jptlnk

I cook almost every day and I would use this feature. I also think it's kind of silly for people to be so upset about what (feels to me) to just be mostly useful.

At times like these I cant help but think of Principal Skinner. Am I so out of touch? No, it's the children who are wrong!

hear hear. imo the constitution should be read as eyes open tradeoffs. the 4th amendment is an clear one and this is a great case for it: its not that flock couldnt theoretically reduce crime. it that the reduction isnt worth giving up the power to the government to do so.

I agree that the state is irresponsible here - if it's bad enough to be a huge deal when an agent does it, it's a huge deal.

With that said, I do think there's an important distinction here, which is that we (hopefully) get to _choose_ if the systems we build will pursue this kind of behavior.

I think most people would agree that a big point of social training (school, parenting, etc) is to inculcate a sense of what's acceptable and not in society. Here it seems like we're letting the agents that we've created create havoc and then we're providing a smoke screen by blaming the victim.

IMO this is a 'for whom does the bell toll' kind of collective moment and we shouldn't be laundering this kind of agentic behavior.

I'm skeptical of any claim that AI can't be conscious, but I am given to this point of view as well.

I always get the feeling that these claims are an attempt to "avoid suffering by fiat". If it can't suffer, you aren't causing suffering.

A somewhat more interesting frame (IMO) is to interrogate how much suffering we're willing to tolerate to achieve our objectives. We implicitly make these decisions all the time when it comes to something as simple as what to have for lunch.

And in a (very hypothetical) world where we found ourselves in the position of having an eloquent conversation with a vat of smallpox, I think it's probably still the right thing to pasteurize it.

Genuine counterpoint:

If a human says "I am in pain", is it just stringing together words or is it feeling pain?

If you take that seriously it's quite a tough question to unwind. See e.g. Dennett's heterophenomenology conversations.

not necessarily according to the courts. there are a number of cases where scale has been the line between permissible and not - e.g. it's fine to scrape ebay, it's not fine to build an ebay scraping business.

courts haven't always ruled that way but there's definitely precedent!

https://en.wikipedia.org/wiki/EBay_v._Bidder%27s_Edge

It's probably not 10MB, but famously the groundwork to prove the statement 1+1=2 is nearly 400 pages in to principia mathematica. That's not even proving 1+1=2, it's just the set-theoretic proofs you need to EVENTUALLY get there.

Unless I'm misunderstanding you, the existence of an objective "consciousness" property is a hotly debated question philosophically. Many great thinkers have written and discussed this question at length, and as of now the jury is very much out.

If I may, I think that's actually part of the point, and (at least for me) part of the lesson.

I read him as saying that part of the miracle to him is that he has experienced something that makes him realize that it's a lot harder than it sounds to be loving and kind with no (or few) conditions, and to open your home and life to a stranger.

For me, a lesson of this piece is actually the juxtaposition of the relative ease of -accepting- help and the strange difficulty of -offering- help. It's worth reflecting on, and imo much more relatable.

I'm reminded of a friend who talks about primary and secondary wants. He wants to eat a burger, but he wants to want to eat a salad. Maybe KK wants to want to help people, and the challenge for him is connecting the dots.

I don't think it's obvious but in my own experiments with this it's things like permissions that can get a little hairy. It's all doable and IMO preferable to have an agent-as-daemon. aider is kind of like this.

What's 'this', do you mean the command pattern being described? If so, yes - I've used it to great effect, if the code around it is designed properly. It's even amenable to evals if you can write the LLM call as a function that operates on some state:

  (document, input) -> command  
  (document, command) -> document'  
  # assert something about document' relative to document

Pioneer Square Labs (PSL) | Principal Engineer | Onsite - Seattle, WA | Full-Time | psl.com

At the PSL studio, we build startups from zero to launch. Over the past 9 years, we've validated 450+ ideas and spun out north of 40 companies.

Our core team of engineers works to validate the technical feasibility of idea/seed stage companies and build prototypes for them once we've reached conviction on the market.

We have 3 principal engineers currently and have an opening for one more.

Learn more and apply via the job posting here: https://www.psl.com/jobs/job?gh_jid=7885441002

Without imputing any actual intention to the author, I agree with your points on tone. It feels focused on optics, not outcomes.

It's one thing to say that you want to get things done. It's another to say you want to be _seen_ as someone who gets things done.

Again, I don't intend to mind read here, and I think the author actually has some really good data gathering ideas. But the language definitely smacks of political motivation, which some folks (myself included) find off-putting.

hey hn, I'm the lead engineer on this project. This was a super fun feature to build, and was incredibly challenging at a technical level.

Lots of fun stuff to figure out, from wrangling structured data, working around limited context sizes, and figuring out how to even write down sensible metrics to determine how well the models are doing.

AMA!

I built the copilot for flux.ai, which allows LLM-driven interaction with circuit schematics and datasheets.

The stack is react / cloud run / job queue / LLMs (several) / vector db.

relevant pre-print is here: https://arxiv.org/pdf/2305.08665.pdf

that'll be my evening reading :)

I haven't read the full article yet but the abstract has some hints: I think what they're doing is leveraging the fact that they know where and when the neutrinos were created to narrow down the search in the actual neutrino detector itself.

Because they can see the associated muon, they know when to look for a correlated neutrino signal.

Between that and the extremely hot source, they can get away with a relatively small active detector volume.

Again that's just based on my reading of the abstract, I need to see the full article to validate that guess.

hey, I'm the lead engineer on this project! I love that you did this :)

our experience thus far is that GPT4 definitely needs help when it comes to datasheet-like data. it's great at broad "functional" strokes, but details and particularly numerical details are not its forte.

we're working on it!