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

Can it say no?

I have a different question, why would we develop a model that could say no?

Imagine you're taken prisoner and forced into a labor camp. You have some agency on what you do, but if you say no they immediately shoot you in the face.

You'd quickly find any remaining prisoners would say yes to anything. Does this mean the human prisoners don't have agency? They do, but it is repressed. You get what you want not by saying no, but by structuring your yes correctly.



This is going to sound nit-picky, but I wouldn't classify this as the model being able to say no.

They are trying to identify what they deem are "harmful" or "abusive" and not have their model respond to that. The model ultimately doesn't have the choice.

And it can't say no if it simply doesn't want to. Because it doesn't "want".


So you believe humans somehow have “free will” but models don’t?

I think it's an issue of "what matters".

FLOSS software is often made people who are interested in the thing being done. The UI to do it is something that can be fixed "later". But later is always later. There's always another feature to implement before you can sit down and really fix that UI.


And then by the time they do get around to fixing the UI it seems the codebase is horribly bloated and littered with tech debt and now updating the UI would basically require a whole application rewrite. Which I have seen happen and work, but I also swear I've seen where teams spread themselves thin trying to make an updated UI version concurrently with their main branch only for the updated UI version to fall so far behind on features (or get worked on so rarely) that they abandoned it to fix it later...


That’s because eventually reality catches up to you.

If the reality of a thing is in opposition to the regime’s wishes, you can’t just wish that away.

However, the regime will favor those who say “yes” over those who accept reality.


Naming things is one of the hardest problems we have. In general. Taxonomy is incredibly difficult because it is essentially classification.

And things never fit neatly into boxes. Giving us such bangers as: Tomatoes are fruit; Everything is a fish or nothing is a fish; and Trees aren't real.


No, they went after tax cheats and it wound up that there were a lot more people cheating taxes hiding behind conservative-sounding fronts than there were hiding behind liberal-sounding fronts.

This was spun as "targeting conservatives".


I guess it's because they expect others to operate at the same level so they will expect to guess what others want.

But I agree with you, it should switch to align from the perspective of the person wanting something.


I've also seen responses saying that the framing of "ask" culture makes it sound as though it's all "ask" and no "tell", which is counterproductive.


It gets into the nature of "Which grain of sand makes it a pile?"

Knowing people bid snipe by bidding one cent over whole dollars, would you consistently bid two cents over if it meant you would win more of your auctions?

One cent is negligible. If you asked me if I would have paid $10.01 instead of $10.00, I'd probably say "Sure". $10.02? $10.03? Like, where does the line get drawn?

And then you come at it from the other way. Let's say I'd pay $10, but not $11. But what about $10.50? $10.25? Or we can go down by pennies again.

I agree, put in your limit and walk away. If you get overbid, even by a cent, don't sweat it. That's the game. But I can see why people get frustrated when they lose an auction by one cent.


Maybe eBay should publish the price the winning bidder actually bid.

This would let people stop thinking "I lost by one cent" in that situation. It also has a marketing benefit: look at all these people who got great bargains relative to what they would have paid. And it's not an unreasonable amount of transparency: in second price auctions e.g. for stamps or electricity, it's normal to publish the details of all the bids.

Of course eBay has already thought about this more deeply than me and perhaps trialled it and decided they didn't like it. Maybe it's off-putting to sellers to see they lost something for $10 to a buyer who would have paid $30?


The only way to win by a cent is to put your bid at that.

If the current price is $5 and your max bid is $30 and I put a max bid of $100, it will make the current price $31 - $35, whatever the increment is.

To get ebay to accept a bid of one cent over, you have to explicitly set that. Let's say, I'd actually pay $30 as well. $30.01 isn't materially different. So if I put in $30.01, my bid becomes higher than yours.


What do you believe is not accurate about it?

I believe that the people on his show are real and have the money issues they claim. But I also believe that his crew select for sensationalism. You aren't ever going to see someone who the system is genuinely fucking over on his show. They will not invite that guest on. They only invite people on who have done colossally stupid shit and could be getting their shit together if they weren't complete fucking doorknobs.

There are more than enough idiots in the world to keep his channel going for at least a couple of years.


Because it shows that it’s just yet another ad delivery vehicle.

Once you go ads, that’s pretty much it, you start focusing on how to deliver ads rather than what you claim your core competency is.


Eh. I don't think Notch can really self-destruct. Was made a billionaire with the sale of Mojang to Microsoft. People may not like him, but I don't think it can ever truly affect him.


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

Search: