Avatar
Vilmar
eed2b88e2998d187816d5ee2492723fada53eb4ba62e3902ab92de5ae990234b
Replying to Avatar Fabio Akita

Interessante. É a porcaria do UOL, então muitos grãos de sal.

Acho que falamos disso. Em vez do Cloudflare ter respondido no mesmo dia a qualquer ordem - o que me soou totalmente estranho e sem sentido. - E quem ficou me enchendo o saco também tinha achado "olha só como todo mundo odebece como deveria menos o X", também foram provados errados.

Matthew Prince, CEO da Cloudflare declarou que não fez nada pro X desbloquear e também não contribuiu nada pra bloquear de novo, como afirma nesse artigo:

https://noticias.uol.com.br/internacional/ultimas-noticias/2024/09/23/cloudflare-ceo-declaracao-x-brasil.htm

Fazia mais sentido que foi o próprio X que entrou no dashboard da infra e reconfigurou pra ser IPs reservados exclusivos em vez de compartilhados. Eles já estavam pensando em atender o bullying mesmo.

Então foi isso mesmo, trabalhando pra atender as demandas absurdas, engoliram e eles mesmos contribuíram, o que também põe mais um prego no caixão do mimimi repetitivo de que eles não querem contribuir.

Mas o palpite que faz mais sentido é que vão criar entrave, não importa se arranjar representante legal e o que mais. Esse nunca foi o problema. O timing foi bom. Agora só vai voltar (pra mostrar que são "misericordiosos") depois das eleições.

Vlw nostr:nprofile1qqswwrcx5dlyv3fhf7nt55qg28qag0msan4cxjdwr69x09fwhzt857spr3mhxue69uhkummnw3ezumt4w35ku7thv9kxcet59e3k7mgpr9mhxue69uhhq7tjv9kkjepwve5kzar2v9nzucm0d5q3vamnwvaz7tmjv4kxz7fwwpexjmtpdshxuet5snh2lc nostr:nprofile1qqsplpv5y65h6g5pr6vx73h73wa6pnwejx5842n96gw6fe6mjkg30pspzemhxue69uhkyet5vyhxummnw3exjmpwvdsk6qfdwaehxw309akxuerfwd3hxtnyw43kkerwwvhx7un89ahx7um5wfex2mrp0yhk54n9veghsm66qydhwumn8ghj7mn0wd68ytnzd96xxmmfdecxcetzwvhxgeg89q2wp

Replying to Avatar Fabio Akita

OpenAI lança o novo modelo O1 hoje, 12/09. O primeiro demo que ele apresenta nas redes sociais é o O1 conseguindo fazer um pequeno joguinho (muito, muito simples, coisa de um único arquivo com um único page down, não fiquem emocionados!)

No Dia do Programador? Coincidência? I think not! 😂

Preparem-se que todo jornalista tá super emocionado de novo. Eu fiz pequenos testes preliminares e vou dizer que não senti tanta mudança assim comparado com o 4o anterior. Sim, parece que ele gasta mais tempo em passos intermediários, tentando quebrar problemas complexos em problemas menores, pra ver se consegue uma solução melhor. É o que chamam de "chain-of-thought". E de fato, pra certos tipos de problemas, parece melhor mesmo.

A idéia do nome "O1" parece ser um "soft-reset". Não vai mais chamar "ChatGPT 4" ou "ChatGPT 5", vai recomeçar do "OpenAI O1", então vez o próximo seja "OpenAI O2"?

Minha teoria da conspiração pessoal (coloquem chapéu de alumínio), é que eles entendem o "peso" que seria lançar "ChatGPT 5". Eu venho falando isso faz tempo: uma versão 5 teria que ser ordens de grandeza superior ao 4, de uma forma que qualquer pessoa comum consiga ver a diferença.

