vitorgrs

28 ms·

301.
by vitorgrs 3y ago
Can you tell all loud in an airport in the U.S that there is a bomb? Would be a lie. But I'm certain that you can't lie about that in public...
302.
by vitorgrs 3y ago
> Is camping and asking for a thing a crime? What is the crime? Article 286 - Inciting, publicly, the commission of a crime: Penalty - imprisonment, from three to six months, or a fine. Sole Paragraph. The same penalty applies to anyone
303.
by vitorgrs 3y ago
Actually, it's a huge metric. Not sure on iOS, but on Android this is so fundamental that usually companies even launch after a "Lite" version exactly because of this.
304.
by vitorgrs 3y ago
That's because the U.S stops Chinese cars. I'll give example on Brazil. Cars were ridiculous expensive lately. Then BYD came in. There were rumors of them launching Seagul (they call Dolphin Mini here) for R$ 100k. Renault had Kwi
305.
Lmsys Arena results are out: Claude 3 Opus behind turbo, ahead of classic GPT4(twitter.com)
3 points by vitorgrs 3y ago|1 comments
306.
by vitorgrs 3y ago
Also, worth to remind that Bing Chat was launched in February 7 with GPT4 already.
307.
by vitorgrs 3y ago
Nop. The model on Perplexity is a finetuned GPT 3.5 (the free one). And the paid versons, well, you can choose between GPT4 (not turbo), Gemini pro, Claude, etc. You can choose their model ("Experimental"), but is not faster tha
308.
by vitorgrs 3y ago
Really depends how good llama 3 will be. I see people thinking it would be better than GPT 4. But they still need to build a model as good as GPT 3.5 even, as llama 2 is... not.
309.
by vitorgrs 3y ago
Japanese, which is in... asia.
310.
by vitorgrs 3y ago
Yes, just like you can get GPT on OpenAI API too. But that's it. You can't get GPT on AWS or any other cloud provider, just like it seems it won't be possible to get mistral closed models on any other cloud providers either.
311.
by vitorgrs 3y ago
That doesn't make sense because GPT4 finished training in August 2022 - before ChatGPT 3.5 release in November 30. You could say they got data to train RLHF after the training, but that seems unlikely. Bing was launched in February 7 w
312.
by vitorgrs 3y ago
Meta actually releases the weights though (for now)
313.
by vitorgrs 3y ago
Which reality? For sure "business man" being 100% white is not a reality for some people in Africa and Asia, don't you think?
314.
by vitorgrs 3y ago
There is AI bias. I think the most common scenario on Dall-e prior to the "fixes", was to ask for "doctors" and only get white people. Never black. The thing is, you don't fix this by changing the user prompt. You f
315.
by vitorgrs 3y ago
Poe has a bunch of them, including Groq as well!
316.
by vitorgrs 3y ago
10 days later answer, but Balanced runs "multiple models", there's a decision making system which choose which model to use. It's mainly focused on search, and being faster to answer. Deucalion is a new MSFT-trained mode
317.
by vitorgrs 3y ago
From my tests, it did better than Gemini Ultra on a few reason/logic questions.
318.
by vitorgrs 3y ago
No. They have early access. Example: MSFT was using Dall-e Exp (early 3 version) in PUBLIC, since February of 2023. In the same month, they were also using GPT4 in public - before OpenAI. And they had access to GPT4 in 2022 (which was when
319.
by vitorgrs 3y ago
The context size is 1 million... They said "they tested up to 10".
320.
by vitorgrs 3y ago
That's not what people are complaining about when they say GPT4 Turbo is lazy. People complain about laziness. It's about code generation, and that system prompt don't tell it to be lazy to generate code. Hell, the API doesn&
321.
by vitorgrs 3y ago
Nokia made a lot of failures, but having too many phones is not the one that affected them. Samsung makes like, the same amount. Xiaomi probably even more. The thing is, if you are in a developed country, you are not seeing half of Samsung
322.
by vitorgrs 3y ago
What? Nokia continued to make smartphones. There was Nokia N96, N97, N8 (which was nice), N900, Nokia 5800... Let alone N9. There was also Nokia 808, which used basically the same camera sensor as Lumia 1020. (will exclude here the Windows
323.
by vitorgrs 3y ago
Which have 0 relations with power generation. There's no lack of energy in Brazil. More likely you went to a state with bad infrastructure (or it was during a storm, pretty possible).
324.
by vitorgrs 3y ago
Maybe, but GPT4 got these puzzles right at the launch.
325.
by vitorgrs 3y ago
They also did a lot of (cringe) ads in Brazil.
326.
Microsoft unveils a new finetuned model for Copilot's Balanced: "Deucalion"(twitter.com)
3 points by vitorgrs 3y ago|1 comments
327.
by vitorgrs 3y ago
Did you tried to use open source models on languages that is not English? Tried asking about things that is not U.S, etc?
328.
by vitorgrs 3y ago
No, you could upload images even before Gemini launch.
329.
by vitorgrs 3y ago
No. Only in Europe Google didn't released Gemini pro for bard.
330.
by vitorgrs 3y ago
You feel Yahoo is less cluttered than DDG?

More ›