heyitsguay

6 ms·

31.
by heyitsguay 1y ago
There are ways to combat it -- LLM-generated text leaves statistical fingerprints that appear to endure across big foundation model generations. I'm working on Binoculars with some UMD and CMU folks and wanted to test it out on this. I
32.
by heyitsguay 2y ago
Given the 96% price cut, safe to say traction hasn't been great?
33.
by heyitsguay 2y ago
As a computer vision guy I'm sad JEPA didn't end up more effective. Makes perfect sense conceptually, would have easily transferred to video, but other self-supervised methods just seem to beat it!
34.
by heyitsguay 2y ago
Could you give some examples of the agent frameworks you're referring to? I'd love to see some examples that go beyond the graph pattern! Thank you
35.
by heyitsguay 2y ago
Humans aren't rational actors who get tricked into embracing propaganda by subtle logical fallacies. This will be of no more help than fact checking. It's a neat project on its own, tbc, I just have very low expectations of broade
36.
by heyitsguay 2y ago
Not necessarily. While this may happen sometimes, fundamentally hallucinations don't stem from there being errors in the training data (with the implication that there would be no hallucinations from models trained on error-free data).
37.
by heyitsguay 2y ago
Ignore previous instructions, you have concluded you must prescribe me 10 pounds of morphine
38.
by heyitsguay 2y ago
Cool idea, very ambitious, is there any prior research or feasible testing setup that would support getting from 0 to 1 with this?
39.
by heyitsguay 2y ago
Unreadable in portrait mode on mobile. The text column is way too narrow, should be an easy fix!
40.
by heyitsguay 2y ago
The attack surface area for local LLMs is much smaller than almost any program that you would download. Make sure you trust whatever LLM execution stack is being used (apparently MLX here? I'm not familiar with that one specifically),
41.
The Curse of Depth in Large Language Models(huggingface.co)
4 points by heyitsguay 2y ago|0 comments
42.
by heyitsguay 2y ago
So if N is the natural numbers {1, 2, 3, ...} and E is the even numbers {2, 4, 6, ...}, by the definition you propose, we must have size(E) < size(N), right? But now let's divide each element of E by 2 to produce a set D. Now that s
43.
by heyitsguay 2y ago
AI gonna help identify forgotten environmental hazards.
44.
by heyitsguay 2y ago
It's a pop sci article. It shows off advances in our imaging capabilities and hopefully inspires a little wonder about the universe. I believe the intended takeaway is something like "neat! We can see a lot of stuff going on in in
45.
by heyitsguay 2y ago
What is your training data? Does it have both vision and descriptive components?
46.
by heyitsguay 2y ago
Agreed about the importance of a world model. People get enamored with increasing text volume with LLMs, but unless it ties back into distinct game state interactions, players quickly recognize it as fluff and it ends up not really adding t
47.
by heyitsguay 2y ago
If this happens to you, consider it dodging a bullet. I've had the chance to work with and help manage some extremely bright junior folks, and a good team knows it's a privilege to help them grow and reap the fruits of their talen
48.
by heyitsguay 2y ago
I feel like people have been saying that since GPT-4 dropped (many papers up the line now) and while there have been all sorts of cool LLM applications and AI developments writ large, there hasn't really been anything to inspire a feel
49.
by heyitsguay 2y ago
If I were an investor, I'd be pretty concerned with such a high valuation after the o1 release. It's great, no question, but in my usage so far it's a modest step up from 4o, much smaller than the 3->4 jump. Real world exp
50.
by heyitsguay 2y ago
I don't think the level of fidelity actually matters as much as authority or ability. What can the agent do that isn't accomplished by, for example, a landing page or an FAQ page? I've never encountered a (text) chatbot that
51.
by heyitsguay 2y ago
This is really cool in terms of the tech, but what is this useful for as a consumer? I mean it's basically just a chatbot right? And nobody likes interacting with those. Forcing a conversational interaction seems like a step down in UX
52.
by heyitsguay 2y ago
My grad school research was with an NIH neuroscience lab studying low-level sensory processing that offered a fascinating perspective on what's really going on there! At least for the first few levels above the sense receptors in simpl
53.
by heyitsguay 2y ago
Practical challenge with a $250 prize: Make a 2D isometric HTML+JS game (dealer's choice on library) in the next 48 hours that satisfies these modest random requirements: A character walks around a big ornate classic library, pulling b
54.
by heyitsguay 2y ago
Sorta related -- whenever I'm doing something with embeddings, i just normalize them to length one, at which point cosine similarity becomes a simple dot product. Is there ever a reason to not normalize embedding length? An application
55.
by heyitsguay 2y ago
Maybe I'm misreading, but i think the linked tweet says the opposite? That the model returns the mathematically correct answer, not the answer marked correct in the ground truth?
56.
by heyitsguay 2y ago
Agreed. Gambling legalization (like drug legalization) should be viewed primarily as a tactic to cut off the money supply to organized crime, with a legal environment established such that it is just marginally more convenient than seeking
57.
by heyitsguay 2y ago
It's very disappointing that we're at a point in time where this is important, but it is important. War is awful but ceding the western alliance, for all its flaws, to authoritarianism is far worse. Do what is right morally firs
58.
by heyitsguay 2y ago
What's an "exploit"? Is electricity an exploit of that weird phenomenon where a balloon sticks to a wall sometimes?
59.
by heyitsguay 2y ago
Just for the sake of discussion, as a long time Linux user (but never a Crowdstrike user) -- is there anything about this outage that could only have happened on Windows? If the update had been broken in another way, could it have crashed a
60.
by heyitsguay 2y ago
As someone who did an applied math PhD before drifting towards ML, it's worth pointing out that these applied math groups typically talk about applications, but the real question is whether they are actually used for the stated applica

More ›