Se falharem em lançar um "5" superior, eles arriscam explodir a bolha de I.A. Por isso eu sinto essa "relutância". Notem que eles travaram no 4. Tentaram sair com "4o" e agora "O1", mas nada de "5". Não querem se comprometer, mas também não podem deixar todo mundo esperando pra sempre.

É uma saída de marketing. O O1 parece mesmo superior ao 4o, mas benchmarks não servem pra nada. Você pode simplesmente enviesar o treinamento pra fazer passar melhor em testes de benchmark. O lance são testes reais.

E em testes de código, eu não senti tanta diferença em relação ao 4o. Lógico, tem que testar mais tempo, com problemas mais diversos, por enquanto é só uma anedota. Mas de novo, não senti nenhuma grande diferença pra justificar jornalista falando "isso vai mudar o mundo!!" e toda essa emoção.

Até peguei um textão de um jornalista e passei pelo O1 pra ver o que ele acha. Alguns trechos que achei engraçados:

O1: "Check OpenAI's official channels to confirm the release of "OpenAI-01" or "o1" and gather accurate details about its capabilities."

OpenAI precisa começar a treinar os modelos pra saberem seus próprios nomes e características pelo menos. Eles nunca sabem!

O1: "Benchmark Results: Re-examine the reported benchmark performances, such as the 83% score on the International Mathematics Olympiad (IMO) and the Elo rating of 1807 in coding competitions. Ensure these figures are accurate and sourced from credible evaluations.

Comparisons with Human Experts: While it's notable if the model performs similarly to PhD students on challenging tasks, it's important to contextualize this achievement and acknowledge any limitations."

O próprio O1 manda tomar cuidado quando se fala de benchmarks e não assumir "uau, é mais inteligente que um PhD do MIT"

O1: "AGI Timeline: Exercise caution when predicting the arrival of Artificial General Intelligence (AGI) by 2025. Acknowledge the speculative nature of such predictions and the ongoing debates within the AI community."

Obviamente, jornalista emocionado de novo falando "AGI está perto" e o próprio O1 pelo menos tem bom senso de falar pra baixar o tom.

O1: "Avoid Hyperbole: Statements like "It will change the world" and "It's a day to celebrate" may come across as overly enthusiastic. Strive for a neutral tone that presents facts and lets readers form their own opinions.

Balanced Viewpoint: Acknowledge any limitations, ethical considerations, or areas where further development is needed."

Achei super legal que o O1 avaliou o tom do jornalista como entusiástico demais, pra ser mais neutro e pra explicar também as limitações e áreas de melhoria. Até a porr de uma GenAI tem mais bom senso que um jornalista médio. Tá foda.

Testem vocês mesmos e me digam o que acharam: sentiram muita diferença em relação ao 4o?

https://openai.com/o1/

Rachei o bico kkkkk

Replying to Avatar Lyn Alden

Good evening Nostr!

Here’s a show review, which some here have been waiting for. It's one of the few shows I've liked in years.

But before that, it’s also about how Nostr social graphs can make reviews better. I tend to write long, but I do so with a purpose. If you don't give a fuck about the context and want to cut this down by more than half, I'd happily recommend you scroll down past the context session toward the review section.

*Context*:

My excitement and enjoyment hit rate of liking shows and movies lately has been quite low. It’s probably because I’m at odds with my general culture. As we go through a transitional period or weird zeitgeist or “fourth turning”, it’s rare to actually find visual content I like. I’ve had to turn to novels instead over the past several years. Visual productions have a lot of people watching over them and fixing them (i.e. making them worse) whereas a book still has a main author that can kind of put his or her concept out there, which is neat.

Back in my teen years, I liked anime a lot. So I’ve got a nostalgic base there. Cowboy Bebop was my defining favorite, but also Trigun, Neon Genesis Evangelion, Baccano!, Code Geass, Full Metal Alchemist, Akira, Samurai Champloo, FLCL, Durarara!!, and both Dragon Ball Z and Naruto kind of embarrassingly, and then also of course Studio Ghibli productions like Princess Mononoke, Howl’s Moving Castle, and Spirited Away.

