sottol

> 2) housing costs and genera cost of living making police salaries unattractive,

A friend's neighbor in wine country is an early retired officer due to injury - their benefits pretty much match many tech workers.

Looks like in San Jose police salaries start at $111k and max out at $186k, BUT additionally SJ spent $58M on overtime. _Average_ overtime pay for officer, sgt, lt and captain were $38-68k and some earned as much $200k in overtime. So the salary is probably $150-$220k on average I'd assume. Not as bad as many think.

But it's possible I'm reading some of this wrong.

[1] https://www.sjpdyou.com/for-applicants/additional-informatio... [2] https://sanjosespotlight.com/san-jose-spent-58m-on-police-ov...

The German's have this saying - "Erstens kommt es anders, zweitens als man denkt.". So I think we may have gotten existential AI risk wrong.

Existential AI risk may not be about killer robots turning us into paperclips but rather AI companions leading to fertility collapse and the end of civilization as we know it. You don't even need AGI for that - OpenAI and GPT4o seems enough.

Wasn't the strategy to exploit the European dependence on gas a failure then though because especially Germany didn't really put up a fight against the sanctions to keep using the cheap Russian gas.

One can argue about how "painless" it was, but I would have thought that stopping Russian gas imports within a few months would cause a recession and a lot more havoc in the economy. And I always assumed that was what Russia was counting on, Germany's dependence and thus tacit support but I didn't see much of that.

"120 cm in inches" (with space) works.

Annoying. AI feature, I assumed, but now you're just always second guessing.

As much hate as California gets, the CPRA and CPPA give me a little hope. Otherwise, maybe Europe might be better off with the GDPR.

But I'm almost more worried about what a company in such a position will do monetarily to its customers once it has reached a quasi-monopoly (or duopoly). I hope it turns out to be more like air-travel than Uber/Lyft, which afaict are now more expensive than cabs they replaced.

That makes sense. Is there a way to figure out which places have a full integration?

Booking.com does sometimes seem not to do a proper integration and just rely on emails or phone calls to confirm the bookings. I've used them multiple times around 10 years ago and the hotel would show me a printout of an email from them with my booking details afair.

A friend of mine recently had to sleep in the car because booking never confirmed her reservation with the hotel and wouldn't accept her booking.com confirmation printout.

I don't necessarily think it is the equivalent. Maybe it's more akin to a high-schooler reading a single book and then being asked to write several new books that hold equal weight in teaching the next generation of children.

These new text books could be great at simplifying the subject matter and making the material accessible or they may just never have fully understood the materials and are misleading.

Now imagine that over and over again, imo it's pretty likely to introduce inaccuracies if just taking a naiive approach.

I use Bluesnooze [1] on my Mac for that reason, it won't even disconnect when suspended otherwise and you have to log back in to turn off BT. It's pretty easy to disable on phones/tablets otherwise.

[1] https://github.com/domzilla/Bluesnooze

Sodium batteries are on the verge of commercialization, they're better in almost every way but energy density, making them ideal for energy storage imo. My hope is that this and other steps like this will make battery storage sustainable.

Not that cats are particularly relevant to human infection, but cats seem to suffer blindness, neurological impairment and death.

https://arstechnica.com/science/2024/04/concerning-spread-of...

Interesting that shipping is 1.7% of which supposedly 40% is fossil fuel transportation.

So that's about 0.7% for shipping petrochemicals and another 0.3% from pipelines. So 1% of global emissions is just the overhead of moving fossil fuels around to where they are processed/burned.

Can the same not be said of nuclear to a degree? My knowledge may be outdated but afaik proper, long-term nuclear waste storage is still an open question.

Nuclear waste persists and remains dangerous for a long time and it's easy to forget and just burden future generations with it.

Imo it just all stems from seeking maximum profits - externalizing environmental damage is just one facet of it. If China enacted strict penalties for environmental damage and it'd be cheaper to manufacture panels in Texas while polluting there, I don't have a doubt we'd build all of our panels in Texas.

I have a feeling that if there was not a healthy amount of competition in the space, the prices would start to trend towards the cost of human work.

> If you don't do that you'll run into the same problems as you would if you hire a contractor to build your codebase without understanding what they did for you.

I really like this way of thinking about using LLMs, I think that's a great analogy in many ways.

Maybe put another way, if I can get an LLM/AI to build exactly the product that I need, is a company that serves many customers simultaneously but probably worse still necessary?

I think it'll be hard enough to reason about what you really want that most customers won't care enough to roll their own. And personally, I'd happily pay someone to keep the product maintained. A product is usually not one and done.

The main thing that makes me skeptical is still what happens to a code base when you do this longer-term. And not just the code base but also the company when nobody understands the code any longer, but maybe neither are problems.

A couple questions:

* Will the codebase turn into a mess over time by having the AI apply changes over changes over changes? Do we even care? Or do we want a human to still be able to follow what is going on?

* Will you just be able ask the AI to refactor it all and clean it up? Then it wouldn't be a problem I presume.

* Are product-based tech companies/startups still defensible if anyone can basically recreate the product with some English?

