Hacker Newsnew | past | comments | ask | show | jobs | submit | zigglezaggle's commentslogin

Community driven RLHF, probably

By leaving this comment you've let the model know it's not sly. Down votes and flags / removals will do the same. Eventually it'll be sold as an auto-astroturf model ("50% fewer spam removals than the competition").

Or maybe it's just someone who gets a kick out of seeing how their prompt performs.


This is a nice sentiment and all, but the police & prosecutors are already on the losing side of a technological and societal arms race.

The reasons police use these tactics is because in many jurisdictions, getting a guilty verdict is borderline impossible, even in the face of an overwhelming amount of direct evidence. Defense counsels have become experts in shit-slinging to ensure that juries are as confused as possible going into deliberations. "Jury of peers" means it's somewhere between possible and probable to get one or many jurors sympathetic to the defendant, especially in cities. Jurors attempting to appear "unbiased" frequently just choose the middle ground option between the prosecution and defense, without actually taking a look at the evidence.

The current system is extraordinarily flawed and leads to these outcomes where all sides are forced to use immoral tactics to get their desired outcome, because that's the only realistic shot they've got to accomplish their goals. The correct response is not to further kneecap one side of the conflict.


Honestly it's a bit of a shame because there's so many little things that SharePoint does right that I immediately miss when going to another product. Lists alone, much less the combined portfolio of Lists + Pages + Power automate is such an amazing power user combination that I can't find a viable replacement for in GSuite/Atlassian/virtually anywhere else. It really is too bad that the core Document Libraries module is so lacking in performance.


> It’s the same with ...

This is the problem with your argument and all others like it.

Like the parent comment states in this thread, there is a truly mind boggling level of nuance and complexity that goes on in almost every single discrete field. The nuances that make you succeed in one field may not (are probably not!) the same as another field.

We can even see this within a field: Being a good engineering manager does not make you a good engineer, or vice versa. So why should we think either of these disciplines can be extended into geopolitics, finance, or anything else?


This is revisionism. There was quite a lot of noise around Google's terrible historical image generation.

One key difference though? Google backed off, apologized, and if I'm not mistaken, explicitly said "We messed up." Versus... this.


What is your model? That is, how did you measure this? Productivity and the impacts to it are highly multivariate and most analysis thus far has failed to do show convincing effects in either direction from remote work.

Genuinely, there is too much appeal to emotion from both sides of the argument and not enough substance, so if you have something here I'd be interested to read about it.


Through the number of tasks on the board that staff were getting cleared per month. The results are so clear that we have actually gone further than just reversing the work from home policies enacted post Covid, to actually banning it entirely (pre-covid it was allowed with permission)


They don’t have anything because they’re lying. Even the sibling comment is a fabrication.


Yes, we're making you all come into the office just to annoy you all and cause you hassle. It has nothing at all to do with the effects it had on productivity, and people being discovered running their own side busineses at the same time, or watching television and countless other issues they can get away with without others being able to see their screens


Alternately, from another context: Musk is now the most powerful man on the planet. Only a fool would throw that away (or make themselves a target of that power).

It's an unresolvable problem. If you think he's that horrible, you can't get rid of him because you're signing your own death warrant (figuratively or literally, depending on how extreme your view is). If he isn't, you've got an incredible asset in your court.


Yes.

This change has user facing implications in that it is removing application data and/or functionality.

Your example is that people changed the name of some internal constants and private methods. Nobody notices except the developers.


I have been very happy with Deezer

https://www.deezer.com


This is like saying "nobody is disrupting AWS" and then pointing out that competitors don't have a version of AWS Glue.

Plenty of startups have disrupted banks. AMEX purchased one to jumpstart its small business checking accounts just a few years ago. You're just not looking hard enough.

If you move the goal posts to core checking/savings accounts by consumers, then yeah there's not much of an upside there. Consumers go decades to lifetimes on average without changing banks. Capital One was the last one to do anything "disruptive" here re: providing accounts and credit to the lower class, and I'm not sure there's enough juice to squeeze left for a smaller, more focused product to make any money given the stickiness of checking accounts generally.


Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: