Yes, but that wouldn’t work as clickbait.
4,556 karma · joined August 10, 2016
Why do you believe that?
It works for the billionaire class, why shouldn’t it work for everyone else?
This is what’s ultimately going to kill what’s happening in the US today: when people in power decide that the rule of law is just a suggestion, they’re going to find that there are people willing to use that against them in ways that they’re not going to like.
You mean the terrorists who used to work for ICE before it was defunded.
Luckily the US has a lot of prison capacity.
Why not? Our data science team uses it exclusively.
No, it's an "equilibrium" state, i.e. "a state of balance where opposing forces ... are equal" (from the MW definition.)
The general concept is that a stable system has at least one equilibrium state, in which no processes go out of control and potentially destroy the system.
A simple(?) example would be global warming - if the atmosphere keeps accumulating carbon, temperatures keep going up, the planet becomes uninhabitable, everything dies. Of course, that's subjectively an undesirable outcome for us - not an "ideal state", as you say - but the point is that any state in which no equilibrium can be reached (or at least approximated) exists is subject to such risks.
You could say that equilibrium is a necessary condition for an ideal state, but it's not sufficient. There can also be equilibrium states that are decidedly non-ideal, like the "everything dies" one I mentioned.
In this respect Gandhi was essentially on the same page as RFK Jr. He thought people should try to lead healthy lives - diet, hygiene, exercise etc. - but ultimately that nature should take its course.
The relevant MW definition for "reasoning" is: "the use of reason, especially: the drawing of inferences or conclusions through the use of reason."
And "reason" is: "the power of comprehending, inferring, or thinking especially in orderly rational ways."
Functionally speaking, i.e. in terms of observable behavior, LLMs exhibit comprehension, inferring, and reasoning. If someone wants to object to that, they'd need to explain what relevant property prevents a conclusion drawn by an LLM from being counted as involving reasoning.
(I suppose a religious or otherwise superstitious person might introduce the soul into this, but I haven't come across anyone actually willing to defend that hill.)
It's also very risky to frame it that way, because that immediately leads to the idea that mathematical endeavors should be judged and funded according to their expected utility. That's product development, not research.
The point is that working with natural language tokens is very different than tokens that represent an image.
A simple relevant example is that if you ask an LLM to write a psychological thriller about a poor former student who commits murder and deals with intense moral guilt, in classic Golden Age Russian literature style, it is unlikely to sign it with "Fyodor Dostoyevsky."
It does that when generating images because, at a high level, image generation doesn't benefit from the kind of reasoning that language generation is able to.
Redefining sycophancy as you're attempting to do would erase the distinction between models that are trained to be sycophantic via RLHF, and models that are just responding based on what was in their pretraining corpus.
> Yes or, in other words: there's sufficient contradictory data
The idea of "contradictory data" is a projection that you're imposing. (We might even call this an hallucination.) In the model weights are just data, and different responses are elicited by different prompts depending on their relationship to to the data.
> That the contradiction is beyond the LLM's scope is part of the problem.
Well yes. But the problem is in your mind and the minds of others who think like this.
This quote is a pretty solid argument that you need to understand the technology you’re trying to criticize better. This issue has nothing to do with LLMs. LLMs are not image generation models.
This is completely silly. If you don’t think LLMs can reason, you’ve either never used them to do tasks that require reasoning, or you don’t understand enough to recognize what’s involved in the responses you get.
In this case it’s clearly the latter, because you’re confusing image generation models with LLMs. There are very big differences between the two. No-one is claiming that image generation models are capable of reasoning.
Worth noting that it wasn't "those in power" who decided of their own accord that they needed to improve the situation. It was organized labor, unemployed workers, protests, farmers, and electoral politics.
"Those in power" almost never seem to make such changes of their own accord. They have to be forced. The only question is how much force is necessary.
There's no such consensus. Read Hardy's "A Mathematician's Apology" to get a better understanding.
Here's a quote which gives some idea of where Hardy was coming from: "I have never done anything 'useful'. No discovery of mine has made, or is likely to make, directly or indirectly, for good or ill, the least difference to the amenity of the world."
Hardy valued math for its beauty, elegance, and permanence, and he's far from the only mathematician who would push back on the idea that mathematics should be "useful to humanity", except perhaps in the sense that art is useful to humanity.
If you're trying to make an argumentum ad populum, well, there's a reason that's a logical fallacy.
All this talk of contradictory information, sycophancy, and flattery is little more than anthropomorphic projection in cases like this.
One prompt will elicit a response based on one set of weights that are most closely related to the prompt. A different prompt will elicit a response from a different set of weights, for the same reason.
There’s no contradiction from the point of view of the model and its responses, and no need to invoke sycophancy or flattery as a reason for the behavior. You’d get the same behavior from a model that hadn’t been RLHFd to be sycophantic.
The DSM even has to include an explicit exception to prevent the clinical definition of “delusion” from applying to religious belief. Without that ad hoc exception, religious belief would be classified as clinically delusional.
You're trying to use technology to handle difficulty with minor adversity.