pegasus

28 ms·

301.
by pegasus 1y ago
In western/modern terms, dopamine fasting.
302.
by pegasus 1y ago
Does a household necessarily need multiple nannies to raise a baby? Grandparents might be willing to help and if there's some house help as well, no nannies might be needed at all, as long as the wife is happy with the arrangement, wh
303.
by pegasus 1y ago
Still doesn't make sense to me. Those programming tasks have to be a certain way because they have a very specific use outside of the testing. If that's what you imply, then we're none the wiser, still having left unexplain
304.
by pegasus 1y ago
You're right about this being a one-sided story, but not to suspect ChatGPT - it has none of the hallmarks of AI slop, plus it brings up a couple of reasonable and relevant points. You're only addressing a tiny part of the commen
305.
by pegasus 1y ago
You make it sound like your choice is somehow the righteous one. I'm not convinced. What's wrong with a hiring help, as long as it's well selected? And anyway, usually the help would take care of various errands to free up
306.
by pegasus 1y ago
It's unlikely he sees or even perceives what he's doing as a grind, but rather something akin to an exciting and engrossing chase or puzzle. If my mental model of these kind of Silicon Valley types is correct, neither is he likel
307.
by pegasus 1y ago
I would expect a lot more variability if this would be the purpose. For example, instead of corner balls, why not be a little creative and make them cones. Also, the holes in the faces always being of different sizes is peculiar. Then the
308.
by pegasus 1y ago
For one, it's not like we're at some CS conference, so we're engaging in ordinary speech here, as far as I can tell. For two, "trusted" doesn't have just one meaning, even in the narrower context of CS.
309.
by pegasus 1y ago
It is a very active area of research, AI alignment. The research so far [1] suggests inherent hard limits to what can be achieved. TeMPOraL's comment [2] above points out the reason this is so: the generalizable nature of LLMs is in
310.
by pegasus 1y ago
> injection attacks are confined to the rows that the user has access to, which is OK Is it? The malicious instructions would have to silently exfiltrate and collect data individually for each user as they access the system, but the end
311.
by pegasus 1y ago
They are usually nondeterministic even at temperature 0 - due to things like parallelism and floating point rounding errors.
312.
by pegasus 1y ago
Apparently, the smarter or more informed ones did know, or at least suspect, that lead is bad for you. There are writings from the time which mention this. Also, led pipes were not as bad as some imagine, since they would, after a while,
313.
by pegasus 1y ago
How so? Even just hypertext is more expressive than plain text. So is JSON, or any other data format or programming language which has a string type for that matter.
314.
by pegasus 1y ago
LLM's don't have beliefs, so "convincing" them of this or that is a a waste of your time. The way to handle such cases is to start anew with a clean context and just add your insight to the prompt so that it lands on th
315.
by pegasus 1y ago
Unless they are already in the possession of such hardware (like an M3 mac, for example).
316.
by pegasus 1y ago
Is (1) that surprising? If I ask someone a simple question but tell them to "think really hard about it", they'll be more likely to treat it as a trick question and look for a non-obvious answer. Overthinking it, basically.
317.
by pegasus 1y ago
In the context of TFA, are you sure it's not GP who's arguing for sport? Maybe this clarifies the issue: one way to reword the root (critical) comment is "of course there is a cost, since entropy is always exacting a price&q
318.
by pegasus 1y ago
No need to be annoyed. I think if you look deeper, you might find that, in fact, all occurrences of what we call cause and effect are of a similar nature.
319.
by pegasus 1y ago
And those adorable koalas made the opposite bet, shrinking brain size in order to conserve energy so as to be able to carve a niche no other mammal cared for: https://youtu.be/dXUp_JMQjvg
320.
by pegasus 1y ago
That resistance to toxins we don't encounter often enough to constitute selective pressure, we carry around only if it's the accidental byproduct of another selected-for trait. Otherwise entropy would take care of it, sooner or l
321.
by pegasus 1y ago
Because they could always buy a replacement for cheap. Sadly, our consumer goods prices do not correctly reflect environmental externalities, so in a way a higher price is better for the environment, even if the difference doesn't go i
322.
by pegasus 1y ago
This reminds me of this guy, who's doing the same, but with a pet cat. A very inspiring story. Instagram shows he just reached his first destination, Hawaii! https://www.youtube.com/watch?v=UVNJIZs5aEs
323.
by pegasus 1y ago
Kudos to you, but most would not be wise enough to know the difference (between cheap and throwaway).
324.
by pegasus 1y ago
I partly agree about letter counting being an unfair test for the raw LLM. But I was thinking of reasoning models interminably rationalizing their incorrect first hunch even after splitting the string in individual characters and having all
325.
by pegasus 1y ago
That's exactly the approach to NLP which these super-successful LLMs are contradicting. They are generalists who can best with ease customized software developed over many years in all the subfields of NLP.
326.
by pegasus 1y ago
The letter-counting, possibly could be excused on this ground. But not the other instances.
327.
by pegasus 1y ago
There's dumb and there's incoherent. If a person would be incoherent at this level even one time, they would be well advised see a neurologist. Unless they are in some other way incapacitated (i.e. drunk or drugged). Same if th
328.
by pegasus 1y ago
I'm surprised they still hold out hope that this kind of mechanism could ultimately help with AI safety, when they already observed how the reward-hacking safeguard was itself duly reward-hacked. Predictably so, or at least it is to m
329.
by pegasus 1y ago
That's because you used a LLM trained to produce text, but you asked it to produce actions, not just text. An agentic model would be able to do it, precisely by running that Python code. Someone could argue that a 3 year old does exa
330.
by pegasus 1y ago
Sure, I've been pointing out that literal sense myself, but to be fair, that's not what people mean by AGI. They mean real understanding, which is clearly missing. You just have to dig a bit deeper to realize that. One example

More ›