woko

Gemini 4 Argon (High) looks comparable to Claude Opus 5.5 (High). https://artificialanalysis.ai/models/comparisons/gemini-4-ar... From that perspective, the benchmark is not too disappointing, given that there are only 8 days between the blog posts (September 30 vs. September 22).

The group is not afraid, but the individuals who got charged probably stopped all black-hat activity: there was an individual in his early twenties who "was sentenced to three years in prison and ordered to return $5 million", which is life-changing to say the least.

> Very interesting, potentially useful. But if doesn't spit up exactly what you want can't edit it further.

DALL-E 2 spits as many outputs as you want. Then you choose the one you prefer.

This situation becomes obvious when you want to use the search box to retrieve old conversations. I use Discord every day, so I fall in the first category in your post. However, every time I try to find an old conversation, Discord UX and search engine make it super hard, if not downright impossible. I can feel the pain of a user who would have just arrived on the server.

Somehow, I missed that. Thanks! I did not realize logging was part of the demo. :D

> I work at a company with >1000 google workspace users.

> That's enough that someone at Google will acknowledge what you're reporting is a bug on their end, and that they can reproduce it. But it's not enough to get the bug fixed.

I will try to remember that the next time I deal with an open-source project, either as a user raising an issue, or as a project contributor helping solve such issues.

Often, we do not realize how lucky we are that contributors to open-source projects help fix the bugs which we report.

> The non-lethal flashbang grenade temporarily blinds anybody within its concussive blast, making it perfect for flushing out closed-in areas.

cf. https://counterstrike.fandom.com/wiki/Flashbang

These two people are respectively 22 and 28 years old. They have spent their time in cyber-crime environments since their teenage years.

Respectively, a third of his life for the first one, and half of his life for the second one. And that is a lower bound.

The first one had an account on a cybercrime forum 8 years ago, so he was 14!

> These last two nicknames correspond to accounts on several top cybercrime forums way back in 2013.

The second one had an account 13 years ago, so he was 15. Wow!

> [T]hat forum’s database says a user by the name “Damnating” registered with the forum in 2008.

It is clear by the mention of "several orders of magnitude" that we are talking about a database which fits in a few MB of disk space on a machine with a few GB of RAM.

If the interviewee has never worked with a database which would fit in the memory of the machine, and if the company only deals with such databases, then it is clear that it was not a good fit. Failing the interview does not mean that the interviewee is bad, just that they don't know how to count that low, and thus were not the person for the job.

Basically, you are presented with a small-scale (possibly toy) problem: the database consists of 50k entries and fits in 3 MB of disk space.

Then the interviewee starts wondering out loud whether he could load it in full on a machine with 8 GB of RAM.

From what I remember, the suspect is believed to be the leader of an anarchist group responsible for squatting places, to struggle against the move of wealthy people to North-Eastern Paris.

They have made the news for squatting one place in particular: the restaurant "Le Petit Cambodge", as it is one of the restaurants targetted by terrorists on November 13, 2015.

Here is an article in French about them: https://www.lefigaro.fr/societes/a-paris-un-local-du-restaur...

For instance, it is discussed in this press article [1] published in 2016 and available in French. It is mentioned that nuclear power costs 60€/MWh, compared to 80€/MWh for wind power. It is also mentioned quite often in the French press: I have seen it mentioned a few days ago. The source mentioned in the 2016 press article is an official report [2] published in 2014.

However, it is true that the maintenance cost increases as the power plants get older.

[1] https://www.francetvinfo.fr/replay-radio/le-vrai-du-faux/le-... (in French)

[2] https://www.ccomptes.fr/fr/publications/le-cout-de-productio... (in French)

The government policy for a reduction to 50% is not based on issues with nuclear power itself. As far as I have seen, there are mainly 2 reasons mentioned in the public debate: i) a concern about the disposal of nuclear wastes and the long-term cost of their storage, and ii) the desire to follow other countries' policies, with a focus on wind, water and sunlight, rather than leading the way with the cheapest and most modern solution for electricity production.

The decision was written in the law in 2015 [1,2], with a statement that renewable electricity (wind, water) should reach 40% of the production, which implies that nuclear power should not account for 70% in the future.

Overall, I believe this is not a bad idea because i) renewable electricity is bound to get better thanks to worldwide research, and ii) it is smart not to put all your eggs in the same basket.

[1] https://www.legifrance.gouv.fr/codes/article_lc/LEGIARTI0000... (in French)

[2] https://fr.wikipedia.org/wiki/Loi_relative_%C3%A0_la_transit... (in French)

If you are a company whose business is advertisement, you should not get a free pass to *require* *unnecessary* personally identifiable information from your target audience.