While American television in the 1990s and early 2000s was episodic and happy, Japanese shows tended to be darker, more emotional, and tended to have continuity from one episode to another and tell a more linear and complete story, as though I was watching a continuous graphic novel from beginning to end. That caught me, and decades later I still remember their plots.

After that, from my mid-twenties and beyond as I was more seriously focused on my professional life, I occasionally tried to watch newer animes in the 2010s and thereafter, but I rarely connected with them like I once did. They felt too immature, or something wasn’t clicking. I liked action, but I didn’t like kid/teen drama and immature heroes, or something adjacent to that. I loved what I grew up on, but new things felt weak, as though I outgrew them. Occasionally something was decent. That’s where Baccano! and Durarara!! and the Brotherhood version of Full Metal Alchemist kind of filled the transitional gap. But I grew apart from the genre.

As an insanely busy person, one of the things I lack for mature anime is a good discovery mechanism, which imo relates to the social graph and Nostr here. I don’t go out and look for good anime anymore, and even if I were to find a high-rated one, I wouldn’t know if those ratings are from a hundred-thousand 15 year olds that might not appeal to me in my 30s. Those reviews aren't context-related.

Some weeks ago, Shinobi aggressively shilled Blue Eye Samurai on Twitter as one of his highest recommendations in years. Over many years I’ve followed him for bitcoin tech content, some of which I agree with or disagree with, but always find value in seeing his viewpoint, and over those years I also know somewhat about a bit of our overlapping anime tastes including Cowboy Bebop, and this was one of the few times where he aggressively shilled a new animated show. So, it caught my attention.

That brings me to my point about social graphs. An aggressively positive review from someone you know with overlapping interests with you is worth 100x random reviews. I hadn’t even heard of this show despite it being released over six months ago, and I watched this 80% because a single taste-relevant connection aggressively shilled it, and 20% because upon googling it the other mass reviews said it was unusually good too.

*Review*:

I watched the 8-episode first season of Blue Eye Samurai. It will have a second season, so it’s not complete, which admittedly kind of annoys me.

But holy hell is this show fucking good. I loved it, which for me is unusual. We start with an interesting protagonist, the blue-eye samurai, and then we increasingly learn about their backstory, which is better than expected. It could have fallen into all sorts of cultural traps but did not; instead it just told a good story. Like the old days.

The animation is great, the story is great, etc. Great protagonists, great antagonists.

The first three episodes are solid and expected from watchers of the genre, but then in episode four it went into completely unexpected territory. I was like, holy shit! And then episode five was also absolutely amazing. That middle period is where I confirmed that this show was truly special, and that a lot of thought went into it.

The next few episodes were great, and without spoilers, my criticism is that it could have ended on a more satisfying and convincing note. I need to see follow-on seasons to truly judge the show. The middle-season was the strongest part thus far, but the early and late season parts were still great.

Rated R, highly mature in terms of violence and sex, but also brought back a sense of nostalgia. Amazing visuals, perfect voice acting, and not quite but nearly perfect storytelling. I have some criticisms, increasingly toward the end, but still open since the show hasn’t ended yet. My husband also brought up similar criticisms.

As a frequent cynic, disappointed with what I watch, I was absolutely thrilled with this, which is rare for me. Later seasons might fuck it up, but in terms of potential, this first season is great, and brings me back to my nostalgic mature anime days.

I’ll watch the second season as soon as it’s out, good or bad.

Ok. You convinced me to watch it haha

what was the real target? what was inside the building?

nostr:npub1a2cww4kn9wqte4ry70vyfwqyqvpswksna27rtxd8vty6c74era8sdcw83a, ela fala bastante sobre liberdade

Replying to Avatar Ivan

Sorry, I'm not from the US. What was in the smaller on?

There's a manifestation planned for today to say no to the authoritarianism of the judiciary in Brazil.

Unfortunately, I don't think it's going to result in any changes, there were already others last year.