scarmig

48 ms·

781.
by scarmig 4y ago
I worry more about the 3%.
782.
by scarmig 4y ago
It takes lots of smarts to fool yourself.
783.
by scarmig 4y ago
Just train a model to break the returned text into a set of claims, evaluate each claim for likelihood of being an ad, and then excise those claims that are likely ads.
784.
by scarmig 4y ago
> google’s private compute
785.
by scarmig 4y ago
Google has far more resources for training models and inference. Likely more than all their nearest competitors combined.
786.
by scarmig 4y ago
It's an image similarity search engine, slapping on the tag SD attribution for marketing purposes. Note that just because an image is similar (to human eyes) doesn't mean that it played a more significant role than a seemingly mor
787.
by scarmig 4y ago
The best way to fleece a flock is to wrap yourself in a cloak of good intentions. You see it in EA, in churches, in Black Lives Matter. Basically as soon as someone proclaims themselves as beyond criticism because of their superior moral po
788.
by scarmig 4y ago
If an organization believes spending a marginal dollar of money on their programs is the best way to improve the world, then spending $10 to get $11 in donations allows them to spend an extra dollar on it. It's rational and even morall
789.
by scarmig 4y ago
The issue with Reddit is that the story is as likely as not to be fake. Particularly, here, I don't think people are at risk of serious withdrawal after only two weeks of heroin use. Though heroin use is obviously one of the dumbest th
790.
by scarmig 4y ago
Does this work with any SAML IdP? E.g. a private Keycloak instance?
791.
by scarmig 4y ago
The difference: When nuclear energy goes wrong, thousands of people die. When fossil fuels go right, millions of people die.
792.
by scarmig 4y ago
In practice, linear regression is highly effective and likely to be better founded than your million parameter neural net.
793.
by scarmig 4y ago
Ancient.
794.
by scarmig 4y ago
Closest thing I can find that seems relevant in there: "Compared to overall private industry, the high tech sector employed a larger share of whites (63.5 percent to 68.5 percent), Asian Americans (5.8 percent to 14 percent) and men (5
795.
by scarmig 4y ago
Probably like 75% male. But at most 25% white male.
796.
by scarmig 4y ago
Well, urban educated women need protections too. The bitter pill to swallow is that there are people of all sexes and demographics who are shitty out there and whatever set of laws there is will be exploited by them for their own ends. Some
797.
by scarmig 4y ago
That's generally the right way to approach things. And fortunately it seems like there will be some kind of paper trail here, so it should be relatively easy for it to be resolved quickly (probably settled out of court, if his descript
798.
by scarmig 4y ago
There's a tension between the belief espoused by many on this thread that a man struggling dating is purely a matter of him being an unkempt, terrible person and the fact that arbitrary things like race can make dating a struggle.
799.
by scarmig 4y ago
The flawed assumption here is that every match a guy gets is a wonderful, invested woman, while the vast majority of the matches women get are evil depraved men. In practice, the proportion of good matches are similar for men and women. The
800.
by scarmig 4y ago
Well, if you copied a famous piece of art from memory... That'd be a copyright violation.
801.
by scarmig 4y ago
For the concrete situation I mentioned above, they were in a part of Google that was profitable, but just not profitable enough. (The Google graveyard is going to have a lot more tombstones over the next couple months.) But, yeah: all oth
802.
by scarmig 4y ago
At least at a big company, RIF decisions are made at higher levels than anyone who knows anything about you or your work. Obviously people on PIP are in trouble, but beyond that there's not much you can do, unless you're Jeff Dean
803.
by scarmig 4y ago
The cultural and moral beliefs of predominantly white, college-educated professionals living in major metropolitan areas of the USA in 2023 map one-to-one with the objective Truth, of course, unlike all rest of people through space and time
804.
by scarmig 4y ago
I'm just excited that soon I'll be able to deploy my own scarmigGPT to argue against all the other GPTs that disagree with me!
805.
by scarmig 4y ago
I don't think that's an accurate statement of facts on the ground. For instance, if you poll the general population about trans women participating in sports, a large majority of respondents are opposed to it: https://w
806.
by scarmig 4y ago
No one has a total claim on truth; worse than that, people who have wildly diverging opinions from truth are more likely to hold them very strongly and will be upset when the model tells them they're wrong.
807.
by scarmig 4y ago
Indeed. Though, in fairness, an unbiased model would probably end up alienating closer to 100% of people instead of any particular half of them.
808.
by scarmig 4y ago
Only a miniscule part of reality is digitized, and what data does exist passed through the biases of people before being available to train on.
809.
by scarmig 4y ago
It would depend on what exactly I was building. Maybe it needs to be able to generate texts on controversial topics. I agree that it alienates people, but the choice is less between alienating half and alienating no one but more alienating
810.
by scarmig 4y ago
My result there was pretty reasonable and on point, actually, without any fluff or ideological throat clearing about how he's evil.

More ›