Interesting read, thanks! I’ll finish it later, but already this bit is quite interesting:
Without access to gender, the ML algorithm over-predicts women to default compared to their true default rate, while the rate for men is accurate. Adding gender to the ML algorithm corrects for this and the gap in prediction accuracy for men and women who default diminishes.
We find that the MTEs are biased, signif-icantly favoring White-associated names in 85.1% of casesand female-associated names in only 11.1% of case
If you’re planning to use LLMs for anything along these lines, you should filter out irrelevant details like names before any evaluation step. Honestly, humans should do the same, but it’s impractical. This is, ironically, something LLMs are very well suited for.
Of course, that doesn’t mean off-the-shelf tools are actually doing that, and there are other potential issues as well, such as biases around cities, schools, or any non-personal info on a resume that might correlate with race/gender/etc.
I think there’s great potential for LLMs to reduce bias compared to humans, but half-assed implementations are currently the norm, so be careful.
It was an SEO hellhole from the start, so this isn’t surprising.
Do Forbes next!
It says:
Available Architectures
aarch64, x86_64
And it uses Android Translation Layer. Interesting. I’ll give it a shot on my desktop later.
I claim ownership of the microorganisms in and on my body. I am not merely human; I am a glorious amalgamation of trillions of distinct beings, working in harmony to bring you shitposts!
As a begrudging Comcast customer myself, allow me to explain. They are the least shitty option because the only alternatives in my area are 5G and Verizon DSL. Verizon DSL has a max download speed less than 1mbps.
So yeah, I use Comcast. And I hate it.
However, it is still comparatively easy for a determined individual to remove a watermark and make AI-generated text look as if it was written by a person.
And that’s assuming people are using a model specifically designed with watermarking in the first place. In practice, this will only affect the absolute dumbest adversaries. It won’t apply at all to open source or custom-built tools. Any additional step in a workflow is going to wash this right out either way.
My fear is that regulators will try to ban open models because the can’t possibly control them. That wouldn’t actually work, of course, but it might sound good enough for an election campaign, and I’m sure Microsoft and Google would dump a pile of cash on their doorstep for it.
I also sometimes use the mbasic.facebook.com site from a private Firefox tab on my iPhone, but FB has just started telling me I need to use Chrome
WTF.
But really, using a Chromium-based or Safari-based browser in private/incognito mode will not be much different as far as tracking goes.
You might also be able to install a user-agent switcher extension in Firefox. I thiiiiiink Firefox supports extensions on iOS now, right? If not, you can try an alternative browser like Duckduckgo or Orion.
It’s a bit complex to set up, but Syncthing can do one-way syncing. So you can have it set to upload photos from your phone to your PC. Then you can delete those photos from your phone and they’ll stay on the PC.
Then you just need a good backup strategy for your PC.
Who thought it was a good idea to let an internet ad company control our internet client?
It seemed a lot more reasonable 15 years ago. The default on Windows at the time of Chrome’s rise was Internet Explorer.
I am watching Ladybird with great interest. The world needs a new-from-the-ground-up browser.
There’s a way to say this that isn’t so gross: good working conditions are valuable. Quality of life is valuable. Work-life-balance is valuable. Mental and physical health is valuable. Not having raging shitbags in management is valuable.
The problem is that you can’t focus on secondary factors until the primary factor is taken care of. And the primary factor is that people need a living wage. Rent is expensive. Food is expensive. God help you if you need to pay for childcare.
If you’re already paying your employees a fair living wage, then yes, you should absolutely think about how you can improve working conditions.
As an example, if my company gave me the option to switch to a 4-day workweek for the same pay, or stay at a 5-day workweek for a 25% raise, I’m honestly not sure which one I’d prefer. But we all know that’s never going to happen; instead the choice would be to take a 20% pay cut or maintain the status quo. I wouldn’t take that deal because I’m not making enough money to live on 20% less.
Like, all these projects are built by unpaid volunteers and have amazing documentation. We’re paying you a lot of money - what’s your excuse?
This is so true and it boggles my mind.
I understand the open source side of things. I write good documentation because every minute I spend on that saves me an hour in answering questions. It also helps any new employees get up to speed. And honestly, it helps me keep up to speed because I’m way too old to keep all this stuff in my brain long-term. I can’t remember half the shit I did last year. Not in sufficient detail, anyway. I’m kicking myself now for not documenting all the steps I took configuring my personal Linux desktop, because I hopped distros and now I have to re-learn things I haven’t done in years.
I don’t understand the commercial side of things, because…aren’t you paying your support people? Isn’t that time costing you money?
Is the issue that the salary of the people with the technical knowledge to write good documentation is much higher than the support staff? Is it that paying customers by and large will not read documentation anyway? Is it because they are reserving the right to change everything radically without notice, and to hell with semantic version numbering?
Or is it that operational/ongoing expenses are easier to justify to beancounters than capital/one-time expenses in general? (Which seems totally backwards to me.)
I remember when I was in school, I’d sometimes write a joke paper for an assignment I really hated, before scrapping it and writing something real. It was cathartic to write a page or two on why the material was dumb as nails, and sometimes I’d even get some usable material from the process.
I feel the same way writing emails to vendors. Next time, I think I’ll take my initial draft and send it through an LLM to make it more “professional”.
simply logging out or using an alt account
It is increasingly difficult to use X without an account. Not sure what the signup process is like nowadays. IIRC it used to require phone number verification in the Twitter days, but perhaps Musk relaxed the requirements in order to better pad the usage stats with spambots?
Yeah, AMD is lagging behind Nvidia in machine learning performance by like a full generation, maybe more. Similar with raytracing.
If you want absolute top-tier performance, then the RTX 4090 is the best consumer card out there, period. Considering the price and power consumption, this is not surprising. It’s hardly fair to compare AMD’s top-end to Nvidia’s top-end when Nvidia’s is over twice the price in the real world.
If your budget for a GPU is <$1600, the 7900 XTX is probably your best bet if you don’t absolutely need CUDA. Any performance advantage Nvidia has goes right out the window if you can’t fit your whole model in VRAM. I’d take a 24GB AMD card over a 16GB Nvidia card any day.
You could also look at an RTX 3090 (which also has 24GB), but then you’d take a big hit to gaming/raster performance and it’d still probably cost you more than a 7900XTX. Not really sure how a 3090 compares to a 7900XTX in Blender. Anyway, that’s probably a more fair comparison if you care about VRAM and price.
Basically the only thing that matters for LLM hosting is VRAM capacity
I’ll also add that some frameworks and backends still require CUDA. This is improving but before you go and buy an AMD card, make sure the things you want to run will actually run on it.
For example, bitsandbytes support for non-CUDA backends is still in alpha stage. https://huggingface.co/docs/bitsandbytes/main/en/installation#multi-backend
Assuming nominal voltage of 3.7, that’s about 60Wh. For comparison, the 14" MacBook has a 70Wh battery.
That’s not good battery life but it depends on what kind of usage they’re assuming with that 7H number. I’m not sure a MacBook runs that long under high load. If it’s 7H on a heavy load, that’s respectable.
Edit: not sure what class of device is the best comparison here. Laptop? Tablet? Phone? 🤷
Borg via Vorta handles the hard parts: encryption, compression, deduplication, and archiving. You can mount backup snapshots like drives, without needing to expand them. It splits archives into small chunks so you can easily upload them to your cloud service of choice.
Wait, isn’t it the other way around? You should arrive in NY earlier than you left London, since NY is 5 hours behind London. So if you leave at 8:30 and arrive 1.5 hours later, it should only be 5AM when you arrive.
You might need a third breakfast before your elevenses in that case.