lukev

12 ms·

181.
by lukev 1y ago
There's multiple kinds of novelty. Remixing arbitrary stuff is a strength of LLMs (has been ever since GPT-2, actually... "Write a shakespearean sonnet but talk like a pirate.") Many (but not all) coding tasks fall into this
182.
by lukev 1y ago
This says more about Kalanick than it does about LLMs.
183.
by lukev 1y ago
This is precisely the wrong way to think about LLMs. LLMs are never going to have fact retrieval as a strength. Transformer models don't store their training data: they are categorically incapable of telling you where a fact comes
184.
by lukev 1y ago
As has been said, actual evals are needed here. Anecdotally, the worst and most common failure mode of an agent is when an agent starts spinning its wheels and unproductively trying to fix some error and failing, iterating wildly, eventuall
185.
by lukev 1y ago
Classification (as used by real science) is descriptive, which is the opposite of normative.
186.
by lukev 1y ago
> healthy You're just using more normative words, implying that you can tell what someone "really" is aside from literally any definition you can articulate, since we've established that gamete size does not apply in
187.
by lukev 1y ago
That sound you hear is literally every anthropologist laughing hysterically.
188.
by lukev 1y ago
Should produce... given what definition? I thought gamete size was definitive? Look, I get what you're saying. Such cases aren't "normal". The claim that "humans have two biological sexes" is true in same way
189.
by lukev 1y ago
It's been really funny watching this definition become the main right wing talking point, as every other definition they've attempted has been shot down conclusively. By this logic, individuals like those mentioned in the article
190.
by lukev 1y ago
There are plenty of cultures around the world that conceptualize gender differently than we do, as well. Including many where there are options where individuals can choose their gender, in various circumstances or for various purposes. So
191.
by lukev 1y ago
The big players actually like this sort of thing, since it stifles any nascent competition. Sure, it hurts margins a bit, but also really entrenches their position.
192.
by lukev 1y ago
This presumes either of those things are damaging in the first place. There's a difference between fictional depictions and reality. The argument is that if we can understand and accept that difference when it comes to violence, we can
193.
by lukev 1y ago
Oh this brings me back. My first real coding job in 2003 was with a startup building dynamic user interfaces with XML, rendered with XSLT from the underlying data. It wasn't perfect, but it did feel correct , in the sense that a UX sh
194.
by lukev 1y ago
Have fun! I'm going to keep trying to build new things that are more excellent and original than the median training data.
195.
by lukev 1y ago
At least in this scenario it cannot utilize CoT to enhance its non-aligned output, and most recent model improvements have been due to CoT... unclear how "smart" a llm can get without it, because they're the only way it can a
196.
by lukev 1y ago
That's fundamentally not a language processing task. That's a decision making task with a huge impact on individual scientists and the scientific community. Not something that should be delegated to a machine, no matter how soph
197.
by lukev 1y ago
I'd like to believe you, but there's credible evidence that (e.g) DOGE has been using LLMs to cut funding for NSF or HHS using prompts in the vein of "is this grant woke." Which is obviously stupid. So if stupid people a
198.
by lukev 1y ago
AMD has never implemented the CUDA API. And not for technical reasons.
199.
by lukev 1y ago
There's a massive financial incentive (billions) to allow existing CUDA code to run on non-NVidia hardware. Not saying it's easy, but is implementation difficulty really the blocker?
200.
by lukev 1y ago
So to make sure I understand, this would mean: 1. Programs built against MLX -> Can take advantage of CUDA-enabled chips but not: 2. CUDA programs -> Can now run on Apple Silicon. Because the #2 would be a copyright violation (specifi
201.
by lukev 1y ago
This effect is well known but not well documented so far, so great job here. It's actually even more significant than it's possible to benchmark easily (though I'm glad this paper has done so.) Truly useful LLM applications l
202.
by lukev 1y ago
This brings up a tangential question for me. Clearly, companies view the context fed to these tools as valuable. And it certainly has value in the abstract, as information about how they're being used or could be improved. But is it re
203.
by lukev 1y ago
So, this is cool and a per-user memory is obviously relevant for effective LLM use. And major props for the temporal focus. However, keeping a tight, constrained context turns out to actually be pretty important for correct LLM results ( ht
204.
by lukev 1y ago
Did you discover it from first principles by yourself because it's a natural conclusion anyone would reach if they ponder the subject? Or because western culture reflects this theme continuously through all the culture and media you&#x
205.
by lukev 1y ago
Yeah this really hurts. If your goal is to precisely mark up a document with some structural elements, XML is strictly superior to Markdown. The fact that someone would go to all the work to build a model to extract the structure of documen
206.
by lukev 1y ago
Very interesting to consider, and there's probably some truth here. Correlating "rich person behavior" with the drug of choice in various decades certainly feels right, though I'm not sure you can actually draw a causal
207.
by lukev 1y ago
Genuine question: have there been any successful lawsuits on the basis of "false advertising" in recent times? It seems so prevalent everywhere, I'm really curious if there's any repercussions for it (no matter how egreg
208.
by lukev 1y ago
Disagree. I actually think no evil person has the thought process of "this is bad, but I will personally benefit, therefore I will do it." The thought process is always "This is for the greater good, for my country/famil
209.
by lukev 1y ago
Absolutely not. I would argue the defining characteristic of being evil is being absolutely convinced you are doing good, to the point of ignoring the protestations or feelings of others.
210.
by lukev 1y ago
I dunno. I feel like a lawsuit would need to demonstrate damages. Real-world productivity improvements due to AI in terms of actual metrics or financial outcomes remain stubbornly undemonstrated.

More ›