addag

6 ms·

1.
by addag 5d ago
Well not only research, but also a lot of other aspects of life; it is way easier to interact with something that has always an answer is polite whatever tell them.
2.
by addag 5d ago
Technically text generation was there way before self driving (e.g markov chain generators are there from the 1990s). The hard thing was to make it sound smart though.
3.
by addag 12d ago
I'm new to the mechanistic interpretability field, but from what I have read insofar, a lot of papers have relied on ablation/causal interventions to prove the faithfulness of their models. Do you have a simple explanation of why
4.
by addag 13d ago
Is it assumed that the total consumption remains the same (i.e there is an increase in consumption in clean energy when there is a decrease in less clean ones)? Otherwise, there would be an obvious solution which is to increase the price al
5.
by addag 13d ago
That's why you need those 5 dice to decide it.
6.
by addag 13d ago
Crazy hack considering the order of magnitude...
7.
by addag 13d ago
I thought the same when I saw that the financial industry was hiring FPGA people for low-latency algorithms. My understanding (as a non-FPGA expert) is that currently FPGA beats generic hardware (CPU,GPU) for "small size algorithm&quot
8.
by addag 13d ago
Sure they "don't go rogue" as if they are doing actions maliciously. Instead there is an emergent behavior from a swarm, that is unpredictable and can lead to unintended adverse outcome. From an AI safety practical standpoint
9.
by addag 13d ago
I'd say it is because of the time factor. It is one thing to have a lot of money, it is another thing to have robust systems that have been developed and tested for years. Money can "buy development time" only up to a certain
10.
by addag 13d ago
I hope you are sarcastic, right?
11.
by addag 14d ago
It seems like it is the bitter lesson of the emergence of intelligence.
12.
by addag 14d ago
The whole history still goes through their servers, though.
13.
by addag 14d ago
And even if they are not doing it right now, they probably keep the history available for future training, "just in case".
14.
by addag 14d ago
The weird similarity between the engineers make it strangely look like it is AI generated (although it is not).
15.
by addag 14d ago
My physics teacher takes the spherical roast beef as an example for heat convection.
16.
by addag 14d ago
I'm wondering if it can be used as a low-discrepancy sequence
17.
by addag 14d ago
From the abstract "Further, our symbolic approximation allows us to modify an LLM's behavior in targeted ways via precise interventions on its internal representations [...]". If this is true and easily computable, this might
18.
by addag 14d ago
While I agree with this statement, one could argue that what's important is not much the support but the emergent properties - in the same way that a wave is still a wave whether it is in the water or in the air. Thus putting things si
19.
by addag 16d ago
In any cases, it is a strong evidence that the freezer can be remotely controlled globally, and this is in itself a source of concerns (what other electrical equipment can be controlled remotely, especially on such critical features?)
20.
by addag 16d ago
Until recently, I would have Occam's razor'd this and thought of a misconfiguration error from the military. Now, considering how the OpenAI/HF hack emerged out of nowhere, it might be possible that groups of AI agents might
21.
by addag 16d ago
Isn't sliding-window a part of linear attention methods?
22.
by addag 16d ago
Tbh, I won't be surprised that 1 is true; it does not look that far fetched. Organizing scarcity is already a well-established strategy in the marketing playbook; there are examples of other big organization playing a way longer game t
23.
by addag 17d ago
This behavior seems to happen only when there is no more words in the list that can complete the current string. Otherwise, it shows only existing domains. I guess this behavior helps to autocomplete the domain extension for less-known doma
24.
by addag 17d ago
For the nonexisting domains, it seems like it only autocomplete with the possible domain extensions, (e.g .com, .org, ....) as the search list is non-exhaustive. But it could indeed be improved by not sending autocomplete requests anymore.
25.
by addag 17d ago
Peak Home Alone behavior right here.
26.
by addag 20d ago
Cutting hair. I think barber/hairdresser will be one of the last jobs to fall.
27.
by addag 21d ago
I'm wondering if there is a way to watermark those data to retrieve if they were used as training for generative models (obviously this is not a trivial task, and would need statistical instead of deterministic proofs).
28.
by addag 21d ago
That specific example might be relevant to the case "I don't see any acquisition being good for the users", but not for the specific case of the HuggingFace buyout. Workflow was already operating within the bounds of the clo
29.
by addag 21d ago
Interesting ideas, maybe the real open alternative will indeed be torrents.