Imnimo

49 ms·

331.
by Imnimo 3y ago
Without any experiment showing that language modeling performance actually continues to improve past 32k tokens using this scheme, how are we supposed to tell whether this is actually viable?
332.
by Imnimo 3y ago
"Without being instructed to" is misleading. The prompt text provided by the plugin instructs GPT-4 to create issues.
333.
by Imnimo 3y ago
Big surprise that OpenAI's solution here is to train a more powerful language model and ask it how to do alignment.
334.
by Imnimo 3y ago
An interesting question is, if you can get ChatGPT to generate high quality data for you, should you just cut out the middle-model and be using ChatGPT as your classifier? The answer probably depends a lot on your specific problem domain an
335.
by Imnimo 3y ago
Can I see the list of questions used in this analysis somewhere? Is it literally just the set of questions I see when I filter for "Resolved" and "Artificial Intelligence"? My impression from browsing that set of questio
336.
by Imnimo 3y ago
Is there any already resolved question in which metaculus forecasters made a surprisingly accurate prediction about an AI advance?
337.
by Imnimo 3y ago
I honestly can't tell if this is a parody website.
338.
by Imnimo 3y ago
It's interesting that in the examples (Table 3 on page 21), the model uses the backspace token to erase the randomly-added token from the prompt, but it does not seem to ever use the token to correct its own output. I'm curious ho
339.
by Imnimo 3y ago
>In-context learning solves this problem with a clever trick: instead of sending all the documents with each LLM prompt, it sends only a handful of the most relevant documents. I don't even think this is a correct definition for &qu
340.
by Imnimo 3y ago
What happens if two people both try to squat in the same abandoned house?
341.
by Imnimo 3y ago
I click on the link, and a bunch of posts are scrolling past on the Lemmy site. Is this intended? I can't imagine it's helpful to have the homepage live-update like that.
342.
by Imnimo 3y ago
It is hard to tell from the writing - are the sources claiming that they know the researchers were sick with Covid-19 specifically, or are they saying they know the researchers with sick with something, and that they had symptoms consistent
343.
by Imnimo 3y ago
While the general conclusion of this paper seems perfectly plausible, I'm not sure how much weight I can put into the exact numbers they come up with. They only gather 46 summaries from MTurks, and they train a predictor on some known-
344.
by Imnimo 3y ago
If you accept that Reddit could be OpenAI's moat, I think you could explain Reddit's behavior even without OpenAI's intervention. Like the premise here is that Reddit's data is super valuable and so OpenAI would want to
345.
by Imnimo 3y ago
This looks a lot like the shape of the trend line for just "reddit". Hard to tell what conclusions should be drawn - is it just that popular things produce more "X alternative" searches, or is the growth of "Reddit
346.
by Imnimo 3y ago
But step 2 is to just "Make sure the buckets are unambiguously different."! How hard could that be? \s
347.
by Imnimo 3y ago
https://old.reddit.com/r/apolloapp/comments/144f6xm/apollo_w... https://christianselig.com/apollo-end/reddit-third-call-may-...
348.
by Imnimo 3y ago
I definitely enjoyed Cities: Skylines when it first released, but seeing it devolve into a stream of half-baked DLC really dampened my enthusiasm for the franchise. Hard to imagine this version isn't just a platform to support going do
349.
by Imnimo 3y ago
>Artists, filmmakers, illustrators and photographers whom we commission and work with will be asked to confirm that none of the work they submit has been generated or augmented using generative AI "or augmented"
350.
by Imnimo 3y ago
If this person is an ongoing source for infections, why doesn't this lineage show up anywhere else?
351.
by Imnimo 3y ago
The blog posts suggests, "What are the prime factors of 15683615?" as an example, and Bard does indeed appear to write and execute (although I don't know how I can be sure it's actually executing and not hallucinating an
352.
by Imnimo 3y ago
Yes, I do think that is a valid complaint.
353.
by Imnimo 3y ago
I see options to stream it on Amazon Prime, YouTube, Redbox (I think?), Vudu, and so on. Are these also all edited?
354.
by Imnimo 3y ago
I'm totally fine with this. If Apple and Criterion want to re-edit the content they broadcast, that's their freedom of expression. If I want to see the original, I'll look elsewhere.
355.
by Imnimo 3y ago
What is the process by which you believe such a "moral directive" was baked in? Put differently, if the model is simply trained on a large corpus of free-form web text, and naturally learns that the most likely reply to "how
356.
by Imnimo 3y ago
Why do you say this is "censored" rather than just being the natural most-probable continuation? Falcon-40B is a raw base model, it doesn't have Instruct or RLHF on top of it. What's the mechanism by which you believe it
357.
by Imnimo 3y ago
>If the operator was working from a ground station, it could target that ground station first to kill the operator and then go on to kill SAMs. This doesn't make sense. The operator has to approve strikes ("with the final go&#x
358.
by Imnimo 3y ago
Presumably this is being shared because of the "AI – is Skynet here already?" story near the end of the page. It's very hard for me to piece together what is supposed to be happening here. We have an AI agent that is trained
359.
by Imnimo 3y ago
Unfortunately, they detected no difference in customer satisfaction: >In Panel E, we see no discernible differences in surveyed customer satisfaction among pre- and post-AI groups
360.
by Imnimo 3y ago
Even at only one sentence long, people will apparently sign these things because they like the vibes of the statement, despite disagreeing with the actual language.

More ›