Skip Navigation

InitialsDiceBearhttps://github.com/dicebear/dicebearhttps://creativecommons.org/publicdomain/zero/1.0/„Initials” (https://github.com/dicebear/dicebear) by „DiceBear”, licensed under „CC0 1.0” (https://creativecommons.org/publicdomain/zero/1.0/)F

FlowVoid

@ FlowVoid @lemmy.world

Posts
1
Comments
1301
Joined
2 yr. ago

  • AI company Anthropic agrees to pay $1.5B to settle lawsuit with authors

    Jump

  • A new model needs training data, it doesn't matter if the data is new or old. But generally, a more advanced model needs more training data, so AI devs generally need at least some new training data.

  • AI company Anthropic agrees to pay $1.5B to settle lawsuit with authors

    Jump

  • No, the old model does not have the training data. It only has "model weights". You can conceptualize those as the abstract rules that the old model learned when it read the training data. By design, they are not supposed to memorize their training data.

    To outperform the old model, the new model needs more than what the old model learned. It needs primary sources, ie the training data itself. Which is going to be deleted.

  • AI company Anthropic agrees to pay $1.5B to settle lawsuit with authors

    Jump

  • It's true that a new model can be initialized from an older one, but it will never outperform the older one unless it is given actual training data (not necessarily the same training data used previously).

    Kind of like how you can learn ancient history from your grandmother, but you will never know more ancient history than your grandmother unless you do some independent reading.

  • AI company Anthropic agrees to pay $1.5B to settle lawsuit with authors

    Jump

  • They will need actual training data when they want to develop the next version of their LLM.

  • AI company Anthropic agrees to pay $1.5B to settle lawsuit with authors

    Jump

  • I doubt it. How would they keep disgruntled ex-employees from spilling the beans?

  • AI company Anthropic agrees to pay $1.5B to settle lawsuit with authors

    Jump

  • They were also forced to delete the pirated works from their training data. Which means they might have problems training the next version of their LLM, and if so that revenue is going to dry up.

  • President Trump says he may send National Guard to New Orleans next instead of Chicago

    Jump

  • No, he said "ICE, Customs and Border Patrol, the Department of Homeland Security, and other similarly situated federal agencies" had started staging.

  • President Trump says he may send National Guard to New Orleans next instead of Chicago

    Jump

  • Pritzker said he received information about Trump's plans, not that the Guard already deployed.

  • President Trump says he may send National Guard to New Orleans next instead of Chicago

    Jump

  • What weapons? Illinois has very strict gun control, including an assault weapons ban.

    There are exceptions for federal officers. But if Trump can't find a way to federalize the Texas Guard...

  • President Trump says he may send National Guard to New Orleans next instead of Chicago

    Jump

  • I doubt that's Texas National Guard, everyone would know if they already deployed.

    Much more likely that Great Lakes is hosting a surge of ICE agents

  • Leaked ‘Gaza Riviera’ plan dismissed as ‘insane’ attempt to cover ethnic cleansing

    Jump

  • Dismissed by whom?

  • Democrats vote down resolution calling for arms embargo on Israel, recognizing Palestinian state

    Jump

  • The previous genocide wasn't done by Jews, and barely anyone noticed.

  • Democrats vote down resolution calling for arms embargo on Israel, recognizing Palestinian state

    Jump

  • I thought it was straightforward.

    Democrats voted on a resolution for an arms embargo, and the vote failed.

  • Scientists Have Translated the Inner Monologue. The Implications Are Incredible.

    Jump

  • Normally words are all trained together, not one after another. Subjects do a monitored task that includes the entire vocabulary list, data from that task is processed extensively off-line, and finally subjects return to test all the vocabulary words.

  • Scientists Have Translated the Inner Monologue. The Implications Are Incredible.

    Jump

  • Just so it's clear, each subject spent hours training their own AI to recognize words they intentionally produced as "inner voice". At the end, each trained AI could only recognize those trained words, and only from their trainer.

    That's quite a bit different than an AI that could recognize new words in a completely different person. I think that's as unlikely as training ChatGPT using only English texts and then expecting it to understand Chinese.

  • Deleted

    Permanently Deleted

    Jump

  • Of course Swiss law matters. This article is on a Swiss website and reports on a Swiss trial.

    If you think someone convicted of rape is not necessarily a rapist, that's on you.

  • Deleted

    Permanently Deleted

    Jump

  • The only definition that matters is the one used by Swiss courts. And Swiss courts found him guilty of rape.

  • Is moderate drinking actually healthy? Scientists say the idea is outdated.

    Jump

  • This was a prospective study that looked at lifelong abstainers, not people who gave up drinking. Furthermore, reduced overall mortality was found in light/moderate drinkers even after excluding participants with pre-existing conditions.

    That's the correct way to control for the issue you raised.

  • Is moderate drinking actually healthy? Scientists say the idea is outdated.

    Jump

  • Despite what the linked article claims, it's still not clear whether alcohol's known cancer risk is outweighed by any health benefits (for example, reducing the risk of stroke/CVD). At least one recent paper concluded that light or moderate alcohol use does reduce overall mortality.

    Compared with lifetime abstainers, current infrequent, light, or moderate drinkers were at a lower risk of mortality from all causes [infrequent—hazard ratio: 0.87; 95% confidence interval: 0.84 to 0.90; light: 0.77; 0.75 to 0.79; moderate 0.82; 0.80 to 0.85], CVD, chronic lower respiratory tract diseases, Alzheimer’s disease, and influenza and pneumonia. Also, light or moderate drinkers were associated with lower risk of mortality from diabetes mellitus and nephritis, nephrotic syndrome, or nephrosis. In contrast, heavy drinkers had a significantly higher risk of mortality from all causes, cancer, and accidents (unintentional injuries).