hereonout2

Yep, it's these simple thingsthat are actually the killer features for me. Alexa is so cheap too (often given away free with other products like ring cameras) that I have them in most rooms.

Mainly I just use them as alarms, timers, a talking clock and for listening to radio. I occasionally ask it what time a local shop may close or similar.

For the little money I've spent on them I'm perfectly happy using them in such a superficial manner, they make life a little tiny bit easier and that's ok.

Older programmer too, tables for layout is very retro though! Hopefully you're not still adding a spacer.gif here and there.

I recently picked up some front end work for the first time in about 15 years. Things have progressed so much, CSS grid is a dream but also the things that were horribly hacky in the 2000s with JS are now solved so well.

Not even talking about frameworks or react, with just vanilla CSS, JS and the mdn docs you can build really nice sites that were so so awkward to make before.

I'd encourage you to spend a few afternoons delving deep into this, a lot of the stuff I learnt years ago is still relevant and brings a solid foundation, it didn't take long to update my approach.

Still using it but the regions issue is annoying.

We currently deploy to an entirely odd region way outside our geographical location because it had the most available models when we set up, even then it still lagged behind and only got gpt-4o recently.

I find the whole concept of a "deployment" odd, clearly it's not deploying anything other than a config enabling access for your account. Great, now I have to keep track of every model release and have a DevOps process to keep on track!

Also yes we noticed slow downs, but as we've stopped using openai directly I don't know if this just applies to azure.

Absolutely, I try to "automate" or at least have scripted almost everything I do.

I do this almost always for reproducibility and documentation purposes first.

I'm assuming this is academia and teaching to undergraduates the foundations of programming / comp sci using Java (which I'm not sure should be many peoples first choice these days!).

This situation is pretty common and has existed for time immemorial - undergraduate tutor's are not likely to be seasoned developers with significant real world experience - at least outside the top tier of institutes.

Many new hires / junior devs seem to have had this experience, and often come in lacking any knowledge of pretty important concepts like version control, testing, http, etc. It's ok, they have the basics and can learn on the job but I do often wonder what it would be like if the curriculums mirrored the real world better.

OP, hopefully no offence taken, the fact your coming here asking makes me think you'll do a fantastic job of this course!

I don't have a java book recommendation, but if you could try and squeeze in aspects like testing, http frameworks, git and the basics of a Linux terminal you'd really be helping the students with prospective job interviews :)

Edit: Wow just read course curriculum, Swing, CORBA, Servlets - throw in some XSLT and I am reliving the late 90s. OP - any chance you could swap some of these out for more modern solutions? Web interfaces / Rest / Dropwizard, spring, etc

Indeed, some big, rare and once costly things are just too cumbersome to have a huge market.

Who has the space to store an antique mainframe?

Same thing happens in other markets, industrial grade 12 foot bed CNC lathes can be picked up for similar prices to small hobby machines. Great deal if you have a spare barn, 3 phase power and the logistics to move it. Not many do!

> They live by a sort of unspoken code that they believe exists but which other people may or may not be aware of.

I try to be a nice guy but also believe (hope) I'm competent!

On the above that's something I've recognized though, I do feel like I bring some sort of unwritten moral code to my work (as do we all I'm sure) but realizing that nobody else is actually aware of it or follows it is quite important.

I think this goes some way to being able to be "nice" without getting resentful or bitter when things don't go your way.

From the kind-hearted colleague who's freindly at the water cooler but actually very difficult to work with - through to the people who are just rude, dismissive or belligerent - acknowledging that they operate by a different set of principals to you and accepting that somewhat release's you from their effects.

Expensive in general but combine some decent tooling and spot instances and it can be insanely cheap.

The latest Nvidia L4 GPUs (24GB) instances are currently less than 15c p/h spot.

T4s are around 20c per hour spot, though they are smaller and slower.

I've been provisioning hundreds of these at a time to do large batch jobs at a fraction of the price of commercial solutions (i.e. 10-100x cheaper).

Any problem that fits in a smaller GPU and can be expressed as a batch job using spot instances can be done very cheaply on AWS.

I've worked professionally over the last 12 months hosting quite a few foundation models and fine tuned LLMs on our own hardware, aws + azure vms and also a variety of newer "inference serving" type services that are popping up everywhere.

I don't do any work with the output, I'm just the MLOps guy (ahem, DevOps).

You mention expense but on a purely financial basis I find any of these hosted solutions really hard to justify against GPT 3.5 turbo prices, including building your own rig. $5k + electricity is loads of 3.5 Turbo tokens.

Of course none of the data scientists or researchers I work with want to use that though - it's not their job to host these things or worry about the costs.

> While MDMA is sometimes termed an “empathogen,” empirical studies have struggled to clearly demonstrate these effects

Clearly the scientists carrying out these studies need to change their working hours.

Visiting any European dance music venue on any given weekend in the last 30 years would have provided adequate demonstration of the effects.

100k records is still pretty small!

It feels a bit like the hype that happended with "big data". People ended up creating spark clusters to query a few million records. Or using Hadoop for a dataset you could process with awk.

Professionally I've only ever worked with dataset sizes in the region of low millions and have never needed specialist tooling to cope.

I assume these tools do serve a purpose but perhaps one that only kicks in at a scale approaching billions.

Indeed, it's almost a cliche.

I believe it satisfies some sort of primeval urge magnified by sitting in front of a screen for decades.

I just paid to get my lawn cleared for a new patio. A guy single handedly shovelled 8 tonnes of clay soil into a skip over two days.

