Trump Orders Every Federal Agency to Stop Using Anthropic
(truthsocial.com)
You're viewing a single comment thread. View all comments, or full comment thread.
Comments (21)
sorted by:
Sorry, fren. You clearly aren't using it on a daily basis. It is transformative - unlike anything that has come before. And, Anthropic's Claude is currently the absolute leader when it comes to BUILDING anything. I'm now able to do in a day, what it would take me a team of 10 TALENTED developers months to create just a year ago. It isn't the devil. It isn't Skynet. IT IS A TOOL. A powerful tool. And our DoD deserves every ounce of every powerful tool we have at our disposal. Full stop.
I work directly on it, its been my exact job, training and building models exactly like what you are using. They have not increased in capability at all, the only thing increasing is adoption and thereby the amount of training data for it to pattern match.
Thats all its doing. And its now to the point where it already peaked... about a year ago I'd say. Up to that point it was reading mostly good data, produced by humans with full logic and human thought. Now tons of new data being injected was created by itself... and is subtly flawed. Little hallucinations here and there, and they gradually degrade the models, and the more human produced training data disappears the worse it will get.
You'll see.
I have found that using an architecture.md file explicitly with every single prompt has dramatically lowered hallucinations.
To your point, junk in junk out. However, the reverse is true as well. Clean in, clean out and I have found it amazing.
Maybe I just keep lucking out. 🎲🎲
i have all the .md files. none of that matters when it is actually generating code.... i'm not sure you understand what I'm saying. You can prompt all you want. The AI isn't really thinking, its just probabilistically matching against things its been trained on to generate next token after next token.
Its like a compression algorithm... artifacts creep in... if you keep just relying on combining and iterating eventually you will not even recognize the result.... and its all just garbage. If humans are not in the loop anymore it will just keep getting worse.
I don't have PhD in the subject like most of the people I work with on it, because I'm self taught, but generally they all agree. Short of some unknown tech that will suddenly be revealed that is genuine intelligence... its not what the media keeps saying it is and there will never be "singularity"
I enjoy using it, and in my opinion, it keeps getting better and better. That is just my opinion, though. I am actually more worried about it getting better too fast than about it stagnating or getting worse. We'll see how it turns out. Thanks for your opinions. Have a great day.