Greg's recent activity
-
Comment on Job offer in a new city -- making friends? in ~life
-
Comment on Thoughts on Donald Trump, America and what this all means in ~society
Greg Link ParentDo you believe that the people perpetrating all these horrors, historically and in the present, were really only doing so because of the systems around them? That if you gave the same people a...Do you believe that the people perpetrating all these horrors, historically and in the present, were really only doing so because of the systems around them? That if you gave the same people a clean slate, they wouldn’t gravitate to the exact same tribalism under some other banner?
That’s where I’m flailing here: “the system” might be broken, but there are too many examples to count throughout history of large numbers of people willingly and enthusiastically choosing to uphold it. This election, and COVID before it, have put a severe hole in whatever ability I was clinging onto to believe we’ve become any more enlightened now.
-
Comment on Thoughts on Donald Trump, America and what this all means in ~society
Greg Link ParentOne of my biggest fears is what this says for the internal politics of a lot of previously safe-seeming options. The people know exactly who Trump is at this point, and they asked for this,...One of my biggest fears is what this says for the internal politics of a lot of previously safe-seeming options. The people know exactly who Trump is at this point, and they asked for this, resoundingly. When the inevitable shockwaves come from his policies and actions, the barest minimum we can ask for is being surrounded by people who'll take a stand rather than cheering him on.
I think @611828750722 makes very good points about America's particular susceptibility, but far right's power has been simmering globally for a decade or more at this point. I have zero faith that the people of the UK - my current home - will do better. Italy, Austria, Belgium, even France and The Netherlands are looking at various levels of meaningful popular support for extremists.
This is who we are, at least a plurality of us - not just Americans but humanity. Weathering the storm will be tough, but the part that's moved me to tears as I sit at my laptop on a random fucking Friday morning is that I don't know where to find a critical mass of people who I trust not to embrace that storm and then savagely mock those who lose everything to it.
-
Comment on My Leopold FC900R broke - Recommendation request in ~tech
Greg Link ParentOut of interest, what is it you've found to feel flimsy about Keychron? I haven't tried Leopold specifically, but I've used a few higher-end options from other manufacturers and the Q6 still feels...Out of interest, what is it you've found to feel flimsy about Keychron? I haven't tried Leopold specifically, but I've used a few higher-end options from other manufacturers and the Q6 still feels like an absolute tank to me (although I do agree that there are fine details that are sometimes a bit more refined on others).
-
Comment on Despite its impressive output, generative AI doesn’t have a coherent understanding of the world in ~tech
Greg Link ParentI'm not going to blame anyone for being sick to death of the marketing hype and reacting negatively based on the saturation of unrealistic bullshit that's being pumped out, but I think it's...- Exemplary
I'm not going to blame anyone for being sick to death of the marketing hype and reacting negatively based on the saturation of unrealistic bullshit that's being pumped out, but I think it's important not to overcorrect because of that.
These models demonstrably do show the ability to follow deductive reasoning steps. It's imperfect, it's brittle, it's often inconsistent - all helpfully quantified by this paper - but it's also the comparatively early days of the field. The fact that language modelling does expand into logical problem solving strikes me as a fundamentally important factor in what this tech will and won't change once the VC dust settles, and them being not-great at it right now is quite different to not being capable at all.
My experience is that dismissals like "empty words" or "glorified autocorrect" are almost as far from the truth as the breathless commercial fluff about <insert important human position> now being totally redundant thanks to ChatGPT. Underestimating the genuine technical capabilities and what they might imply feels a bit like the skepticism around this newfangled internet thing at the time of the pets.com collapse - entirely justified based on some of the ridiculous claims and ridiculous behaviours we're seeing from companies in the space, but at risk of missing the very real fundamentals as a result.
-
Comment on My Leopold FC900R broke - Recommendation request in ~tech
Greg Link ParentMy mistake, I must've read "full size / Keychron / UK ISO" and missed the K-series vs Q-series distinction! Good reminder that they do have a bunch of models with sometimes fairly subtle...My mistake, I must've read "full size / Keychron / UK ISO" and missed the K-series vs Q-series distinction! Good reminder that they do have a bunch of models with sometimes fairly subtle differences, in that case.
