Don't think somebody like this can improve ;)

I’ve been thinking about how to grow the fediverse in my neck of the woods as the global south has very little adoption. Besides hosting a localized (Spanish in my case) server, most people don’t even know the fediverse exists. So maybe some old school flyer’s and zines to get the word out about the Fediverse. As it stands, most people aren’t making a choice to not be on the Fediverse, they don’t even know it exists.

absolutely, I'm trying to raise awareness as much as I can, you should come to FediPropaganda@lemmy.world, let's strategize together on how to get more people over here <3

million

Let's continue to do our best to promote, over time people will find the way :))

Doesn't this assume, that it is giving all training data the same importance? I'm pretty sure they rate the quality of the samples and give it different value. An example for this theory is the data by human professionals, hired to train AI.

Yes, I also read several articles stating that many developers felt they were being more productive using AI, while data showed, they reduced in productivity. Wonder how much Claude really changed since then, I was told: that it improved dramatically with the past 2 updates.

Yeah, this guy is also going into software architecture, likely for that reason. However he did say with the recent model the query doesn't even have to be great anymore to get high quality results.

Yes I have read that multiple times and also found the code by normal chatbots e.g. chatgpt to be inefficient and long, however when I brought the performance issue up, he said this is a query problem and the code he is getting from the latest Claude versions is efficient and cloud ready.

Pretty senior, team lead, in his 50s. Also testing/using AI for a while, so not just somebody hyped.

He did say that it isn't great at Unity, as there seems to have been insufficient training.

Interesting take, much of this I have not considered before and can certainly see your logic. The team code vs individual code might certainly offer the explanation. He is now moving into software architecture he said. As AI is not really creative, it might be much easier to compete with it - in that field.

hahahahah 1000000%

ambitious

The costs already ballooned from a few billion to 50 and then probably 100 or 200, not like that's a lot of money in this region of the world though ;)