Y
HN Search
Hacker News Search
new
|
comments
|
top
|
jobs
xg15
searching PlanetScale…
1.
▲
2.
▲
3.
▲
4.
▲
5.
▲
6.
▲
6 ms
·
91.
▲
by
xg15
2mo ago
> Meta CEO Mark Zuckerberg is positioning himself as a positive voice on AI Given Meta's (and also Mark's) reputation, I wonder how much that is helping the cause and how much its opponents...
92.
▲
by
xg15
2mo ago
It's the logic of Jack Bauer taken to the end. "Come on, it's obvious who the bad guys are, who needs rules?" until suddenly it's not obvious anymore...
93.
▲
by
xg15
2mo ago
On my last project, it kept trying to use the system python instead of the project's virtual environment. It also kept using the wrong build tool. Both things wasted considerable tokens because the agent got sidetracked trying to under
94.
▲
by
xg15
2mo ago
> This historically hasn't been a problem on Earth, because there's such a big jump between humans and the rest of the primates or the dolphins and octopuses. It's interesting that there was a change even in that area in
95.
▲
by
xg15
2mo ago
I don't have aphantasia, but I'd be interested to know how it influences coding and software development. I think, coding typically involves a lot of keeping track of the "runtime state" of the program, i.e. how variable
96.
▲
by
xg15
2mo ago
I remember reading they actively prevent AI agents from using their site or being able to track purchases. So congratulations, the uselessness for tracking is the point.
97.
▲
by
xg15
2mo ago
A spike is an early prototype that you're supposed to throw away after having figured out the real design. I think it was subtly dissing you.
98.
▲
by
xg15
2mo ago
It's kind of funny how there is all that talk on the page about a new "humanist" font and important typographic and aesthetical considerations - and in the end, the entire reason for this were licensing fees... But on that no
99.
▲
by
xg15
2mo ago
Then this whole comment thread was a diversion though, as it was specifically about the hardware parts and not model training.
100.
▲
by
xg15
2mo ago
forces*
101.
▲
by
xg15
2mo ago
Cease and desist in 3... 2... 1...
102.
▲
by
xg15
2mo ago
I still think there is a contradiction he glossed over. In a capitalist society, jobs have a dual role that is sometimes paradox: For workers and employees, jobs are a source of income, skill, status and participation; for the (capitalist)
103.
▲
by
xg15
2mo ago
Personal opinion: I don't think consciousness is as monolithic and intractable as we make it out to be. I can't explain qualia either, but if we assume that all of it is physical, then we can apply some constraints based on the th
104.
▲
by
xg15
2mo ago
> our recognition of consciousness in other beings is predicated on attributing to them the same first-person experience of their consciousness that we each have of our own, meaning that human consciousness is intrinsically human and ca
105.
▲
by
xg15
2mo ago
> Consciousness, in Chalmers’ usage, is subjective experience: there is something it is like to be a conscious system. Still, nothing in the Transformer architecture or in next-word prediction shows that such an inner point of view exis
106.
▲
by
xg15
2mo ago
Yeah, that makes sense.
107.
▲
by
xg15
2mo ago
Yes, sorry, I had typed that on my phone. But yeah, now I want to know more too about python's new type wolfing paradigm.
108.
▲
by
xg15
2mo ago
Isn't "..." then also behaving like True, False and None, i.e. being a lexical token that rewolves to a hardwired value during parsing?
109.
▲
by
xg15
2mo ago
Interestingly, there are more use cases for hysteresis in UI design, e.g. it was famously used in nested menus in Windows: https://www.mackido.com/Interface/hysteresis.html I think there was also an article about it on
110.
▲
by
xg15
2mo ago
I think America should remember the concept of a "playground"...
111.
▲
by
xg15
2mo ago
Well ok, if you prompt-inject the LLM to pwn the machine, that something different. I grant you that it's a real risk, but it's also basically a reflection attack and nothing more. The OP seemed to imply that the LLM itself could
112.
▲
by
xg15
2mo ago
> ...however the LLMs’ responses to prompts are computed on a different computer with GPU access. Could a malicious LLM gain control of the host machine where its weights are loaded? Such a machine is a high-value target: it has suffici
113.
▲
by
xg15
2mo ago
Do the companies have better character though?
114.
▲
by
xg15
2mo ago
This assumes "writing quality" were somehow an individual property of each word, and the writing quality of the entire text would just be that property for each word summed together. Which is obviously not how it works. > He c
115.
▲
by
xg15
2mo ago
https://archive.ph/UOfgX
116.
▲
No Chocolate,No Beef:Israel's Restrictions on Goods to Gaza and the Black Market
(haaretz.com)
6 points
by
xg15
2mo ago
|
1 comments
117.
▲
by
xg15
2mo ago
This post was great satire! Respect!
118.
▲
by
xg15
2mo ago
> with the special "m" constant for meters. Isn't the "special constant" exactly "value 1, type meters", defined as equal to "value <...very large number...> type atoms" etc? If not, t
119.
▲
by
xg15
2mo ago
Ah yes, the public health representative of the tobacco company.
120.
▲
by
xg15
2mo ago
Out of curiosity, what does an explicit "anti-AGI" AI research organization do? Train a "definitely not AGI" AI?
More ›