux, except that GKH says it took 1 hour to fix all of them because most weren't bugs and the rest were almost all completely trivial, unimportant, and/or not severe
I'm not saying we should assume they'll do anything as long as they believe they can evade punishment, although perhaps we should?
I would say that we get the one (public outrage at specific companies/people) but not the other (focus on the legislation). Unfortunately, the latter has the controlling direct effect.
"A Federal Law Gray Area Allows Meta a Potentially Massive Tax Break."
... or something similar.
The focus could be less about Meta and more about the law. It's not like there's something special about Meta. The vast majority of corporations would do this if it was profitable.
Regardless, the overwhelming majority of the article is about Meta's actions: when, why, how. Very little of it is about the laws governing the taxes, who passed them, when, what's been proposed to amend them, etc.
The general assumption should be that at least some people/corporations will do ANYTHING that is legal as long as it is profitable.
Wreaking "untold damage" sounds like something where a law should exist to prevent it unless there's some incredibly important counter benefit.
> The tax break dates to the first year of the Reagan administration, when Japan...
It seems to me that there could be a lot more discussion and analysis on what led to the laws, who was involved, efforts since to amend them, etc.
The general assumption should be that at least some people/corporations will do ANYTHING that is legal as long as it is profitable.
Previously we've seen "The Secret IRS Files" from Propublica as well as other coverage of tax avoidance that never resulted in tax evasion charges. Coverage wide enough that it's arguably safe to assume that no laws were broken because there was quite the demand for an enforcement response.
The funny thing is that a more accurate title for these pieces would be, "Federal Government Allows [tax-paying entity] a Massive Tax Break." And then the piece would be all about how the senate, congress, and president either explicitly wanted there to be this extreme tax loophole or was incompetent and allowed something they didn't mean to allow.
Folks are continually getting upset and riled up about players in the economic system instead of who and what makes the rules.
An analogous example: everyone got cranky about Martin Shkreli, mostly because he raised prices on Deraprim, although also because many would label him a brash jerk. But he went to jail for finance stuff, not raising prices. Why? Because raising prices in the way he did wasn't something the feds thought was illegal or wasn't something they thought he could be convicted of.
Folks get mad at Musk, Zuckerberg, etc... if they're doing something illegal, ok yea that makes sense. If folks are mad about actions that are legal, it makes zero sense to be mad at the actor. Be mad at the people/entities who created the legal loopholes and laws.
> Through careful analysis and manual review, we have identified what we consider to be clear and concrete indicators in one or more of your homework assignment solutions that it was partially or entirely generated by an AI/LLM tool like ChatGPT, Claude, Copilot, etc....
> You are required to submit a response to this form....Failure to respond will result in a grade of 'F' for the course.
> Honest responses that align with our analysis will result in a score of O for the identified assignment(s).
> You also have the option to meet and dispute our findings. Should you choose to do so and we determine that our original findings of AI/LLM usage stand, we will instead apply the standard course policy from the syllabus and assign a grade of F for the course.
The above letter sent to a bunch of students.
> It is an open question whether this is reflective of their actual behavior or instead something that underscores the purported, and unintentional, potentially coercive nature of the form.
Sounds like the professor discovered (or disputes the possibility of?) the concept of forced confessions.
"Honest responses that align with our analysis": did the professor consider the possibility of dishonest responses that align with the analysis?
"Should you choose to do so and we determine that our original findings of AI/LLM usage stand" Not much needs to be said about that.
If the professor was confident enough about the analysis to send the letter, then all of the students getting the letter should have gotten an F and no letter at all.
Is it reasonably expected? Yes, definitely, if that means "much more likely than violence against non-teachers": a random K-12 teacher in the USA is ~2.5-7x more likely to be assaulted than a random person in the USA.
Seems like it could still help but also feels like one of those things where it becomes vastly more complex and difficult to debug.
Hrm. I'd say that all of these things are possible when using services of companies that are valued at billions/trillions in the stock market. So "impossible" might be too strict.
bbb.name can ONLY be registered if it is not already registered AND there are no 3rd levels registered on bbb.name currently.
My "no worries" comment was a bit tongue-in-cheek. Who knows how far the tech will go or how cheap it will get.
These folks compete in a reasonably zero-sum game.
Presumably, the actual text of the message was reviewed by a human? So the human(s) didn't get the humor either?
EDIT: by "humans" I mean the FBI / cops.