rytill

9 ms·

61.
by rytill 3y ago
Did you try using the API and providing a very clear system message followed by several examples that were pure JSON?
62.
by rytill 3y ago
What kind of novel UX are you imagining?
63.
by rytill 3y ago
You're exactly right. I'm not sure what the point of this is.
64.
by rytill 3y ago
Why do you find that funny?
65.
by rytill 3y ago
Right, I’m surprised to see so few engaging at the level of the actual logic the proponents have. Many people on HN seem to disagree with the premise: they believe that AI is not dangerous now and also won’t be in the future. Or, still beli
66.
by rytill 3y ago
What do your users want? Will they understand your offering faster if you explicitly mention AI in your messaging?
67.
by rytill 4y ago
Does a primitive definitely need to be easy to test or deterministic?
68.
by rytill 4y ago
You're not seeing this the right way. You are saying the equivalent argument of: "Look at how much hand-holding this processor needs. We had to give it step by step instructions on what program to execute. We are still a long way
69.
by rytill 4y ago
Why would you assume that?
70.
by rytill 4y ago
Can you describe how this software could sustainably be developed to serve you, and where you would fit in to that ecosystem? How would someone who makes software to help you interact with your database and write queries using AI buy food a
71.
by rytill 4y ago
Why not get a loan? Can someone explain what fundamental difficulty there is in taking out a temporary loan to make payroll until the FDIC sorts things out and you make X% of your money back?
72.
by rytill 4y ago
Paper: https://www.cs.toronto.edu/~hinton/FFA13.pdf
73.
by rytill 4y ago
If WolframAlpha were more open, there would be more discussion about it. That's the price WolframAlpha pays for its extreme walled-garden, black box strategy.
74.
by rytill 4y ago
> For some references, I recommend classification of mystical vision (psychedelic-assisted or not). Could you clarify what you’re referencing here? I wasn’t able to find something that looked like what you were referencing by searching “
75.
by rytill 4y ago
Are there model weights?
76.
by rytill 4y ago
For a temperature of 0, the highest probability token is predicted for each step. So “my favorite animal is” will end with “a dog” every time. With higher temperatures, lower probability tokens are sometimes chosen. “my favorite animal is”
77.
by rytill 4y ago
Do you know of examples of graphic designers who have shared their img2img workflows online?
78.
by rytill 4y ago
Do other large language models perform well on this task, e.g. EleutherAI models?
79.
by rytill 4y ago
Can you help me understand how you are imagining that? As in, you have texture images already and you want to apply them to the 3D object intelligently? Is that the case you’re talking about? Or texture generation and UV unwrapping in one?
80.
by rytill 4y ago
This is an interesting answer. What in your opinion are the main blockers to this ability?
81.
by rytill 4y ago
Google gives the following definition of apocryphal: > (of a story or statement) of doubtful authenticity, although widely circulated as being true. So it seems to be certainly apocryphal.
82.
by rytill 4y ago
What if that’s just an analogy though and it doesn’t actually apply to the real situation? The first tasks we’re going to solve are relatively easier ones, it doesn’t mean they’re the only tasks that will be solved with a technique.
83.
by rytill 4y ago
I don’t know. It seems like deep learning might take us really far. > A third possibility, which I personally have spent much of my career arguing for, aims for middle ground: “hybrid models” So this person witnessed the last 5 years of
84.
by rytill 4y ago
Because almost everyone doesn’t know what AGI is or why it would be dangerous. Among the people who could know that it’s dangerous, many of them don’t accept it because it conflicts with their existing worldview too strongly.
85.
by rytill 4y ago
Bet on your belief; you could make a lot of money: https://polymarket.com/market-group/ethereum-merge-pos
86.
by rytill 4y ago
> I believe programming, in this regard, can learn something from writing: when writing the first core of a new system, when the original creator is still alone, isolated, able to do anything, she should pretend that this first core is h
87.
by rytill 4y ago
I took that as an "obvious exaggeration", but it felt pretty out-of-place.
88.
by rytill 4y ago
Full-stack means you can contribute to code pretty much anywhere in the product’s stack. Why would you have to do CPU design and assembly programming to contribute to an average software product?
89.
by rytill 4y ago
> But as far as I can tell, we don’t seem to have any mechanism for “observing” our consciousness. That is the missing freewill component, which may have only existed in our first run. Do you think the "first run" version of yo
90.
by rytill 4y ago
What about diffusion models? (And all the other things with less publicity that aren’t that)

More ›