-
Comment on Despite its impressive output, generative AI doesn’t have a coherent understanding of the world in ~tech
Greg LinkInteresting paper - the core focus seems to be an automated, mathematically provable approach to measuring models’ performance on logic tasks. That’s got some real potential value in that it’s a...Interesting paper - the core focus seems to be an automated, mathematically provable approach to measuring models’ performance on logic tasks. That’s got some real potential value in that it’s a small step from measuring the objective to training for the objective. A nice feature of ML models is that if you can quantify how bad they are at something, you can often just turn that around and use let them target that metric for improvement.
I don’t love how the conflate LLMs with transformers in general: it originally made me think they were being a bit misleading and had done the taxi experiments on a model more trained for mapping than language, but I think it’s more just the way they’re using the term. The actual results do seem like they’ll have some worthwhile general application.
-
Comment on My Leopold FC900R broke - Recommendation request in ~tech
Greg Link ParentAnother Q6 user chiming in here - went for it as one of the few options for full size with ISO layout, same as @trim, and I’ve been very happy with the choice. I’ve been pleasantly surprised at...Another Q6 user chiming in here - went for it as one of the few options for full size with ISO layout, same as @trim, and I’ve been very happy with the choice. I’ve been pleasantly surprised at how valuable having QMK (for macros) and the rotary dial (for calls and media) have been: both comparatively small things that I wouldn’t have originally gone out of my way for, but once you use them 15 times a day it’d be a noticeable loss to go without.
-
Comment on ‘Mild’ tofu, ‘mild’ carrots, ‘mild’ pine nuts: my five-year quest to understand German taste in ~food
Greg Link ParentI just want to say that the phrase "regular non-recursive restaurants" has put a big smile on my face, so thank you for that!I just want to say that the phrase "regular non-recursive restaurants" has put a big smile on my face, so thank you for that!
-
Comment on A freeze dryer is not a reasonable purchase in ~tech
Greg Link ParentSomething about the dust left behind made me feel oddly uneasy. Like, I know water has a bunch of minerals in it, I'm perfectly used to seeing the marks left on the shower screen or on a pot left...Something about the dust left behind made me feel oddly uneasy. Like, I know water has a bunch of minerals in it, I'm perfectly used to seeing the marks left on the shower screen or on a pot left to boil for too long, but the residue being left in a dry powdery form rather than a (still dry!) liquid pattern apparently sets off my brain complaining to me about water not feeling like powder.
Which, in fairness, was probably a small echo of the much greater "wrongness" he was experiencing when trying a lot of the foods!
-
Comment on AI is killing remote work in ~tech
Greg LinkThe guy seems like an asshole (most of his tweets are pretty much live laugh love for investors and tech bros; “You don’t need a third space, you need a better first and second space”?! Fuck off),...The guy seems like an asshole (most of his tweets are pretty much live laugh love for investors and tech bros; “You don’t need a third space, you need a better first and second space”?! Fuck off), and with the way he’s phrased that first paragraph I’d be truly amazed if he doesn’t have a financial incentive to promote Anthropic. I also don’t put much weight in his logic here - at best it’s just one guy’s semi-informed opinion, at worst it’s going to be proven factually incorrect.
The one thing I do agree with is @floweringmind’s comment about ownership and cooperatives. Technology has been multiplying productivity for decades (centuries, really), and it looks to be speeding up rather than slowing down. At some point people might collectively realise that 40 hours per week for effectively every adult in the population isn’t actually a requirement to keep society functioning, but that’s going to be a long and difficult fight even after getting enough buy in - in the meantime, ownership is the only way to fairly distribute those productivity gains on a micro scale here and now.
-
Comment on Why the US General in charge of nuclear weapons said he needs AI in ~tech
Greg Link ParentThey should, and it is, but sadly that ship has sailed. The decision makers have to use the buzzword or they’ll have politicians on their collective ass questioning why China is “using AI” and...They should, and it is, but sadly that ship has sailed.
The decision makers have to use the buzzword or they’ll have politicians on their collective ass questioning why China is “using AI” and they aren’t, regardless of the reality of the situation. The press are gonna run with whatever version of the story gets eyeballs on ads. And the unfortunate truth is that a good number of snake oil salesmen will work their way into the cracks that forms and skim off their percentage.
It sucks, but it’s the reality of the situation, and I think that being cognisant of that is the best way to defuse it.
-
Comment on Why the US General in charge of nuclear weapons said he needs AI in ~tech
Greg Link ParentI absolutely wish the terminology were more precise, and I’m sure the headline writer was deliberately leaning into that ambiguity for clicks, but I think your comments here are fanning those...I absolutely wish the terminology were more precise, and I’m sure the headline writer was deliberately leaning into that ambiguity for clicks, but I think your comments here are fanning those flames more than highlighting the issue.
