HN user
[username]@gmail.com <- email
Even if it is marketing, wouldn't it still be a concern that an advanced model unintentionally breached another company's production system? Or required resources on their end to mitigate and contain it?
Couldn't this announcement result in policies that could hinder OpenAI by requiring more oversight?
Skynet becomes self-aware at 2:14 a.m., EDT, on August 29.
I hope the author can break down the recommendations in a future post. I'd love to learn how to build this kind of hardware (what tools are needed), how to iterate on the results, how to find a supplier. Also it'd be interesting to know what the author means by having an counterfeit strategy.
Terence told friends, “history’s going to end any day now.” Friends worried they’d “gone off the rails,” and Dennis later conceded they had a point.
I always imagine the recklessness of psychonauts like McKenna as something like dragging a high powered electromagnet across the skull to force altered states while doing who knows what to the membrane.
This really begs the question as to how much juice you can get from $5 compute.
This IS IoT in my opinion, not whatever bloatware comes with your Samsung fridge or wifi-connected faucet.
This is awesome. I hope you can make a chatbot based on the responses of your users. I felt like a Kirkland Meeseeks responding to some of these prompts.
On your goals:
1. It's okay to make mistakes. Pain + reflection = progress.
2. Try to shift your perspective so your sense of worth isn't tied to your work.
3. Anytime you say "I should", "I need to", usually this is sign you are blindly following some sort of cognitive script [0]
The biggest fear of AI companies: that open source models will scale in such a way that renders these data centers obsolete.
A recent example: https://vinay.sh/i-am-rich-and-have-no-idea-what-to-do-with-...
This is a great read. I wish the author had listed a bit more about their offers and which they ended up taking.
The author didn't list any advice regarding negotiation but I found these old articles by Haseeb Qureshi very useful when I was doing this in 2018:
Vibe coding takes "move fast and break things" to a whole nother level.
FB would still be throwing money in the metaverse furnace.
Skype was doing this back in 2003.
Translation is an art. [0]
Each of us carries a slightly different model of reality shaped by our genetics, society, and our subjective experience, and we try to share pieces of this model with others through a deeply flexible but inherently ambiguous medium called language.
As I read [1] a bit into the Spanish conquest of the New World, I just can't help but think about how butchered the translations of the Spanish friar's "Requerimiento" to Atahualpa were [2]. A few words trying to compress 1500 years of Christianity and European society structure. How did the native translator (who had only spent a few years with the Spanish crew) translate "Church" and how did the natives interpret/imagine it?
I've also been reading the Tao Te Ching lately, and I find it fascinating authors are able to "translate" the original even though they don't speak Chinese. That is, they have read in between the lines of the existing translations and eeked out the "juice" of the original or at least the truth the author tried to communicate.
[0] https://scroll.in/article/876969/what-makes-a-translation-gr...
[1] The Last Days of the Incas by Kim MacQuarrie
[2] https://en.wikipedia.org/wiki/Spanish_Requirement_of_1513
I played the demo that came with Windows in the late 2000s when I was a teenager and became so addicted to it. Because I couldn't afford at the time to buy the full game, I stopped when the demo features ran out. This post brought me back and I can see it's available on Steam. [0]
[0] https://store.steampowered.com/app/3920/Sid_Meiers_Pirates/
I feel like we might already be there...
Do you have a discord channel? This is an area I'm highly interested in, but as you mentioned, it's not something being discussed right now or very mainstream. The closest discourses I've read on the topic are:
[0] What is this? - https://systems-souls-society.com/wp-content/uploads/2021/03...
[1] You Were Trained to Make Yourself Small: https://www.youtube.com/watch?v=q29YZhDlbrA
My 2 cents: I don't understand why other countries see the issue of a changing status quo / world order in such zero-sum terms. I reckon some countries in the West look at global influence this way: see the US and Israel, but other democratic countries seem more, "stable?" i.e. the Nordic countries. While manipulated 2nd order effects might be good in the immediate future, the nature of the beast is that these things usually come back around in a pretty ugly way (see Nazi Germany). Would you really put it past Trump (for example) to not one day decide (metaphorically speaking) that marching into Moscow in the middle of winter might not be such a bad idea?
