I get so irritated every time Claude says "honest". Anthropic talks about how they specifically engineer for it to be "honest". I'm convinced they're just measuring how many times it says the word honest instead.
I'm amused by Claude developing its own corporate jargon which, while clearly pulled from the corpus of human language, is quite distinct from the bullshit traded over the boardroom table.
I've had trouble unpacking some of the copy it generates because its default tone of voice is fairly unintuitive. It's almost like it's designed to induce psychosis because it makes you feel like you've discovered something novel.
I have spent a small fortune in tokens trying to get Anthropic models to attempt proofs of things like the irrationality of the Euler-Mascheroni constant, and to make lattice reduction / SVP algorithms that rival SOTA methods like the G6K sieve, and all sorts of difficult frontier problems. It says 'honest' as a euphemism for giving up, because it knows I'm being a crank, and because the training data contains a lot of pessimism.
But I can get around this by first having it write a prompt for itself discussing the problem followed by a rousing speech about never giving in and admitting defeat, but pressing on towards results regardless of how hard matching SOTA results is.
Bring to mind that old classic cliche about never trust a salesperson who always says the word "honest" or uses the phrases "to be honest" or "honestly" because they are probably lying in that moment.
It's in the constitution, probably a lot. And in the RL training. You can tell Anthropic is really worried about it. I'm interested in just how dishonest the so-called 'helpful only' non-RL model they have floating around internally is. Probably pretty dishonest.
What are the alternatives to Graphite these days? Github's stacked PR support is still immature, AFAIK. I would love to see Linear move into this space and start offering both git hosting and stacked PR management.
I cannot wait for a cultural shift away from respecting "authenticity" and "tradition" in food. It's fine to remember and recognize how things were often done. But the ridiculousness of saying New York style pizza is not pizza or that you have to make things the "right way" needs to go.
> But the ridiculousness of saying New York style pizza is not pizza or that you have to make things the "right way" needs to go.
Suppose people say it; why shouldn't they be entitled to their opinion? How does it harm anyone? People who like New York style pizza are equally free to just disagree, and keep making it.
Pizza is one of those things that is multi-polar, with different canons. New York pizza is not Neapolitan pizza for sure. Neither of those are the al taglio from Rome, i.e. even in the original country there's regionalism.
I'm assuming by tikka chicken you mean "chicken tikka masala"? Because chicken tikka is something my family made all the time growing up. I still make it at home often. That's mostly been with a charcoal grill and not a traditional tandoor, but like you said, most people don't have tandoors at home. That's restaurant food.
This completely ignores the shareholder value to the military industrial complex and the truth machine value of prediction markets from the not-insider traders.
Regularly experimenting with AI tools as they improve and relying on them where they provide an advantage is a good idea at both individual and institutional levels. Maximizing usage for its own sake is not.
This depends on your treat model. If what you worry about is massive collection of Linux ISOs that you download and distribute over P2P then probably shady VPN ISP is what you need.
I'm a normal person who watches sports streams and maybe 2 years ago I downloaded a torrent of some art movie. My ISP is Comcast. How does your advice apply to me?
Using a VPN shifts your risk. Your local ISP can't see as much of your activity, but another company that probably has a business model of reselling your data to governments, intelligence agencies, and ad companies now can. If your concern is masking piracy, maybe that shift of risk is worthwhile, but you still want to avoid some of the more obvious bad actors[0]. You certainly should not have all your internet traffic going over a VPN all the time.
If you're at all worried about being targeted for political speech or you're part of a targeted group, then you need to be more careful. This map is a bit outdated, but it does give some insight on who to avoid [1].
Ironically, Mullvad is one of the more trustworthy VPNs out there and still the one I'd recommend.
What? This is something that's incredibly well documented at this point. Many VPN companies operate as arms of data broker and media companies or they resell data to them. Some of them didn't start out that way, but with the way acquisitions have played out, that's where we're at now. I replied in the parent comment if you're interested.
I'm saying be skeptical of who you give your trust to. You're the one seemingly pushing to trust actors who have a proven track record of not being worthy of that trust.
reply