neumann

The architectures of these models are a plenty good scientific basis for this statement.

Agreed, but maybe that's what you call it when you get your configs wrong

likewise

I don't understand your question. What does citizenship got to do with this?

I agree to a degree, but I am in that camp. I subscribe to alphasignal, and every morning there are 3 new agent tools, and two new features, and a new agentic approach, and I am left wondering, where is the production stuff?

Thank-you! It is great to read a honest and self-aware journal covering this.

As someone who fantasizes about running my own thing, working for a few start-ups have made me very stringent on what my requirements are for starting something (co-founders, investment, location, market). And also that these requirements have become so very risk-averse that I probably am not the personality profile to run a business! Nevermind the endless imposter syndrome.

Approach it from the other angle - what scenario would it be bad in. It's not hard to see very real possibilities in the short term where it does matter: A 16yro looks up on chatGPT how she can check discreetly whether she is pregnant or not and what potential avenues she has. The advertiser could literally by anyone targeting pregnancies, including government or action groups who now have some information about that user's conversations in this scenario.

Any data exfiltration or reporting on the users would quickly be developed by the industry to merge this information and improve inferences with confidence values on target populations/individuals.

yes! PhonieBox - But you built it yourself [0]. You make your own cards with nfc/rfid stickers in them, put a nfc/rfid reader somewhere nice, and hooked up to phoniebox rpi with spotify to a nice sound system.

https://github.com/MiczFlor/RPi-Jukebox-RFID

This is very cool. And I can imagine sometimes it's not even about keeping them accountable, but demonstrating how easy a solution is to a problem that council bureaucrats would not have the capability to solve - likely have to outsource to a vendor that would charge for 'out of scope'.

Thank you. It is always intimidating to see how people are using 5-10 agents for coding, but haven't seen practical examples of the process and how people do this specifically.

Is there a mini tutorial demonstrating best practices for people keen to try but who haven't used agents like this?

This is definitely something to be aware of - especially with larger companies that aren't growing fast and this culture begins to be baked in. You see so many colleagues going the extra mile past their role requirements to earn that rare promotion, essentially jockeying to be in the running. All hands are about 'calling out' great performers and thanking them. Thank them by paying them more please.

As soon as they get the promotion, the work piles on even more, and they won't be given the amount they would if they switched companies.

I don't think the size of the working stockpile matters. It's that they have enough. Even the successful detonation of one working nuke would be catastrophic.

oh god - please tell BoM in Australia. Either ML is not keeping up with clime change unpredictability, or SOTA is worse than what we had 10 years ago.

This is actually nice and balanced, but the title is misleading. I feel like ALL I hear about maintaining an open source project is how hard it is and how people burn our. I almost never read a blogpost or comment declaring how rewarding it is. So, this was a nice (slightly) more balanced view.

Ouch. I think most people would assume he got permission, so this is a bit ugh

Yeah. I bought a black duo out of nostalgia and wanting pebble to succeed, but not interested in the time and realized I don't love them enough to want to wear a white one. Fickle me, I guess.

Yes please. Or Jesse armstrong and Adam McKay

Scale and purpose is what you are missing.

If you paint a Studio Ghibli totoro on your cup, pillow, PC, T-shirt, nobody is going to care. If you do this a thousand times it obviously is an issue. And if you charge people to access you to do this, it is also obviously an issue.

I've given up on America. Democracy has essentially been destroyed in front of our eyes. So many of the current ICE and administrations acts are clearly setting up/testing response to secure the midterms through fear, intimidation, corruption, and blatant disregard for the law. And it will succeed because there is no one at this stage in authority to speak up, and they already fired anybody who wasn't aligned/acolytes.

Now I just hope these Nazi tactics don't continue to become a playbook in my country. Meanwhile, actual Nazis are emboldened and encouraged.

Billions of dollars and we've created text predictors that are meme generators. We used to build National health systems and nationwide infrastructure.

This is exactly what we use it for

What is NINBYism?

I think mentioned elsewhere here, https://github.com/MiczFlor/RPi-Jukebox-RFID is great for this. I did something similar with an opp shop Fisher Price[0] record player, with the RFID reader under the turntable and each 'card' is a laminated record cover with the rfid stuck on it. Lots of good photos of different implementations in their issue threads.

We also use it for kids podcasts (autodownloads them weekly). I added a TTS script that generates a friendly audio message from a text file that can be triggered to play from an alarm or for a specific record. This announces the weather with a Dad joke at the end. I tried to automate the last one with various sources (db, LLM, etc - but felt too cold, so I just dictate it to the server from the phone) and usually add a customised message about our family calendar (wear a jacket for rain. cousins are coming today).

[0] https://www.amazon.com.au/Fisher-Price-Classics-Record-Playe...

Raining is essentially doing as a 'built in' feature for what users found earlier that requesting longer contextual responses tend to arrive at a more specific conclusion. Or put it inversely, asking for 'just the answer' with no hidden 'reasoning' gives answers far more brittle.

I have been using code + vscode extensively for coding, but in the last few months it has been a frustrating downgrade compared to the same prompts and code being pasted into chatGPT.

Is this going to be the way forward? Switching to whichever is better at a task, code base or context?

Yes, they are going to blow a lot of money on blowing hot air into this

Rather than anti-AI, I think most of the HN crowd are educated enough about underlying technology and it's capabilities to see the snake oil being pushed. Most of us would have given it a good go (and probably still use it daily), just know the difference between its actual limitations and the dreams that are being sold.

oh, I was definitely rocking a 17" CRT in 2001! Pickup up 2 other guys with CRTs and lugging it to a friends for a LAN party we couldn't fit them in the boot of the Ford Meteor and had to cram them in with the passengers in the back seat.