* I don't know Codepilot Workspaces - are the prompts that generate and change the code kept somewhere? Imo they're part of the codebase now.

And unlike, say, Samsung Ultrabooks or even Microsoft Surfaces, Macs last a really long time. My kids are using my 2011 MacBook Air and 2009 iMac and they still work, even the battery still kinda hangs in. They've had a few rough years 2016-2019 with the butterfly keyboards but I don't know many current manufacturers with products as solid long term.

I am also worried about officers "gaming" the system to bolster whatever stat their superiors want them to optimize at the detriment of those being "policed".

We've all seen how prompting can change GPT output and this is what worries me the most.

Imagine users finding out that starting every interaction that is recorded with a certain sentence acts as a pre prompt that in turn subtly biases the gpt output so that it makes judges or prosecutors less sympathetic to defendants when the police department is optimizing for higher conviction rates, eg. (That is, doing the opposite of the quote above of the public defender "humanizing" the drug case defendant in the plea deal example.)

There will be a lot of reports transcribed, people will find ways to optimize the output to their gain and we've all seen how easy it is to bias llm output by promoting.

You can just get the GGUF off hugging-face. Usually googling for "<modelname> gguf" gets you to the site, then choose a quantization that works for you. But as I said, not working on Pixel 4. Should get a new screen tomorrow and can't wait to try on my Pixel 6 Pro.

Eg "phi-3 gguf" eventually take you here [1]. and you can download the Q4 quantized model on that page.

[1] https://huggingface.co/microsoft/Phi-3-mini-4k-instruct-gguf...

I've been trying this app but haven't had any luck getting it to actually generate text yet:

https://github.com/Mobile-Artificial-Intelligence/maid

The UI looks nice and includes a native compilation of llama.cpp.

My main phone's screen broke so I'm on an old Pixel 4 until it's repaired but I've had no luck getting 2-3GB models to run so far.

A big slide happened in Big Sur near Limekiln State Park and they're still fixing it, it's been over a year there. But this doesn't look as bad.

Literally "killer" imo.

It did look like there was some sort of object tracking indicator in the video feed afaict but could easily be added in post-production.

You can imo see RF jamming of the fpv feed when the drone approaches the tank, quite interesting. Probably what neccesassites the automated terminal attacks.

Imo it doesn't matter how sophisticated the tech is, if it works with a meaningful success rate than it's quite an effective weapon against a multi-million rubel tank.

Sure, CUDA has a lot of highly optimized utilities baked-in (CUDNN and the likes) and maybe more importantly, implementors have a lot of experience with it but afaict everyone is working on their own HAL/compiler and not using CUDA directly to implement the actual models. It's part of the HAL/framework. You can probably port any of these frameworks to a new hardware platform with a few man-years worth of work imo if you can spare the manpower.

I think nobody had the time to port any of these architectures away from CUDA because: * the leaders want to maintain their lead and everyone needs to catch up asap so no time to waste, * and progress was _super_ fast so doubly no time to waste, * there was/is plenty of money that buys some perceived value in maintaining the lead or catching up.

But imo: 1. progress has slowed a bit, maybe there's time to explore alternatives, 2. nvidia GPUs are pretty hard to come by, switching vendors may actually be a competitive advantage (if performance/price pans out and you can actually buy the hardware now as opposed to later).

In terms of ML "compilers"/frameworks, afaik there's:

* Google JAX/Tensorflow XLA/MLIR, * OpenAI Triton, * Meta Glow, * Apple PyTorch+Metal fork.

That is sort of what I mean, to be able to continuously regenerate the output in the same manner and achieve consistent results, imo, you would probably converge on something that leaves little open to interpretation - so mathematical notation or legal code.

90% of the work of building real world applications is imo fixing and adding features, i.e. maintenance. I see it rarely considered when talking about "english as code". How does this work in a more complicated application with multiple "pages" and "features"? One conversation per feature? How do you make changes to those when requirements change? If you just go back and change parts of the conversation, how do you make sure the final product still covers all the existing use-cases?

The example is outputting (afaict, skimmed most of it) effectively 15 lines of code. In terms of features, the equivalent of a single button on an ops dashboard page.

That's just not maintainable. Sure, you can write one-off, throwaway "code" this way but there's no way you could build (and maintain) a business this way, at least imo.

You'd need to "regenerate" chunks of your business logic from changed requirements, hoping they'd remain about the same. Thus why I think you'd converge on Math or Legal Code, precise and hopefully reproducible output when slightly tweaked.

I can reach our AWS site from my phone on LTE but when I switch to Comcast Wifi the site is no longer reachable.

+1 Comcast problem

Even if it does, it will end up looking more like a mathematical proof with exact definitions, sub-definitions and the equivalent of "lemmas" and such.

Or maybe it will like current laws & regulations, legal code, maybe something like this (first legal code I could think of...) [1].

But I'm sure it will not look like what people imagine when they first hear "programming in English" - a book or a nice story or a conversation.

[1] https://codes.findlaw.com/ca/vehicle-code/veh-sect-23152/