There is no good argument for making it a *requirement*. If the platform has issues moderating its content, then there are solutions other than requiring their users to identify themselves. Plus, as we saw, asking for ID does not help fight against problematic content and its spread. The root of the problem may lie elsewhere, e.g. in the system which boosts user-generated content which draws clicks (clickbait) and likes (outrage and fake news). Asking for ID is only good for the targetted advertisement business, telling otherwise is a decoy.

An introvert is "a typically reserved or quiet person who tends to be introspective and enjoys spending time alone" [1]. So I agree with you that it is surprising to be a "complete introvert" and to look for social interactions with strangers, even though they are 1-to-1. The people whom I met and who are extremely shy or timid would tend to avoid these, I think.

That being said, I can see how the poster would define himself as an introvert (avoiding social gatherings with a crowd), and be more comfortable alone or with 1-to-1 discussions (actually 3 people, since there is her friend as well). The adjective "complete" in "complete introvert" might be exaggeration though. :)

[1]: https://www.merriam-webster.com/dictionary/introvert

In practice, this has not happened yet, and it has been 3 years since the forum inception.

One obstacle which I forgot to mention is that an account cannot report posts or warn other members unless the account is 3 months old *and* the account has created at least 300 posts. Both conditions have to be met. I guess it is a sufficient hindrance for most Internet trolls to forget about the forum if they had no intention to take part in the community in the first place.

That is post-modernism [1], the tendencies of our era. Boghossian's commentary [2] published in 1996 is still relevant these days, though it would need to be put in context.

> Postmodernists like to respond [...] that both claims can be true because both are true relative to some perspective or other, and there can be no question of truth outside of perspectives.

[1] https://en.wikipedia.org/wiki/Postmodernism

[2] https://as.nyu.edu/content/dam/nyu-as/philosophy/documents/f...

It depends on the scale. I can vouch that crowd-moderation works fine for a small forum (~ 1000 members) which I am part of. And there is no karma system. You get to report posts (3 reports mean that the post is deleted), and warn users (24-hour ban after 3 "active" warnings, and then it scales up to a permanent ban after 15 "active" warnings). Warnings become "inactive" after a month.

It also depends on the threat model. If the community is the target of an harassment campaign coordinated by external actors, then you might need additional tools, or people dedicated to the job. However, this won't necessarily solve the problem, as external actors could double-down, and moderators can lose their minds (suspicion of a troll behind every post, abuse of power, absence of control of the moderators, possible presence of a spy/agitator among the moderation team, etc.). I won't name the forum and the community, but I have a specific one in mind. It does not help that it is a source of information for gaming media, which means that it is often linked to in press articles, which attracts much attention from all kinds of people.

That being said, I get back to the subject: user-generated content on platforms (and not just forums). If the goal is to reach a large scale, then I fully agree with you.

There is ImgBB. You need to create an account, but it is free.

https://imgbb.com/

Keep in mind that the real goal here was to "test scalable alignment techniques", and OpenAI ended up adopting "recursive task decomposition".

That is basically a "divide and conquer" approach for simultaneously learning a complex task *and* allowing humans to evaluate it.

> To test scalable alignment techniques, we trained a model to summarize entire books. [...] This work is part of our ongoing research into aligning advanced AI systems, which is key to our mission. As we train our models to do increasingly complex tasks, making informed evaluations of the models’ outputs will become increasingly difficult for humans. [...] Our progress on book summarization is the first large-scale empirical work on scaling alignment techniques.

> People remember.

No. People don't remember over a sufficiently long period of time.

For instance, you mention China, Russia and the British Empire, but if you look further in time, you get to see other things which nobody remembers:

> The Islamization of Xinjiang started around 1000 AD by eliminating Buddhism. [1].

> Many Buddhists fear that their countries will lose their culture and become Muslim, as had been the case in many parts of modern day Central Asia, Xinjiang, Afghanistan, and Pakistan, which were majority Buddhist before the arrival of Islam in the 7th-11th centuries. [...] When the Muslim Turkic Qarakhanids captured the Buddhist city of Khotan in Xinjiang in 1006 CE, one of their poets penned this verse: “We came down on them like a flood/We went out among their cities/We tore down the idol-temples/We shat on the Buddha’s head.” In the Islamic world, a destroyer of idols came to be known as a but-shikan (بت شکن), a destroyer of but, a corruption of the word Buddha. [2]

Long before the Islamization of Buddhist Uyghur, there were "Caucasoid" people, which would be impossible to know without the discovery of the Tarim "Celtic" mummies (~2000 years BC) [3].

> From the evidence available, we have found that during the first 1,000 years after the Loulan Beauty [~4000 years ago], the only settlers in the Tarim Basin were Caucasoid. East Asian peoples only began showing up in the eastern portions of the Tarim Basin about 3,000 years ago, Mair said, while the Uighur peoples arrived after the collapse of the Orkon Uighur Kingdom, largely based in modern day Mongolia, around the year 842. [4]

[1] https://en.wikipedia.org/wiki/Xinjiang#Demographics

[2] https://thediplomat.com/2017/10/buddhism-and-islam-in-asia-a...

[3] https://en.wikipedia.org/wiki/Tarim_mummies

[4] https://www.independent.co.uk/news/world/asia/meeting-civili...

This quote is specifically about Diophantine equations.

You can find a similar statement as "Proposition 3.2.2" in Chapter 3 "Modular Arithmetic" [1] of this introductory course [2] to Number Theory. Here is a picture of the statement [3].

Basically, if there exists a number n such that there is no solution modulo n, then you know there are no solutions. Here is a simple application [4] of the proposition with n = 2 and 3.

[1] <http://www2.math.ou.edu/~kmartin/intro-nt/ch3.pdf> [2] <http://www2.math.ou.edu/~kmartin/intro-nt/>

[3] <https://i.ibb.co/8mNHW0Q/2021-09-20-10-34-25.png> [4] <https://i.ibb.co/Qffr9k2/2021-09-20-10-56-29.png>

Moreover, I don't understand how having a built-in adblocker is seen as a selling point. Wouldn't it be more powerful to rely on a third-party add-on like uBlock Origin?

Here are a few examples of human figures from 30,000 years ago [1], and 35,000 to 40,000 years ago [2, 3]. For comparison, the eroded camel sculptures are from 8,000 years ago at best.

Moreover, beware of possible biases which could explain what we see today: there is a difference between what we observe and what was. Remember for instance the destruction of the Buddhas of Bamiyan [4], which happened less than 1,500 years after they were built. It is extremely unlikely that a work of art survives for 40,000 years, and some may be more likely to be destroyed than others, depending on location, weather, traditions, history, etc.

[1] https://en.wikipedia.org/wiki/Venus_of_Galgenberg

[2] https://en.wikipedia.org/wiki/Venus_of_Hohle_Fels

[3] https://en.wikipedia.org/wiki/Adorant_from_the_Gei%C3%9Fenkl...

[4] https://en.wikipedia.org/wiki/Buddhas_of_Bamiyan

It depends on the offer. Remember that recent piece of news?

> Newly-released emails reveal how Google hoped to pay $50 million for YouTube in 2006, and why it ended up paying $1.65 billion

> https://www.businessinsider.com/email-reveal-google-hoped-bu...

There are several issues with your statement.

> The mail service that handed over data of a customer to a foreign government

First, ProtonMail can only hand over meta-data, because data is encrypted.

Second, "ProtonMail does not give data to foreign governments; that’s illegal under Article 271 of the Swiss Criminal code. We only comply with legally binding orders from Swiss authorities."

> and changed the privacy statement on their site

The privacy policy was not misleading if you read it carefully. It was not "changed" as in removing a lie from the statement. At best, it was clarified to ensure *everyone* would correctly understand it in the future. It is accessible at https://protonmail.com/privacy-policy

What may have been misleading was the marketing message on the homepage. If you pondered each word of the one-sentence marketing message, you could have guessed that the expression "by default" was there for a purpose. Companies do not add useless words for marketing, they do it to avoid false advertisement. However, this is not the same thing as the privacy policy. And ProtonMail stated that they would fix that: "we will be making updates to our website to better clarify ProtonMail’s obligations in cases of criminal prosecution".

Quotes are taken from: https://protonmail.com/blog/climate-activist-arrest/

GPU can be useful for Nearest Neighbour as well. In case you have access to a GPU, I would strongly recommend Facebook's FAISS [1,2]. For everything else, sklearn is amazing.

[1] https://faiss.ai/ [2] https://github.com/facebookresearch/faiss

See the FAQ: https://scikit-learn.org/dev/faq.html#will-you-add-gpu-suppo...

It would be interesting to monitor his number of followers before/after the controversy. For information, John Gibson, CEO of Tripwire Interactive, which is a videogame publisher, recently tweeted a controversial opinion about abortion, then had to step down from his job, yet his number of followers increased from 2.5K to 9K in the process.

This can be seen on websites like: https://socialblade.com/twitter/user/rammjaeger/monthly Picture: https://i.imgur.com/poc1QcE.png

Edit: when it comes to Piers Morgan, it does not seem to have been worth it. The last bump of followers happened in March.

Reference: https://socialblade.com/twitter/user/piersmorgan/monthly Picture: https://i.imgur.com/s7145UK.png

You are correct. See:

> Judge seems alarmed at the extent to which impulse spending by "exorbitantly high spending gamers" drives App Store revenue.

> Says it's outside the scope of antitrust but says issue is "worthy of attention"

[1] https://twitter.com/stephentotilo/status/1436365652164493317

> "in most economic ways...the App Store is primarily a game store and secondarily an 'every other' app store"

[2] https://twitter.com/stephentotilo/status/1436365985800507394