Imnimo

44 ms·

541.
by Imnimo 4y ago
I will certainly grant that Elon spent $44 billion for his new toy and so he can do whatever he wants with it, but like...c'mon. At some point, I hope people who were gung-ho about Elon Twitter will re-evaluate their priors.
542.
by Imnimo 4y ago
The way I think of it is that there is some hypothetical continuous diffusion process leading from the original noise sample to the final output. We could get a very good estimate of this trajectory if we took a very large number of very s
543.
by Imnimo 4y ago
This Twitter Files charade has long since crossed the line into being embarrassing for Taibbi. I'm surprised he's still digging.
544.
by Imnimo 4y ago
I'm not sure I buy this line of reasoning about "inputs" rather than "outputs". If there is some prohibition about using an image as an input, regardless of whether any vestige of that image exists in an output, do
545.
by Imnimo 4y ago
If you create a trademark-violating image using an AI model does that demonstrate anything more than that that particular image is violating? Like it's also violating if I hand draw those images, the fact that they're AI-generate
546.
by Imnimo 4y ago
Remember a few days ago when the so-called Twitter Files were supposed to convince us that moderation decisions at old Twitter weren't up to snuff? Maybe Bari Weiss can use her unfettered access to tell us what went into this new deci
547.
by Imnimo 4y ago
Out of curiosity, did Binance's recent faux-audit cover USDC reserves?
548.
by Imnimo 4y ago
And he'll be testifying again on Tuesday.
549.
by Imnimo 4y ago
The picture at the front of the article is SBF testifying in a house hearing. That's not "silent".
550.
by Imnimo 4y ago
Cool idea. My only complaint is that I found the UI for uploading screenshots pretty cumbersome, especially since they weren't initially organized in the order in which I had uploaded them. The small "move up"/"mo
551.
by Imnimo 4y ago
>In its defense, Tesla lawyers said that “mere failure to realize a long-term, aspirational goal is not fraud.” I wonder what percentage of FSD purchasers believed that they were paying $15k for a "long-term, aspirational goal"
552.
by Imnimo 4y ago
So far not very impressed. https://beta.sayhello.so/search?q=how+can+i+use+a+list+of+st... None of these results are really even on topic, let alone helpful.
553.
by Imnimo 4y ago
I would find this a lot more convincing if they had put a customer with an underspecified problem in front of ChatGPT and seen how well or poorly it could nail down the true requirements. I would not be shocked at all to learn that ChatGPT
554.
by Imnimo 4y ago
Before you get to that question, you'd want to first answer the question of whether the deaths of the animals were actually necessary to make progress in the first place. Just because Neuralink has killed 1500 animals doesn't aut
555.
by Imnimo 4y ago
It's interesting that OpenAI slow-rolled the release of some of their other projects due to a worry that they'd cause harm in the wrong hands, but seems to have just let ChatGPT rip. I don't think "makes it easy to poll
556.
by Imnimo 4y ago
It is not at all clear that ChatGPT should follow an internally consistent political philosophy. Just because it answers "agree" when asked a specific question does not mean that it will display that same agreement in it's o
557.
by Imnimo 4y ago
I disagree that that is the only way of constructing a study which is effective in investigating this question. I also don't accept that the fact an RCT is not viable would somehow imply that we should draw out conclusions from studie
558.
by Imnimo 4y ago
How do you decide whether the article too long to fit in ChatGPT's context window?
559.
by Imnimo 4y ago
I clicked on the first study in the linked blog post. Here is what it says: >We found that active smoking was associated with lower odds of having a positive test result. There are several plausible reasons for this result. Active smoki
560.
by Imnimo 4y ago
A graph showing me that SAT scores are highly correlated with getting PhDs is not very convincing. You need a good SAT score to get into many top colleges, and you need to have gone to a good college to get into many PhD programs. They&#x
561.
by Imnimo 4y ago
Given a prompt or partial output, human continuations follow some unknown probability distribution over the set of possible next words. Suppose that ChatGPT has perfectly learned this distribution - it can correctly tell you the probabil
562.
by Imnimo 4y ago
It's not clear to me that the only way to parse/process/generate/transform code is to have a correct mathematical model of the code's behavior. Even human programmers sometimes analyze code based on an imprecise ge
563.
by Imnimo 4y ago
How does the failure to understand elementary math demonstrate that it's copy-pasting? If it were copy-pasting, wouldn't you expect it to be able to handle simple math, since there's plenty correct simple math text available
564.
by Imnimo 4y ago
I tried this prompt: >The Elves take turns writing down the number of Calories contained by the various meals, snacks, rations, etc. that they've brought with them, one item per line. Each Elf separates their own inventory from the
565.
by Imnimo 4y ago
>First, none of our policies have changed. Our approach to policy enforcement will rely more heavily on de-amplification of violative content: freedom of speech, but not freedom of reach. Sounds like a change in policy to me.
566.
by Imnimo 4y ago
I would say that training implies an update of the weights, at a minimum. I would accept fine-tuning GPT-3 as training, but not prompting. I think that "chat bot" is less objectionable, but it presents a bit of a misleading pictu
567.
by Imnimo 4y ago
I really dislike the trend of using the phrases like "I trained an AI chatbot" to mean "I prompted an off-the-shelf language model".
568.
by Imnimo 4y ago
If it were the case that 538 were better than the betting markets, then obviously the bettors would just start following 538's odds until the gap closed. Like if you believe betting markets work at all, you should have always expected
569.
by Imnimo 4y ago
Well, but the metric in the study doesn't count those anyway. It's just about gasoline used by cars, and they get their number by considering how much gas would have been used if all the miles driven by EV had instead been driven
570.
by Imnimo 4y ago
What stops the owner of a mastodon server from performing "data mining"?

More ›