He charged much less than I earned in the same two days configuring AWS lambda functions. It's hard to imagine I had the more exhausting job.

This is true for almost all of us though? Nearly everyone has to work at least 5x 8 hour days a week - some much more. Some have complex family needs that must be attended to before any personal time can even be considered.

There is always some time though, it just needs to be made and scheduled and worked on. We can choose to veg out in front of Netflix, or doom scroll, or peruse hacker news, but we can also choose to do something else possibly more fulfilling.

Of course not as good as having more disposable money and fewer work commitments but painting it as an all or nothing situation feels very defeatist.

During my late 20s I realised I was very good at my programming job but not much use at anything else. It was quite a disappointing realisation.

I started to do less and eventually no programming outside of office hours and instead invest my spare time in different hobbies and experiences.

I found gaining these new skills really helped build my confidence. 15 years later I'm not just a programmer, I'm a also a motorcyclist, experienced carpenter, been a member of certain meet up groups for over a decade, travelled to a few exotic locations, flown upside down in a plane, the list goes on.

None of these things are exceptional but doing this extra stuff has given me enough dimensions that I kind of feel comfortable with the way I've done the last 15 years. This along with starting a relationship and building a family is enough for me now.

Sure, I could have done more but compared to the corporate quagmire of my 20s things are very different.

I didn't read any sympathy in that comment, they just seem to be pointing out that 25 years is not a "low" sentence.

I would also agree that spending a 3rd of your life in prison is in effect destroying it.

An absolutely fantastic publication, nice to see it still going.

If you can find it there's a 1000 page hardback compendium of some of the best articles available too.

This is a bit more useful than things like vscode or flake8 (my preferred option) telling you that you "import xxx" but don't use it.

I'd love a deeper way to identify and selectively reduce the cruft from projects though. For example currently I have about 180 dependencies in my project's .venv - most of them transitive.

These include things like the google-auth library when there is no code in my project that would ever need to talk to Google.

Mostly this is just wasteful and not a big issue, but then come the conflicting dependencies you need to resolve by pinning version numbers. This is doubly annoying if you have no idea what the dependency is needed for or even if it's needed at all.

This is not just a python issue, it's apparent in all languages I've worked in.

It also has no articles, one of the main ones listed is titled "Hello World".

Looks like a toy weekend project by someone learning web development. A worthy endeavour no doubt but not sure why it's on the front page.

Should have read the manual.

People will throw up all sorts of excuses for not just renting a VM for a static site - scalability, ease of use, security.

But in the vast majority of cases you could just take a $5 VM, apt-get install nginx and be absolutely fine. A tiny bit more effort and you can make sure it's always up to date and very secure. Plus you get a VM you can use for other things when needed.

10M tokens is something like 14 copies of war and peace, or maybe the entire harry potter series seven times over. That'd be some legal document!

Unfortunately they don't just sell singles, so you quite often get to number 15!

You can install pytorch with it, I have been doing so for a while.

  [tool.poetry.dependencies]
  torch = { version = "1.13.1+cu116", source = "pytorch-cuda" }
  
  [[tool.poetry.source]]
  name = "pytorch-cuda"
  url = "https://download.pytorch.org/whl/cu116"
  priority = "explicit"

But, it does seem to go off and download many GBs of pytorch packages for all possible python, architecture and OS versions which takes ages. Feels very broken at least.

> The needs of people writing and deploying HTTP/REST servers is just so different from the people writing PyTorch models and numerical simulations that no tool will ever satisfy the very different needs of both camps.

Is this true though? At the end of it they both need python packages and some system dependencies installed (let's ignore models and data for now).

Why is there umpteen tools for doing this in python when there isn't in other languages? I have to deploy both web apps and ML models, the first thing I do is convert any project to use pyproject and poetry.

Whilst Poetry comes with it's own issues I've not had a project yet that this doesn't work for and wish the wider community would just settle on one method. Instead we have stuff coming in with various conda incantations, pip, pipx, poetry, setuptools, setup.py!

Deployment of ML models needs a decent solution too, half the ML code I get goes and fetches stuff from NLTK or Huggingface at runtime. Some of it (like the LLAMA models) needs various API keys set and EULAs agreed to before it'll run, then pings back to Huggingface each time you run it to check the EULA again! This makes life difficult when trying to deploy and adds this massive dependency on 3rd party services.

Indeed, it was just the first one that came up in the search to illustrate the massive difference.

It's about twice the house you'd get for Manchester money. Maybe there's other factors to consider, but the idea that Romanian housing is the same price as the UK's doesn't make much sense to me, and I can't find evidence of it being true.

I just found a newly built 7 bedroom villa with pool in Bucharest for under €500k, not ideal as it's near the airport but if anyone could point to an equivalent in the Manchester metropolitan area I'd be very interested.

Also found an entire building on Bucharest's second busiest street Magheru Blvd for €1.4M. Out of my price range unfortunately but it does include 12 separate 3 bedroom apartments, a floor of offices, and a restaurant, cafe and art gallery at street level.

Behold the wonders of the modern age!

Depends on the definition, but only the very oldest have hit 40, many in their 30s and some still in their 20s. It's too broad to be of use here, those entering the workforce in 2006 were in a different world from those doing so in 2020.

That's not quite true. The EU, US, UK, Japan all have legal minimum ages for employment ranging from 13-15.

I don't think the laws have changed that much from when I was young.

There's more focus on remaining in education sure (which is generally a good idea but often poorly implemented) but I think there's wider societal explanations as to why you see far fewer kids working outside school hours than you once did.