ComplexSystems

8 ms·

91.
by ComplexSystems 1y ago
> Trying to eliminate cases where a stochastic model the size of an LLM gives “undesirable” or “untrue” responses seems rather odd. Why? It seems no less odd than eliminating cases where it gives "undesirable" code snippets wit
92.
by ComplexSystems 1y ago
It isn't a hallucination because that isn't how the term is defined. The term "hallucination" refers, very specifically, to "plausible but false statements generated by language models." At the end of the day,
93.
by ComplexSystems 1y ago
The RAM bandwidth is so slow on this that you can barely train or do inference or do anything on it. I think the only use case they have in mind for this is fine tuning pretrained models.
94.
by ComplexSystems 1y ago
First: while it's not technically incorrect to say that they're learning "patterns" in the training data, the word "pattern" here is extremely deep and hides a ton of detail. These aren't simple n-grams li
95.
by ComplexSystems 1y ago
This all seems totally orthogonal to the statement: "I don't get why people are so resistant to the idea that AI can prove new mathematical theorems." I don't necessarily disagree about the copyleft stuff. Transformers d
96.
by ComplexSystems 1y ago
People are resistant because: 1. There's this huge misconception that LLMs are literally just memorizing stuff and repeating patterns from their training data 2. People glamorize math and feel like advancements in it would "be AGI
97.
by ComplexSystems 1y ago
I would like to see what happens if some company devoted your resources to just training a model that is a total beast at math. Feed it a ridiculous amount of functional analysis and machine learning papers, and just make the best model pos
98.
by ComplexSystems 1y ago
I really just want a UI to integrate this into VSCode...
99.
by ComplexSystems 1y ago
Why would they leave out GPT-3 or the original ChatGPT? Bold move doing that.
100.
by ComplexSystems 1y ago
I thought this article was going to be a bunch of security theater nonsense - maybe the relatively bland title - but after reading I found it to be incredibly insightful, particularly this: > MCP discards this lesson, opting for schemale
101.
by ComplexSystems 1y ago
This is sloppy: "ChatGPT Deep Research, along with Sora by OpenAI, which can generate video, is available on the ChatGPT Plus or Pro plans. If the user asks about the GPT-4.5, o3, or o4-mini models, inform them that logged-in users can
102.
by ComplexSystems 1y ago
o3 is the best OpenAI model but it still makes tons of mistakes. It's got a very strong background in most of undergrad level math, and a decent amount of grad level machine learning stuff, but its tendency to hallucinate means it will
103.
by ComplexSystems 1y ago
I am not in biology, and this is the first time I have ever heard anyone advocate for freedom of knowledge to such an extent that we should make biological weapons recipes available. I note that other commenters above are suggesting these t
104.
by ComplexSystems 1y ago
How do you suggest they solve this problem? Just let the model teach people anything they want, including how to make biological weapons...?
105.
by ComplexSystems 1y ago
It makes too many mistakes and is just way too sloppy with math. It shouldn't be this hard to do pair-theorem-proving with it. It cannot tell the difference between a conjecture that sounds kind of vaguely plausible and something that
106.
by ComplexSystems 1y ago
That "odd one out" game is broken - it just generated 21, 14, 35, 49 and told me the right answer was "14" because it's the only number not divisible by 7. Also the prime number game generated a set of all composite
107.
by ComplexSystems 1y ago
He taught classes and and was in charge of their online deep learning materials for a while at deeplearning.mit.edu, back when the idea of universities putting materials online was still very new and cutting edge. A lot of his video lecture
108.
by ComplexSystems 1y ago
These are limits for the $200/mo plan?
109.
by ComplexSystems 1y ago
I am curious how much of this is real. I mean, it's a great story, and it fits a bunch of tropes about ol reliable systems that don't need a-changin', but it seems crazy to put any actual information about the US's nucle
110.
by ComplexSystems 1y ago
Always liked his "Werner Von Braun": https://youtu.be/QEJ9HrZq7Ro?si=RDu536cWahc2Oct7
111.
by ComplexSystems 1y ago
This has got to be one of the most loaded question I have ever seen on here. Could you perhaps be clearer with what you want to ask? It sounds like you are insinuating that the guy is a bad person in some vague, nondescript way.
112.
by ComplexSystems 1y ago
> LLMs specialize in self-apologetic catastrophe Quote of the year right there
113.
by ComplexSystems 1y ago
Good stuff. You could get much better bandwidth than this by tokenizing and using something like a Huffman or arithmetic code on token frequencies. As a simple example, if you set your tokens to be all English words - let's say there
114.
by ComplexSystems 1y ago
Is this just my device, or is there no way to use this roll-your-own SVG generator to actually roll your own? I can only pick from a tiny subset of preset colors, most of which seem super random and desaturated and not what I want. There&#x
115.
by ComplexSystems 1y ago
Is this different from regular DMT?
116.
by ComplexSystems 1y ago
I have been using this with mostly decent results. I am curious how it compares to other IDEs though - Cursor, Windsurf, Roocode, etc. Any thoughts?
117.
by ComplexSystems 1y ago
That is very different though. You are now saying that for all n, PA proves that it has a proof the Goodstein sequence terminates for that n . But that is very different from the earlier claim: "And, therefore, there must be a distinc
118.
by ComplexSystems 1y ago
We are talking past each other. I am responding to this: "If PA proves that it proves a statement, PA cannot conclude from that fact that it proves that statement." When you say "PA proves that it proves a statement," th
119.
by ComplexSystems 1y ago
If PA proves that a number exists with some mathematical property - including being a Gödel number of a proof of something - then some number with that property must exist in every model, including the standard model. So there would have to
120.
by ComplexSystems 1y ago
I am not sure I follow and it is possible I may not be on the same wavelength, but here is another thought. For any sentence S, if first-order PA (or any first-order theory) proves S, then that means that S holds true in every model of th

More ›