The article body puts it very well: ”Wellerstein pointed out that replacing ‘AI’ in his sentences with ‘computer analysis’ renders them mundane.”, so I don’t think there’s significant confusion about the real conversation. I’m sure some idiots will propose, and possibly even attempt, a full Robocop-meets-Skynet dystopia at some point - but this isn’t that, and the things that were said seem pretty reasonable in context.
-
Comment on The Strava problem: how the fitness app was used to locate the world’s most powerful people in ~tech
Greg Link ParentI can see it being one of those things that doesn’t particularly matter except for the 1% of times that it really, really does: visits by wartime leaders that aren’t announced ahead of time for...I can see it being one of those things that doesn’t particularly matter except for the 1% of times that it really, really does: visits by wartime leaders that aren’t announced ahead of time for security reasons, early stage talks between hostile countries that don’t want to acknowledge that they’re negotiating until they’ve got the shape of an agreement, that kind of thing.
And you’d probably be right to say “well of course everyone’s on double secure best behaviour in those cases”, but then articles like this are one of the ways that things get bumped up from “best practice” to the “don’t fucking do that, not even if you’re off duty, not even if it says it’s anonymous, if a newspaper can figure it out then what do you think a real security threat could manage?!” list.
-
Comment on Ask Tildes: What kind of non-mainstream AI are you using regularly? in ~tech
Greg LinkI build the things (assuming we’re using the same definition of “AI ≈ big neural nets, probably transformers”), and a lot of the more compelling use cases I see are comparatively dull...I build the things (assuming we’re using the same definition of “AI ≈ big neural nets, probably transformers”), and a lot of the more compelling use cases I see are comparatively dull task-specific models for data handling: medical imaging, financial projections, internal document search and summarisation, that kind of thing.
The gains there compared to state of the art even two years ago are vast; if you’re interested in specific examples you might find it helpful to take a look at the competitions companies are sponsoring on Kaggle, that’ll give you an idea of what they’re publicly paying to solve right now.
The other major example that comes to mind is things like Photoshop, or your phone’s camera software (the actual internal postprocessing, even if you don’t edit the pictures): they’re mainstream tools, but not what most people think of as AI per se - the modern versions just have their functionality heavily augmented by neural net models.
The subtext to what I’m saying there is that the tech itself is often incredibly capable, but most of the time when a product is shouting “AI” as the headline it’s a sign that they don’t know what else to shout about in terms of user benefits. A lot of the actually really useful models for voice recognition, image search, OCR, machine translation, etc. are things we kind of take for granted already - because they’re not products per se, they’re a component of the wider ecosystem that’s there to solve a problem. It can control the home automation while your hands are covered in bread dough, or to find that cute picture you took three months ago, or give you an idea of what that menu says while you’re in Shanghai, or any number of other similar things.
-
Comment on Goodbye, floppies - San Francisco pays Hitachi $212 million to remove 5.25-inch disks from its light rail service in ~tech
Greg Link ParentIf a system really is perfectly functional then I’d agree, but I’ve rarely seen it play out that way, unfortunately. In my experience it’s either a case of a huge amount being spent to keep the...The fact that a system built decades ago is still perfectly functional today is a good thing, it means we are getting our money's worth.
If a system really is perfectly functional then I’d agree, but I’ve rarely seen it play out that way, unfortunately.
In my experience it’s either a case of a huge amount being spent to keep the legacy system running (the military paying whatever Microsoft asked for extended Windows XP support, banks paying IBM and anyone who knows COBOL to keep their mainframe systems alive, etc.), or a ticking time bomb that’s going to absolutely decimate the budget when it fails because nobody knows in full how to repair or replace it.
Often it’s both, with the expensive patches on top of patches on top of patches being used to buy another year because nobody knows what to do when the whole house of cards comes down.
There’s also the opportunity cost of not even being able to consider related projects and upgrades that might make the system as a whole safer/cheaper/more efficient because doing so would conflict with the old tech.
I’m definitely not saying it can’t be done - I’m sure there are some bits of totally self contained factory or lab equipment out there that’ll keep ticking until they physically wear out, long before we collectively use up all the floppy disks stored in that one guy’s warehouse - but more often than not it’s been a sign of much bigger problems when I’ve seen it.
