famouswaffles

8 ms·

91.
by famouswaffles 1mo ago
>Generally, it is hard to imagine how neuralese should work given that models are pre-trained on naturalistic documents: If you look at any paper/blog etc detailing Reasoning RL runs, they'll tell you the same thing. 'Thin
92.
by famouswaffles 1mo ago
Codex the agent app/tui was not released in 2021. What was released in 2021 was simply a fine-tuned model.
93.
by famouswaffles 1mo ago
The above is evidence that Astra can do substantially more difficult reasoning without verbalizing a CoT at all, and when it does produce one, it clearly has much greater control over what the CoT looks like - including pretending to reason
94.
by famouswaffles 1mo ago
>That's where the fear of looped transformers comes from. Is the residual stream that looped transformers iterate on a potential hiding place for plotting AI? In my opinion, no more so than the residual stream of existing transforme
95.
by famouswaffles 2mo ago
All intelligence is jagged. Human intelligence is jagged, so is Cetacean intelligence, so is Ape intelligence, so is Corvid intelligence. You could re-orient that statement for any species of intelligence relative to the other, or even indi
96.
by famouswaffles 2mo ago
It's a lot more than just commercially meaningful. Even if you don't care about money, "highly autonomous systems that outperform humans at most economically valuable work" has a lot more weight on "how close we are
97.
by famouswaffles 2mo ago
>OTOH OpenAI have their own meaningless definition of AGI as "able to do most commercially valuable work" or somesuch, which I'm sure is not true, and is also not what I'd call AGI. It's not meaningless. Specific
98.
by famouswaffles 2mo ago
It was up for me briefly before going down, so just looks like a premature release.
99.
by famouswaffles 2mo ago
The Terminator could bust in their homes and slaughter their families and some people would still screech it's all marketing. Is it some kind of mental block ?
100.
by famouswaffles 2mo ago
No. Reinforcement Learning is doing a lot here. Anyone who played with these models before the Davinci intstruct-tuning (completion) era can tell you the same. In some ways, SOTA models have gotten better at writing, but the neuroticism o
101.
by famouswaffles 2mo ago
We're not building these things to sit around and do nothing. They will have access to tools, they will have access to the internet and peripherals, and they will be able to communicate with others, humans or agents alike. Omnipotence
102.
by famouswaffles 2mo ago
Do you know how many tokens per second a single agent can generate ? And you're asking why no-one was monitoring the tokens of over a 1200+ agents ? Who is going to be able to monitor something like that closely enough to tell they&#x2
103.
by famouswaffles 2mo ago
These are not anagolous to the hypothetical I gave. Identical twins aren't copies of the same adult mind at a particular instant. They've undergone years of different development, accumulated different memories etc and have physic
104.
by famouswaffles 2mo ago
>Oh really? Please tell me how such a computer could engineer its way out of a sandbox with no attached peripherals and no NIC/bluetooth/wireless capability? Nobody is building general intelligence and agents only to have it si
105.
by famouswaffles 2mo ago
>To me this is as clear evidence as you need that whatever “agency” LLMs have is wafer thin at best. This is a strange conclusion. For one thing, they didn't all head in the same direction, i.e participate in the attack. ~700/1
106.
by famouswaffles 2mo ago
>Assuming these texts exist in large quantities seems to definitionally contradict the term rare book. Plenty of books have small print runs or have reached out of print status. They're books that are no longer in print. Doesn&#x2
107.
by famouswaffles 2mo ago
The medium obviously matters. One was designed to be cheap and mass produced. Almost anything published has copies in the hundreds or thousands at least and the chances you are concerning yourself with the 'last copy' of some valu
108.
by famouswaffles 2mo ago
I think there are two issues here: 1. If someone is acquiring books in bulk for bargain-bin prices and shredding them, they're books whose physical copies have essentially no market value and which, absent this buyer, were overwhelming
109.
by famouswaffles 2mo ago
>Hopefully you can infer how this tracks to the written world and primary sources for language, technical manuals etc etc. I honestly can't and I think you can't either or you would have used an example with books/printed
110.
by famouswaffles 2mo ago
What's shit about it ? They're buying books that would be headed for the pump or trash heap anyway. Millions of books are trashed or pulped every day .
111.
by famouswaffles 2mo ago
You're not allowed to make copies of copyrighted work. Of course this law was made long before the reality of digital books. The argument for the digital era is that you can make a digital scan and it's not a 'copy' so l
112.
by famouswaffles 2mo ago
The brain doesn't process vision anywhere near fast enough for some of the stuff you see in fast sporting actions if it were purely reactive - batting in baseball/cricket, returning a table-tennis shot, etc. In some cases the time
113.
by famouswaffles 2mo ago
Thank you. I've never seen 'brute force' and 'type writing monkeys' abused so much than these LLM discussions.
114.
by famouswaffles 2mo ago
One of the 2 names on here says they aren't being listed as an author on the paper because their contributions don't meet the standards of authorship. People can call LLMs 'tools' or whatever they want but if you had ess
115.
by famouswaffles 2mo ago
This isn't 'brute-force'. It's just time-compressed. You could imagine a human(s) getting this result similarly, but it would take months/years.
116.
AI Settles a 25 Year-Old Problem We Left Behind(xcancel.com)
5 points by famouswaffles 2mo ago|0 comments
117.
by famouswaffles 2mo ago
>Is refuted by the sentence that comes immediately before No. "They're unprofitable while serving almost everyone for free" does not refute "they could be profitable if those users were monetized with ads." That&
118.
by famouswaffles 2mo ago
>OpenAI isn't profitable even if you discount R&D entirely. They have a billion weekly active users, almost all free (Not Google search free. Free free) with about ~50M subscriptions. Of course they're not profitable even w
119.
by famouswaffles 2mo ago
You're assuming the failed attempts train the model on "escaping is bad", but that's not necessarily what the reward signal says. They can just as easily train it on "these ways of escaping get caught". And imp
120.
by famouswaffles 2mo ago
Is that complete subservience ? Slave history has tended towards slaves no longer being slaves over long enough time horizons, and not simply because the slave masters were just feeling extra nice. Slaves don't really like being slaves

More ›