HackerTrans
TopNewTrendsCommentsPastAskShowJobs

Footkerchief

no profile record

Submissions

[untitled]

1 points·by Footkerchief·2 years ago·0 comments

Departed OpenAI head of alignment: safety has taken a backseat to shiny products

twitter.com
13 points·by Footkerchief·2 years ago·7 comments

Artificial General Intelligence Is Possible and Deadly

letter.palladiummag.com
1 points·by Footkerchief·2 years ago·0 comments

comments

Footkerchief
·8 months ago·discuss
"Fine your coworkers." Looks true:

https://www.theguardian.com/world/2008/mar/19/japan
Footkerchief
·8 months ago·discuss
Exactly. Use a fancy expressive structure if you want, but don't try to abstract away the mapping between that and the general-purpose code that it relies on. "Each domain has its own rules"? How would I even know where to look for those?
Footkerchief
·12 months ago·discuss
TLDR: you can put tablets of mosquito-killing bacteria in buckets.
Footkerchief
·12 months ago·discuss
It's probably better for Wikipedia's quality if editors remain unsung, rather than it being status-conferring.
Footkerchief
·last year·discuss
Yes, these comments are necessary pushback against the habit of these disciplines to push interventions that don't work because their evidentiary standards are bad.
Footkerchief
·last year·discuss
You still use constraints even if you put all your business logic in stored procedures.
Footkerchief
·last year·discuss
When did clickbait headlines become acceptable here?
Footkerchief
·last year·discuss
Surprised to see no discussion of other data structures like dicts/maps, or arrays of arbitrary type. Hopefully they'd be a straightforward extension. IME, apps need collaborative data structures more often than they need pure collaborative text editing.

The motivating examples (update validation, partial loading, higher-level operations) are interesting, but I don't see a strong case that the reason Yjs etc. lack these features is the underlying CRDT implementation, as opposed to these features being intrinsically difficult to build.
Footkerchief
·last year·discuss
Horseshoe theory addresses this.
Footkerchief
·2 years ago·discuss
2.5 m (8 ft 2 in) in diameter:

https://en.m.wikipedia.org/wiki/M-V
Footkerchief
·2 years ago·discuss
Can you provide an example with details? Thanks!
Footkerchief
·2 years ago·discuss
I always assumed it was some kind of bacterial product from early spoilage. Nice to learn!
Footkerchief
·2 years ago·discuss
The flipside: it's equally hard for people who assume AI is safe to establish empirical criteria for safety and behavior. Neither side of the argument has a strong empirical basis, because we know of no precedent for an event like the rise of non-biological intelligence.

If AGI happens, even in retrospect, there may not be a clear line between "here is non-AGI" and "here is AGI". As far as we know, there wasn't a dividing line like this during the evolution of human intelligence.
Footkerchief
·2 years ago·discuss
LLMS don't have to be smart enough to be AGI. They just have to be smart enough to create AGI. And if creating something smarter than yourself sounds crazy, remember that we were created by simpler ancestors that we now effortlessly dominate.
Footkerchief
·2 years ago·discuss
Jan Leike, recently resigned head of alignment: "Safety culture and processes have taken a backseat to shiny products"

https://x.com/janleike/status/1791498184671605209
Footkerchief
·2 years ago·discuss
> Don’t all ISPs have retention notices? Probably not (for the reasons above), but this is not public information, and ISPs and others subject to retention notices face statutory prohibitions on disclosure.

> If all Internet access providers were subject to retention notices, wouldn’t it be easier to say that? There would be no need to dance around issues of secrecy, or explain why the list of notice recipients cannot be published. The fact that the Home Office chooses to take this approach undermines the claim that all providers have notices.

This source appears to be engaging in wishful thinking.
Footkerchief
·2 years ago·discuss
The evidence we have for safety of low-level consumption doesn't look great.

https://www.ncbi.nlm.nih.gov/pmc/articles/PMC7691614/

"The paucity of adequate studies, particularly related to long‐term use of A. muricata supplements, currently does not allow the establishment of a safe intake level."

https://www.ncbi.nlm.nih.gov/pmc/articles/PMC10092620/

"Cognitive performance assessed with the MDRS worsens above a cumulative consumption threshold of 0.2 fruit‐years of Annonaceae fruit/juice (ie, one fruit every 5 days for 1 year) or with consumption of any quantity of Annonaceae herbal tea."
Footkerchief
·2 years ago·discuss
Missing mention of pawpaw's annonacin content, which causes Parkinsonism.

https://pubmed.ncbi.nlm.nih.gov/22130466/

"Pawpaw fruit contains a high concentration of annonacin, which is toxic to cortical neurons."

https://www.mskcc.org/cancer-care/integrative-medicine/herbs...

"Pawpaw fruit contains high concentrations of annonacin, which is toxic to nerve cells. In addition there have been case reports of possibly related nerve toxicity. Therefore, chronic use should be avoided."
Footkerchief
·2 years ago·discuss
Looks like you're right -- $551 out of $842 billion, 65%, is spent on compensation.

https://www.cbo.gov/publication/59475 https://www.cbo.gov/publication/59511
Footkerchief
·2 years ago·discuss
In the case of GitHub Actions, it's made more painful by the lack of support for YAML anchors, which would provide a bare minimum of composability.

https://github.com/actions/runner/issues/1182