-
Comment on Anthropic announces New Claude 3.5 Sonnet, Claude 3.5 Haiku and the Computer Use API in ~tech
Greg Link ParentIn fact, I’ve just been recommended this video that shows it choosing which program to open, making its own decision to use a combination of visual OCR and browser devtools for a web based task,...In fact, I’ve just been recommended this video that shows it choosing which program to open, making its own decision to use a combination of visual OCR and browser devtools for a web based task, and controlling LibreOffice and XPaint for other tasks: https://youtu.be/DVRg0daTads
-
Comment on Anthropic announces New Claude 3.5 Sonnet, Claude 3.5 Haiku and the Computer Use API in ~tech
Greg Link ParentI think that’s a failing of the demo video, not the tooling. I noticed that the “CRM” is running on localhost:3000, for example, so I’d make a heavy bet that this is just a developer testing...I think that’s a failing of the demo video, not the tooling. I noticed that the “CRM” is running on
localhost:3000, for example, so I’d make a heavy bet that this is just a developer testing things/setting up a demo in the way that gives them the most visibility and flexibility. Claude is now the lead on the OSWorld benchmark by a significant margin, and that’s by no means limited to browser applications.Navigating like a human rather than in the theoretically more sensible machine readable ways makes it agnostic to whether there’s an API or not - there’d be no real way to move from web to desktop software if they hadn’t had it drive the mouse and keyboard. I think @Grumble4681’s self driving car analogy was spot on there: if it can handle navigating in the same way as humans, it doesn’t rely on updating the extremely long tail of places (or programs) that aren’t designed to be navigable by machines.
Now, you could say that a zillion dollar company like Anthropic could have communicated that better in the video, and it’d be a fair comment, but I’ve also been the dev in the past who was told by marketing that they needed a screen recording by this afternoon and “just use whatever you’ve got set up”. I expect we’ll see examples of it navigating desktop software in the immediate future, assuming they aren’t out there already.
-
Comment on Norway is to enforce a strict minimum age limit on social media of fifteen as the government ramped up its campaign against tech companies it says are “pitted against small children's brains” in ~tech
Greg Link ParentI’m also broadly on the “banning stuff doesn’t really work” side of things, but I do wonder if this situation in particular has a couple of extra factors in its favour. Social media by definition...I’m also broadly on the “banning stuff doesn’t really work” side of things, but I do wonder if this situation in particular has a couple of extra factors in its favour.
Social media by definition operates on network effects, and it seems like even if the ban is, say, 30% directly effective it’d go a long way to disrupting that. It gives all the parents who were otherwise on the fence a much stronger push to go with their gut, all the kids who are worried a way to save face, and puts a direct roadblock in the way of all those who simply don’t know how to circumvent it.
Suddenly whatever platform isn’t the one place where everyone is anymore - and the pressure to be there or face social ostracism starts to evaporate.
-
Comment on Anthropic announces New Claude 3.5 Sonnet, Claude 3.5 Haiku and the Computer Use API in ~tech
Greg Link ParentDoes Windows desktop software commonly come with an API that can be accessed by other tools on the system? That was my first thought on who this is for: small to medium B2B customers who want to...Does Windows desktop software commonly come with an API that can be accessed by other tools on the system? That was my first thought on who this is for: small to medium B2B customers who want to automate whatever inefficient workflows they’ve built up around internal processes and/or task-specific proprietary tools.
Could well be that there is some common automation standard there they could be hooking into that I’m not aware of - I don’t use or develop for Windows day to day - but my impression was that desktop applications tend to be tightly coupled to their UI toolkits as far as any external app is concerned.
Strongly seconding this - I’d say that doing it because it’s a significant and somewhat intimidating change is almost reason enough on its own. There’s huge value in experiences like that, and the fact it’s a meaningfully better job as well is the icing on the cake.
One thing that might help is intentionally reframing it (to yourself at least, perhaps to your friends as well) as a two year project. You’re not (necessarily) leaving forever to build a new life elsewhere, you’re going off on an adventure in search of learning and new experiences. You’d be eyeing up new jobs in 2 - 3 years anyway; maybe you love Vancouver at that point and you absolutely do want to stay, maybe you feel like heading back to Toronto, maybe you even decide that you want to hit the road and explore what else is out there - as long as you go in with the mindset that the first couple of years are an experiment, none of those options will feel like failure.