adutchman's recent activity

  1. Comment on Need advice about work in ~life

    adutchman
    Link Parent
    Not necessarily, in the Netherlands it's mainly driven by the uptick in EVs, heatpumps and solar panels.

    Not necessarily, in the Netherlands it's mainly driven by the uptick in EVs, heatpumps and solar panels.

    3 votes

  2. Comment on Need advice about work in ~life

    adutchman
    Link Parent
    +1 there is still a living to be made in software, though it is tougher to get in.

    +1 there is still a living to be made in software, though it is tougher to get in.

    1 vote

  3. Comment on Need advice about work in ~life

    adutchman
    Link Parent
    I can attest that, due to the energy transition, skilled electricians are in super high demand.

    I can attest that, due to the energy transition, skilled electricians are in super high demand.

    3 votes

  4. Comment on Chicago Mayor Brandon Johnson finally explains his decision to bid $3.2B for parking meters before walking away in ~transport

    adutchman
    Link Parent
    I agree that it is skewed, because AI tends to be disproportionately usefull as a dev. Even then, I do get sick of vibecoded crap that inevitably breaks (and who has to fix that...). I also hate...

    I agree that it is skewed, because AI tends to be disproportionately usefull as a dev. Even then, I do get sick of vibecoded crap that inevitably breaks (and who has to fix that...). I also hate the fact that lot's of devs take the opportunity to take AI as an excuse to make more shitty stuff instead of making good stuff slightly faster. I also hate dealing with people who answer my thought out question with "this is what Claude said". It's all around a sad state of affairs.

    Forced AI use is the worst, that is just directors/bosses going with the next shiny thing and that never helps anyone. I've heard of companies forcing a minimum amount of tokens used (yes, you read that right). That just ends up with devs generating lorem ipsum to hit the quota.

    With all that being said, it still had it's place, for me at least. I use it to check my texts for logical errors (I write it myself first, then it gets a go at correcting errors I have to approve), take a technical text I wrote and some new requirements and updating it (which I then check the changes of and improve) or make overview images/documents which are a one-off, to name a few. All tasks I could do myself, but don't add much additional value, joy or learning for me.

    You shouldn't be forced to use it, no, but it could still be usefull. Wishing you the best with dealing with the forced AI bullshit at work though, that sucks.

    3 votes

  5. Comment on Chicago Mayor Brandon Johnson finally explains his decision to bid $3.2B for parking meters before walking away in ~transport

    adutchman
    Link Parent
    Okay, so I looked it up: I thought I remembered this from an AI course I took, but I misremembered. Summerisation is one of the finetuning steps you can take (make an LLM more domain specific),...

    Okay, so I looked it up: I thought I remembered this from an AI course I took, but I misremembered. Summerisation is one of the finetuning steps you can take (make an LLM more domain specific), but it is not necessarilly one of the pre-training tasks, that is exclusively next token prediction.

    Sorry about that.

    Seperate from that I want to make a distinction based on my experience: you might be mentioning something like the AI summary from Google, which indeed is lauhably bad. That is probably because it is not a true summary: it takes parts of the search results and fills in the rest, and that is prone to the AI hallucations we're all used to.

    That being said, if you take a modern model, and input a full text that isn't too long as input, it will produce a fairly accurate summary in my experience. The example above shows this as well. AIs follow their input, but input tokens are expensive, so Google cannot input all top artocles in it. If there isn't enough input, the AI starts to go off script which increases hallucations. If the text is fully used as a source, this risk is a lot lower (but not zero).

    In general, I just feel like there is a disconnect between the opnion of AI of folks here and on broader forums, and of people I meet in daily life. Online, it seems people are either AI zealots (no thank you) or AI haters. I'm a dev, and AI is very useful for me in creating real worth, but that is only possible with effort, which is something the AI slop zeolots never put in (see the human attention, human effort blog). It is a tool you need to know when to use and how to use. Writing off AI wholesale seems counterproductive to me personally.

    2 votes

  6. Comment on Chicago Mayor Brandon Johnson finally explains his decision to bid $3.2B for parking meters before walking away in ~transport

    adutchman
    (edited )
    Link Parent
    I was not commenting on the reality of AI slop being horrible, which I wholeheartedly agree on. I pointed out that one of the tasks that LLMs are trained on, is summerizing. That makes them good...

    I was not commenting on the reality of AI slop being horrible, which I wholeheartedly agree on. I pointed out that one of the tasks that LLMs are trained on, is summerizing. That makes them good at it. It does hinge on the output being complete, because an LLM is a machine that produces output that is statistically likely, not necessarily accurate. So you're right on that front.

    In other words: despite AI being slop being bad, summarising an input is a task that they are pretty good at overall. That still does not mean you can trust them fully.

