qnleigh

13 ms·

241.
by qnleigh 1y ago
What happens if you ask it to please continue? Does it start over?
242.
by qnleigh 1y ago
Although it may be true that LLMs will not achieve AGI in the purest sense, they have at least forced us to move a lot of goalposts. I don't know what Gary Marcus was saying a few years ago, but I think many people would have said that
243.
by qnleigh 1y ago
I've seen a similar quote attribute to Bill Gates; "We always overestimate the change that will occur in the next two years and underestimate the change that will occur in the next ten." I think about this quote a lot these d
244.
by qnleigh 1y ago
Thanks for sharing this information. Do you (or anyone else here) know if these trends might be expected based on how the treatment works? For example, given than the treatment is only injected into certain parts of the brain, could we expe
245.
by qnleigh 1y ago
Wait really?? I also wound up migrating to Fedora 75bit for basically the same reason (TopHat doesn't even support hoobastank). But then I couldn't find `file` in the specified directory. I have `library/Lib/library&#x2F
246.
by qnleigh 1y ago
This is kind of sad. I'm probably missing context here, but surely it's not necessary to rush out of a good working relationship with a good mentor. The most important things you'll learn from a good mentor can't be lear
247.
by qnleigh 1y ago
> the distinctive quality of internet indignation is unprocessed, opportunistic rage... It's particularly important to know this in a group context. When a group joins together to vent rage—because an occasion justifies it, even tho
248.
by qnleigh 1y ago
Why do so many websites disable pinch-to-zoom for mobile users? This page is full of interesting close-up photos, but at least on my Android phone, I can't zoom in to see any of the details. Who benefits from disabling this feature?
249.
by qnleigh 1y ago
> Quite literally just gemini.google.com in an iFrame. Hmm, no? It has access to all of the content of all of you're currently open tabs, and is able to parse images on web pages as well. It would be neat if it could also browse on
250.
by qnleigh 1y ago
I got sucked into reading this paper, and it has completely changed my perception of these scams. They are incredibly thorough and realistic. Here are some quotes that drove this home for me. "The bond phase showed that scammers had tr
251.
AI-generated medical data can sidestep usual ethics review, universities say(nature.com)
8 points by qnleigh 1y ago|1 comments
252.
by qnleigh 1y ago
Yikes! How much does it cost you per product at this point??
253.
by qnleigh 1y ago
> when a good algorithm that is known to exist Sure, if a good algorithm exists and is simple to implement, then go for it. But if it is non-trivial, then you have to make a judgement call whether it is worth the trouble to solve in a mo
254.
by qnleigh 1y ago
Yes, especially if the interviewee said something like 'this may not be asymptomatically optimal, but if it's not a known bottleneck, then I might start with constraint solver to get something working quickly and then profile late
255.
by qnleigh 1y ago
I agree with the other comments here that using a constraint solver defeats the purpose of the interview. But this seems like a good case for learning how to use a constraint solver! Instead of spending hours coding a custom solution to a t
256.
by qnleigh 1y ago
It could go both ways though. Phones get more heavy use, and foldables will probably always be more fragile than iPads, so they might wind up with more customers replacing a more expensive device more often. iPads last a really long time, w
257.
by qnleigh 1y ago
Nit of a nit; the energy might take that long, but the photons that reach us on Earth are not directly created by the nuclear fusion reactions in the sun's core. Fusion creates high-velocity nucliei and other particles, but not visible
258.
by qnleigh 1y ago
> designed to look correct more than it's trying to actually be correct This might not quite be true, strictly speaking, but a very similar statement definitely is. LLMs are highly prone to hallucinations, a term you've probabl
259.
by qnleigh 1y ago
> Structural load calculations for a post in a house You're relying on ChatGPT for this? How do you check the result? That sounds kind of dangerous...
260.
by qnleigh 1y ago
These were intrusions into private companies.
261.
by qnleigh 1y ago
This is a great analogy. There's a wider Fermi paradox here regarding business productivity. Where's the 10x economic output?
262.
by qnleigh 1y ago
I suspect not literally this, but the point is one doesn't need to treat reaching out to an old friend with overtures and formalities. Probably you need a bit more context if you haven't seen them in years, but you don't have
263.
by qnleigh 1y ago
If fine-tuning for alignment is so fragile, I really don't understand how we will prevent extremely dangerous model behavior even a few years from now. It always seemed unlikely to keep a model aligned even if bad actors are allowed to
264.
by qnleigh 1y ago
Though it's not obvious to me if you get this association from raw training, or if some of this 'emergent misalignment' is actually a result of prior fine-tuning for safety. It would be really surprising for a raw model that
265.
by qnleigh 1y ago
The article here is about a specific type of misalignment wherein the model starts exhibiting a wide range of undesired behaviors after being fine-tuned to exhibit a specific one. They are calling this 'emergent misalignment.' It&
266.
by qnleigh 1y ago
I see. It'll be interesting to see how much on-device models take off for consumers, when off-device models will be so much more capable. In the past, the average consumer has typically been happy to trade privacy for better products,
267.
by qnleigh 1y ago
> when they finally execute on the blindingly obvious strategic play that they are naturally positioned for What's that? It's not obvious to me, anyway.
268.
by qnleigh 1y ago
Can someone with an ultra subscription ask it how many rs are in the word strawberry? And report back how long it takes to answer?
269.
by qnleigh 1y ago
Something like this does seem like the only viable option. I wonder how much it will actually happen.
270.
by qnleigh 1y ago
3 months doesn't seem like enough advanced warning before deleting essential data like this. What if someone is hiking the Pacific Coast Trail right now? Or is the recovering from a serious medical event? Or just doesn't use the a

More ›