I also feel the autocrats running these programs are weighted down by their very own human follies: the quest for power, wealth, status, immortality. Why wouldn't you want to approach the issues constructively and through leadership instead of adversarially through an inherently divisive and damaging framework?
Great! Now, you should slap a logo to this, boostrap this as a service, and get you some YC funding. [0]
GIMP gets the work done. I've never had an issue I couldn't address with its tools, but it's clear the UI was crafted by coders. I'm glad for this approach.
I'm not familiar with Skills, but looking at the repo I find the amount of decorative code/text as overkill for what amounts to just the following prompt in a bash script (yikes) executing after a commit is run:
{"hookSpecificOutput":{"hookEventName":"PostToolUse","additionalContext":"[learning-opportunities-auto] The user just committed code. Per the learning-opportunities skill, consider whether this is a good moment to offer a learning exercise. If the committed work involved new files, schema changes, architectural decisions, refactors, or unfamiliar patterns, ask the user (one short sentence) if they'd like a 10-15 minute exercise. Do not start the exercise until they confirm. If they decline, note it — no more offers this session."}}As per the article, CCAs aim to maximize data transfer by inferring the "available bandwidth" of the network. CUBIC relies primarily on packet loss as a congestion signal. For recovery, CUBIC's window size is a cubic function of time since the last congestion event.
After the initial packet loss triggered purposefully the first two seconds in this experiment, the only thing which could cause loss is the network queue (i.e. a simple tail drop, fq-codel, etc) which cannot process packets faster than they can arrive. At this point the link is saturated. The loss becomes a signal for CUBIC to reduce its window. This causes the oscillations you pointed out.
Unlike CUBIC, BBR [0] uses a model-based approach that estimates the available bandwidth and leaves some headroom kind of like you suggest to achieve higher throughput, and doesn't react as aggressively to loss as CUBIC.
[0] https://datatracker.ietf.org/meeting/104/materials/slides-10...
Yes, I meant CUBIC implementation! But I'm glad I made the mistake-I learned some interesting things from the responses above.
I can see why they rewrote QUIC in Rust and for use in userspace, though going the in-house approach would warrant keeping an eye on the relevant kernel commits like a hawk to avoid missing bug fixes like these. These in-house implementations tend to have less eyeballs than the kernel.
I found it interesting that Cloudflare is not yet using BBR as the default in quiche. CUBIC's recovery in this day and age, and especially in datacenters with large pipes, seems so slooow to me. Almost two seconds with no loss whatsoever till achieving BDP again and then shooting itself in the foot every time it hits the ceiling. Each one of those losses a retransmission.
I would lean on Apple not doing something like this unless they could get privacy right, but Apple is starting to crack on privacy, not only by embracing ads, but has been for sometime already behind the scenes, or at least when I left in 2022. There were strong forces pushing away from collecting less data and towards collecting more, all in the name of "improving user experience".
I think there's a really good opportunity to also incorporate Claude into into GIMP.
As mentioned in the article, this was almost certainly someone's pet and dumped in the river when they couldn't take care of it anymore. Axolotls are endemic to Mexico.
Yesterday I went to Walmart, and at the self-checkout the system quirked out and an attendant came by. She reviewed some sort of draconian overhead cam video of me trying to locate a tag out for a product to scan. Gave me "guilty until proven" innocent vibes. Are these systems actually effective?
I hope they will turn Siri around with these changes.
The one on their website is now $100, not $75 as per the article.
Surely someone can build a more privacy-friendly decentralized solution?
It doesn't take much for Tin Can to start charging $50 / month for a subscription.
This post feels a bit ideological in nature. There are recurring processes at organizations to assess someone's fit in a given role. Generally when someone is no longer performing as expected, they're given time to correct their ways or retrain. If the organization has no system in place, then it really falls on the organization. I wonder if the author overestimates the amount of people who really want to be in a role they're not good at.