    4 votes

  7. Comment on Chicago Mayor Brandon Johnson finally explains his decision to bid $3.2B for parking meters before walking away in ~transport

    adutchman
    Link Parent
    AI summaries are not completely bogus: it is one of the tasks they were explicitly designed and trained to do. I for one appreciate the overview

    AI summaries are not completely bogus: it is one of the tasks they were explicitly designed and trained to do. I for one appreciate the overview

    10 votes

  8. Comment on Give me your culture clash stories in ~travel

    adutchman
    Link Parent
    /noise damn that does suck yeah

    /noise damn that does suck yeah

    3 votes

  9. Comment on Give me your culture clash stories in ~travel

    adutchman
    Link Parent
    Interesting. In the Netherlands, this would have been fine. If a busdriver needs to leave the bus, they just leave it open for people to get in while waiting for the next one to get on

    Interesting. In the Netherlands, this would have been fine. If a busdriver needs to leave the bus, they just leave it open for people to get in while waiting for the next one to get on

    11 votes

  10. Comment on Dbrand has cancelled their "Companion Cube" Steam Machine case, because they didn't ask Valve for permission in ~games

    adutchman
    Link
    Good on Valve, they should've just asked for permission. Shame nontheless though, looked cool

    Good on Valve, they should've just asked for permission. Shame nontheless though, looked cool

    3 votes

  11. Comment on The Online Accent Database: Explore authentic accents with a growing video library of real-world interviews in ~humanities.languages

    adutchman
    Link
    Too bad it's only English accents

    Too bad it's only English accents

    4 votes

  12. Comment on I'm looking for an adage or "law" (like Conway's law), but for dealing with AI slop in ~tech

  13. Comment on Why carbon capture can't solve climate change in ~enviro

    adutchman
    Link Parent
    That's grossly understating it. CCS is one of the best examples of fossil fuel lobbying and misdirection. For example:...

    That's grossly understating it. CCS is one of the best examples of fossil fuel lobbying and misdirection. For example: https://www.desmog.com/2024/10/07/uks-22-billion-carbon-capture-pledge-follows-surge-in-lobbying-by-fossil-fuel-industry-records-show/. 22 billion is a lot of money that could be used for solutions that are proven to help today. Instead it is used as an excuse to continue polluting.

    7 votes

  14. Comment on Why carbon capture can't solve climate change in ~enviro

    adutchman
    Link Parent
    Again, there's two parts to solving climate change: going net zero, and going negative carbon. Today, every dollar spent on CCS is a dollar better spent on reducing emissions. Claiming investing...

    Again, there's two parts to solving climate change: going net zero, and going negative carbon. Today, every dollar spent on CCS is a dollar better spent on reducing emissions.

    Claiming investing in CCS today is better than investing in any other solution like solar is incorrect.

    10 votes

  15. Comment on Why carbon capture can't solve climate change in ~enviro

    adutchman
    (edited )
    Link Parent
    The point of the article, I think, is not to say that the principle of CCS is wrong, but that fossil fuel companies have been pushing it as a 'solution' to climate change, so we don't have to...
    • Exemplary

    The point of the article, I think, is not to say that the principle of CCS is wrong, but that fossil fuel companies have been pushing it as a 'solution' to climate change, so we don't have to transition away from fossil fuels (or CO2 intensive processes). The fact is that that is plain wrong. CCS does not work today, many other solutions do, but they require us to transition away from fossil fuels.

    Now, CCS after we are largely done with the energy transition? That is an enticing idea, but in the same way fusion research is important: it might help us in the future, but not today.

    Btw, the fact that this discussion is confusing, is because CCS is a well-known fossil fuel industry decoy, here's a video on the subject: https://youtu.be/nJslrTT-Yhc

    17 votes

  16. Comment on Why carbon capture can't solve climate change in ~enviro

    adutchman
    Link
    Great write-up by ProPublica on how Carbon Capture and Storage has been hailed by oil companies a solution to climate change, even though it has never proven to work in the real world.

    Great write-up by ProPublica on how Carbon Capture and Storage has been hailed by oil companies a solution to climate change, even though it has never proven to work in the real world.

    2 votes

  18. Comment on Solar with grid connectivity, but no networking? in ~life.home_improvement

    adutchman
    Link
    Smart features can be nice if you want to manage your energy consumption with an Home Energy Management System (HEMS). I work for an organisation trying to get devices to talk more to each other...

    Smart features can be nice if you want to manage your energy consumption with an Home Energy Management System (HEMS). I work for an organisation trying to get devices to talk more to each other in the home, like you can already do with smart light bulbs and Matter. We're getting there with devices like solar inverters, but it takes time.

    1 vote

  19. Comment on The invention of buses in ~transport

    adutchman
    Link
    Incredible history. Great example of innovation being later and in a different form than you would expect.

    Incredible history. Great example of innovation being later and in a different form than you would expect.

    4 votes

  20. Comment on US battery industry cuts losses, shifts to new ventures amid electric vehicle bust in ~transport

    adutchman
    Link Parent
    Thanks for the correction, I was mixing things up there. The exemption in the US is indeed CAFE. A separate issue is the lax overall testing of things like pedestrian safety for US cars which...

    Thanks for the correction, I was mixing things up there. The exemption in the US is indeed CAFE. A separate issue is the lax overall testing of things like pedestrian safety for US cars which makes most SUVs illegal in Europe, but that does not factor into cost I think.

    5 votes