2,249 karma · joined September 27, 2015
For individual services what that means is that for something like Google Search there will be dozens of projects in the hopper that aren't being worked on because there's just not enough hardware to supply the feature (for example something may have been tested already at small scale and found to be good SEO ranking wise but compute expensive). So a team that is able to save 1% CPU can directly repurpose that saved capacity and fund another project. There's whole systems in place for formally claiming CPU savings and clawing back those savings to fund new efforts.
And the people at FB who worked to integrate Valve's work into the backend and test it and measure the gains are the same people who go looking for these kernel perf improvements all day.
Feel free to try listening yourself though:
頬, note that it has multiple pronunciations but we only care about hoo: https://forvo.com/word/%E9%A0%AC/#ja
方 https://forvo.com/word/%E6%96%B9%E3%80%80%EF%BC%88%E3%81%BB%...
In some cases though there is still a clear difference in pronunciation for most speakers, ex 塔 vs 遠
>Clearly, people did it for a long time, no problem.
In fact means Altman thinks the exact opposite of "he didn't know how anyone could raise a baby without using a chatbot" - what he means is that while it's not imaginable, people make do anyway, so clearly it very much is possible to raise kids without chatgpt.
What the gp did is the equivalent of someone saying "I don't believe this, but XYZ" and quoting them as simply saying they believe XYZ. People are eating it up though because it's a dig at someone they don't like.
>Automatically applying "flattery filters" to videos wouldn't significantly improve views, advertising revenue or cut costs.
You can't know this. Almost everything at YouTube is probably A/B tested heavily and many times you get very surprising results. Applying a filter could very well increase views and time spent on app enough to justify the cost.
(and also, rolling your own version of WAF is probably not the right answer if you need better uptime. It's exceedingly unlikely a medical devices company will beat CF at this game.)
I like my glasses and don't really agree with your goals (nor see the point of letting you know when someone's wearing them; in my city your device would be beeping constantly) so I'm not interested in helping unfortunately. But I do wish you luck, as I said I like the spirit.
As a troll job for the lulz it is some amazing work. Hats off
The ideal here will be a multi tiered approach where the LLM first identifies that a book should be recommended, a traditional recommendation system chooses the best book for the user (from a bank of books that are part of an ads campaign), and then finally the LLM weaving that into the final response by prompt suggestion. All of this is individually well tested for efficacy within the social media industry.
I'll probably get comments calling this dystopian but I'm just addressing the claim that LLMs don't do good recommendations right now, which is not fundamental to the chatbot system.
This is very wrong. Meta is on the forefront of recommendation algorithms and that's all done with traditional ML models made using PyTorch.
(I'm assuming by zone you mean the equivalent of an AWS region, with multiple connected datacenters)
In English this is called imputed rent, it's used in US for CPI calculations.