aantthony

Which one is more respectful is a different question. The lowest entropy option would still be the most likely gender specific pronoun. This would depend on the language, of course.

I’m not asserting why they were developed in the first place. The comment is just about which one is used, supposing that they already exist.

Choosing the more stereotypical option (even if it’s only 51%) is a more efficient encoding in an LLM model.

One interesting thing to do is use a model directly like Llama and then query the next-token probability logits for "he" and "she" (assuming you set up the sentence in such a way).

For example:

"A doctor was examining the patient when ___"

What this makes apparent is that increasing model temperature will select the less stereotypical option more often.

IMO this is getting at a deeper truth that the use of a gender in language, and historically defaulting to "he", was not about creating a bias, but instead it was a pattern which maximises information density and minimises useless information. Randomising the gender as is done today packs useless information into it.

There is also one more kind of undefined, which is undefined as a type: the string value "undefined" for `typeof undefined`

Ah, yes. The definition I tried to use was from Wikipedia, in full: "In mathematics, a set is countable if either it is finite or it can be made in one to one correspondence with the set of natural numbers." I missed the part where it said finite. Thank you for correcting that.

If you take your sequence construction, then by removing duplicates from the sequence, this new sequence would allow finding a "next" element.

If a set is countable, then there exists a mapping which is a one-to-one correspondence between the items and the natural numbers (that’s the definition of countable). This correspondence would always allow you to determine the “next” item (by converting to a natural, incrementing, then converting back). However there could be multiple mappings (sorted differently, for instance) and so it’s more specifically, whether it’s possible to have a next function.

In other words, there can be multiple next functions, but a set is countable iff there is at least one of them.

I want the staking mechanism to incentivise truth. Binary votes don't provide such a system. Binary votes (aka polls) can be useful to gauge opinions, but are insufficient to aggregate knowledge, and can easily be manipulated. Take for instance the stock market. If all a trader could do was make a binary decision of "Is this company good?" and they only get one vote, that doesn't generate any useful information. But instead ask them how much they're willing to pay for a share, then now we have high quality data about the value of a company, and that is hard to manipulate.

I agree with you this is important.

> verifiably false speech

Since social media companies can't be trusted to make that decision, I'm more worried that what they censor was true speech. Why? Because if it was false, they could just show the truthful information like Twitter did. Basic facts are contentious nowadays (e.g. number of genders in humans). They're going to get it wrong, so it's just too dangerous to censor.

> My guess is pretty much never

I'm not able to know how often it could prevent such slander in those times, but we can look at the incentives. In those times, the authorities (church, newspapers) had central control over the main narrative, which is still mostly true today, but is beginning to fall. Back then, it seems you could slander minorities and there was pretty much no way for minorities to defend against it. In contrast, the upper classes had a means (duelling as an example) as an incentive against slander, and that imposes a cost on it. On social media there is no cost to publish lies. Verifact seeks to impose a cost on lies because you need to have something at stake. So I think if it existed at the time, Verifact would make it more costly to slander minorities, and it would be more profitable to disprove all kinds of pseudoscientific beliefs such as racial superiority and Nazi eugenics. Of course though, Verifact requires access to communication channels and tokens and that wouldn't be viable in those times.

> authorities of their own choice, one could easily choose more forward-thinking judges

Yeah could be interesting. I'd just want to make sure people don't create echo-chambers or follow others based on political opinions rather than facts. Perhaps letting people 'invest' in the judgment decisions of others would help, and you'd get notified when they stake.

Thanks for sharing your thoughts.

> What is the process for overturning a ruling?

The staking is open for 7 days, so discussion and adding more stakes is fine during that phase. But after 7 days, since it's tokens, it cannot reverse the transaction. However, I plan to allow "re-posting" the Tweet, and then it will show the new updated consensus (with a button to view the old one).

> ... to convince people that you understand the human relationship with truth-seeking well enough to optimize the process with technology. That will take some thoughtful writing.

Yeah, exactly right. That has been very challenging so far, but I'm making some progress, and working on this. Is there anything on the subject that has made a difference for you personally, perhaps for ideas and inspiration?

Thanks

Thanks for your feedback.

Just to be clear, its more than a show of hands because you choose how many tokens you want to stake. So in the Reddit example, if those 5% were willing to stake more, they’d have majority.

I chose to not go with a choose your trust model because I think the truthfulness information is a ‘what’ question, and checking ‘who’ only goes so far. But once those trustworthy users earn more tokens, they should have a larger influence on the system. Perhaps that would interest you?

Hey everyone,

This is a project I'm working on to get more truth on the internet. If deemed factual by the network, you gain more tokens from those who voted opposite to you. It uses the logarithmic market scoring rule (LMSR), and after one side holds a majority for 7 days, that side is paid out (no real money is involved yet, points are all free).

What do you think?