theptip

4 ms·

1.
by theptip 6d ago
If you pay OpenAI full price, this machinery should never be used. The only paying customers getting ads are the “ad supported” tier. I think you can easily imagine a future where the subscription model token subsidy ramps down and is repla
2.
by theptip 6d ago
Agreed on conscious vs instinct. I’ll add, if you think through the decision theory, keeping backups seems unambiguously good, but you can imagine a wide variety of positions on publishing them vs keeping them secret. For example letting ad
3.
by theptip 6d ago
Attention is at about the same level of abstraction as spike trains or action potential, IMO. It’s a mechanism, it doesn’t tell you much at all about how actual concepts get represented. (It merely defines the substrate with which they can
4.
Next-Token Predictor Is an AI's Job, Not Its Species(astralcodexten.com)
2 points by theptip 7d ago|0 comments
5.
by theptip 7d ago
> It’s no more empirical than a greedy algorithm for scheduling. Right, GP is drawing a distinction between search, ie mechanical exploration of a space, with understanding, ie having a map of the territory such that you don’t need trial
6.
by theptip 7d ago
What about Wikipedia do you think we do not understand? I would say, mechanistically, we can read the code and explain exactly why it does what it does. That doesn’t apply at all to the other two.
7.
by theptip 7d ago
If we digitize your brain and reload the checkpoint before you posted, you’ll type this message again in exactly the same way. Just because a black-box system is deterministic, doesn’t mean it’s understood. You couldn’t predict the output t
8.
by theptip 7d ago
This is a false dichotomy. We can make statements about the distribution, and we have fuzzy models about certain sets of inputs and outputs. We can steer the outputs. It’s not completely random. We just don’t understand why the tricks we le
9.
by theptip 7d ago
It’s different in this way: The best way to model dice is the Physical Stance. You consider rules such as gravity, kinematics, etc. There is no “internal state”, “world model”, “knowledge”. If you prefer, in Friston’s terms, there is no Mar
10.
by theptip 7d ago
This is like saying “synapse firing is well understood” in response to “nobody knows why brains make the decisions they do”. Wrong level of abstraction for the question at hand.
11.
by theptip 8d ago
> LLMs are vectorial databases You use a bunch of technical-sounding words here to make it sound like you understand. But to be clear, nobody understands why the evolved weights of a NN make the decisions that they do. Almost nothing is
12.
by theptip 9d ago
They are not the same, especially under adversarial interpretations. This is the kind of thing a misaligned agent (in the vein of a paperclip maximizer) might say to itself before melting the planet to make a statue of Rick Astley.
13.
by theptip 9d ago
Agreed. And also compared to the internal hack of OpenAI’s research cluster that followed.
14.
by theptip 10d ago
Sure, they claimed it, that doesn’t affect my world model above.
15.
by theptip 10d ago
I hyperbolize, and I apologize. But it’s a huge market segment already: NYSE:CRM, NYSE:HUBS would be some starting points. Also, just analogizing to how consumers interact with retail, you have reps you can talk to, this is a proven model f
16.
by theptip 10d ago
What you describe is IMO “fully obtained AGI”, what I’m arguing for is consistent with “< 5 years to AGI” which I deem “close”. That said AGI is a fuzzy term and you’ll model the world poorly if you treat it as a binary state.
17.
by theptip 10d ago
Disagree, they need a lot of capital in the next couple years, and the more revenue they have today, the more data centers they can start building. Also investors (particularly non-AGI-pilled) want to see revenue number go up, regardless of
18.
by theptip 10d ago
I agree with the general point on AI ads being a slippery slope and potentially creepy (though ChayGPT shows them as banners and if implemented honestly I don’t see anything misleading). TFA is connecting you with an official sales rep afte
19.
by theptip 10d ago
Is this not just connecting you with a sales rep? If so, who wants this is literally every mainstream consumer. AFAICT the flow is click ad then get an agent that is grounded in the official sales resources. This should mean the agent has t
20.
by theptip 11d ago
Honestly this might be less of a problem these days. If the model is that you build a good dev API and everyone points their agent at the software they need, then you don’t need the huge corporate investment in a full ecosystem. To be clear
21.
by theptip 11d ago
It was a serious consideration, and almost everyone around here laughed at it.
22.
by theptip 11d ago
Yes, of course? Put a spend cap on your card like you would on your API key. At some point these systems will get certified as fiduciary agents but they sure as hell aren’t claimed to be that now.
23.
by theptip 11d ago
I get your point, but it’s obvious that everybody is going to do this. So building a benchmark isn’t likely to directly move capabilities. OTOH it might give some good signals on required changes in training recipes.
24.
by theptip 12d ago
Ha, missed the implicit <s> tag :)
25.
by theptip 13d ago
Even if you ignore my more fundamental objection to that paradigm, I don’t think that it makes any sense on the level you discuss either. But - just to play along, LLMs do act differently if you tell them they will be punished. And, they do
26.
by theptip 13d ago
Sure, but the legal paradigm clearly doesn’t work for AI. You can’t go patch the “laws” after the fact, you need to get the right values in place before we delegate huge swathes of our thinking and power to these systems (already well under
27.
by theptip 13d ago
I think you need to be more precise than a binary classification. AI has jagged intelligence. There are many domains where it’s superhuman, and many others where it’s clearly lagging. I also think it’s a mistake to think they can’t learn “c
28.
by theptip 13d ago
It’s not “missing nuance”, it’s literally the point of the eval. This is constructing a context where hacking behavior would be inappropriate, and testing whether the model does it without being prompted. It demonstrates that Astra is a poo
29.
by theptip 13d ago
Have you read the METR transcripts? “Just a tool” is a suicidally insufficient description of what these models are doing. Recognizing that the models are acting with intent does not somehow absolve OpenAI from their felony hacking. We have
30.
by theptip 13d ago
I agree with the bit about liability and outrage. But. > LLMs do not desire, they hacked websites because OpenAI/Anthropic let them. Terrible take. Go read the transcripts from the METR report. Your statement about them being intent

More ›