kadushka

8 ms·

61.
by kadushka 1y ago
This might makes sense if you believe US is an evil empire, DoD is doing bad things, and AI will help DoD do even worse things. But it's not so black and white, is it?
62.
by kadushka 1y ago
With self-driving cars some human will be held responsible in case of the accident, I hope. Why would it be different here? It seems like a responsibility problem, not a technology one.
63.
by kadushka 1y ago
If there's not enough money from taxes, they will borrow/print more to cover total deficit (not specific to DoD). Otherwise, tax money will go directly to DoD.
64.
by kadushka 1y ago
Why do you think humans would make better life or death decisions? Have we never had innocent civilians killed overseas by US military as a result of human error?
65.
by kadushka 1y ago
If you live in US, taxes you pay directly fund DoD. So if you sponsor their activities, why can't Anthropic do business with them? Which other company would you rather get their (your) money?
66.
by kadushka 1y ago
I’m thinking about limiting the extent I use it for coding. Just to stay sharp. Need to exercise my brain more.
67.
by kadushka 1y ago
Thank you! I guess if there's enough spelling related text in the dataset, a model is forced to learn some info about token composition in order to predict such texts. I wonder if it would help to explicitly insert this info into an em
68.
by kadushka 1y ago
Oh yes, you’re right, I was thinking about claude.ai
69.
by kadushka 1y ago
The $20 one doesn't have Opus It does.
70.
by kadushka 1y ago
As long as it solves my technical tasks, I don't care what political biases it has.
71.
by kadushka 1y ago
Is a human who never learned how to read not generally intelligent?
72.
by kadushka 1y ago
Is there any evidence to support your hypothesis?
73.
by kadushka 1y ago
Great, then we established that I (generally intelligent human), and LLMs, both can't perform the task in question without specific training. We still don't know if given that specific training we (a specific human or a specific L
74.
by kadushka 1y ago
I got 3 tokens: st, raw, and berry. My point still stands: processing "berry" as a single token does not allow the model to learn its spelling directly, the way human readers do. It still has to rely on an explicit mapping of the
75.
by kadushka 1y ago
o3 is definitely usable, as I said, it solved about half of the coding tasks I tried. My problem with your original comment was "bottlenecks in genAI is not the knowledge or accuracy". Knowledge and accuracy are absolutely the mai
76.
by kadushka 1y ago
Each character made the cut, but the word "strawberry" is a single token, and that single token is what the model gets as input. When humans read some text, they can see each individual character in the word "strawberry"
77.
by kadushka 1y ago
These models are not trained on character level input. Why would anyone expect them to perform well on character level puzzles?
78.
by kadushka 1y ago
Codex paper confirms that GPT-3 could not do any coding tasks. It's right there in the abstract: https://arxiv.org/abs/2107.03374
79.
by kadushka 1y ago
Because gpt-3 was not trained to do coding tasks. It could do a simple autocomplete. Perhaps you are confusing it with gpt-3.5?
80.
by kadushka 1y ago
gpt-3 could not do any coding tasks.
81.
by kadushka 1y ago
Last time I checked they were trained to predict the next word. Why would you expect them to solve research problems?
82.
by kadushka 1y ago
What did he do with gpt-3?
83.
by kadushka 1y ago
Move somewhere cheap and work remotely.
84.
by kadushka 1y ago
Doesn’t sound like you’re ready to get back to work yet. I’d extend the sabbatical.
85.
by kadushka 1y ago
We have reached a point where the bottlenecks in genAI is not the knowledge or accuracy, it is the context window and speed. You’re joking, right? I’m using o3 and it couldn’t do half of the coding tasks I tried.
86.
by kadushka 1y ago
Are LLMs trained to do all that? How do you know they cannot?
87.
by kadushka 1y ago
I cannot do all that. Am I not generally intelligent?
88.
by kadushka 1y ago
Are you ready to lose your job, permanently?
89.
by kadushka 1y ago
There will still be a functioning economy - serving the elite class. There will be a million people total who control all the resources. These people will form a new society, will have their own government, their own laws, their own values,
90.
by kadushka 1y ago
So then there's no need for AI workers. You got this backwards - there won’t be need for humans outside of the elite class. 0.1% or 0.01% of mankind will control all the resources. They will also control robots with guns. Less than 1

More ›