Good time to subscribe to @wired.com www.wired.com/story/ai-saf... - ThreadSky

timothysnyder.bsky.social • 9 hours ago

Good time to subscribe to @wired.com
https://www.wired.com/story/ai-safety-institute-new-directive-america-first/

Comments

Idealogical bias is a code word for stuff that makes Trump and conservatives look good. However, you can't teach an AI anything that isn't in it's training materials, nor easily affect how it makes connections between things.

jacquelinerussell.bsky.social•9 hours ago

Thank you. Done!

tarotbysherri.bsky.social•9 hours ago

New subscriber here 🖐️ Wired has been doing a great job!

ceu---azul.bsky.social•2 hours ago

rxdavid.bsky.social•9 hours ago

Shades of HAL the computer in 2001: A Space Odyssey.

tomdecanio.bsky.social•9 hours ago

Geezus. If American made LLMs turn into nothing but right wing zealots I guess I might have to look to Deepseek and LLMs built elsewhere. What an unexpected turn.

sorentsvendsen.bsky.social•9 hours ago

French-made Le Chat is an option :)

philfeller.bsky.social•6 hours ago

I found it somewhat disturbing that one of the quoted sources in this piece asked “what does it even mean for humans to flourish?” Sure, the instructions’ use of the phrase is clumsy, but shouldn’t an educated person be aware of its origin in Aristotle’s ethics?

terimorrow.bsky.social•7 hours ago

Read this and then listen to this: Business: Elon Musk resembles the techno-fascists of 1930s Japan https://slate.com/podcasts/slate-money/2025/03/business-elon-musk-techno-fascists-japan-silicon-valley-gold-card-donald-trump-plastic?utm_medium=social&utm_campaign=traffic&utm_source=article&utm_content=web_share

kilocharlie237.bsky.social•7 hours ago

At this point the US looks WORSE than Germany before the beginning of WW 2.
For all of trump's threats, EO's, he can't put anyone in prison or do anything to their family or property. The worst he can do is possibly, fire them.
And yet, the compliance is off the chart.

dustyknews.bsky.social•9 hours ago

Everyone has AI now. China, Canada now operate their own comparable models at a fraction of the cost. The U.S. built a bubble that’s now collapsing.

alkisky.bsky.social•9 hours ago

“Responsible A.I.” will do exactly those things. The computers are going to look at that directive and conclude that the people behind it aren’t very bright.

jaimehylton.bsky.social•8 hours ago

Will do.

fvlib.bsky.social•9 hours ago

While the corporate political press and elite DC punditry are still looking at the horse race and both sides, WIRED has the DOGEbags and Musk's game nailed. They see right thru all of it.

zimadood.bsky.social•8 hours ago

IMO: Trump & Heritage working towards ripping up the American Constitution.
Ironic, it was the Supreme Court that gave him all the powers of a King.
Even more ironic, America's Semiquincentennial, the 250th anniversary of the United States Declaration of Independence, is only one year away in 2026.

historyisimportant.bsky.social•8 hours ago

Scientists do not deal in ideological work. 🍊💩 does, and he does not want you to be smarter than him.

jaimehylton.bsky.social•8 hours ago

That ship has sailed.

pammcnamara.bsky.social•9 hours ago

Done

billga2603.bsky.social•8 hours ago

Ie praise Trump for robbing us.

aguasonic.com•9 hours ago

You don't have to think. We will tell you what to think.

aguasonic.com•9 hours ago

Except bias <> be removed from Large-Language-Models. They reflect the corpus of what was ingested, and the rules for parsing it written by their ( biased ) creators.

dooleyappointed.bsky.social•9 hours ago

It's scary that they want this, but it's interesting to go into even Musk's Grok and ask who the biggest threat to the world is and so forth. It gives a pretty honest answer, so far. In a sequel to 2001, Clarke shows that HAL went insane because they forced him to lie.

gripnoir.bsky.social•6 hours ago

I did in the 2nd week of this administration. It's - connected!

mightyapollo.bsky.social•8 hours ago

Just what we need, LLM models altered to be even less accurate.

steveknapp.bsky.social•9 hours ago

Bias, like beauty, is often in the eye of the beholder. Facts are your firewall against bias.
Tom Brokaw

socialmediaman.bsky.social•9 hours ago

Perceived bias, sure, actual bias, less so. People on the right often claim facts are biased against them.

therese-ekstrom.bsky.social•8 hours ago

Probably why Trump is targeting education aswell.

anthonymugan.bsky.social•6 hours ago

When major powers give up reality for ideology the result is never good for them or millions of ordinary people (cf Catholic Church 17thC, Soviets, 3rd Reich, Islamic societies 20thC). Golden chance for Europe. Thuggishness leads to easy consent to hard choices necessary..if we can get act together

waz001.bsky.social•4 hours ago

This feels like the start of Skynet....

theseneschal.bsky.social•8 hours ago

I’ll remove my model’s ideological bias if Trump removes his ideological bias.

rwhetsel.bsky.social•8 hours ago

Ubuntu sounds helpful.

"Today S. Africa is different from when Rosieda & Ebrahim wed. It has moved past apartheid into a reconciliation movement supported by a national ubuntu in some ways new to the world."

https://en.wikipedia.org/wiki/Ebrahim_Rasool

https://en.wikipedia.org/wiki/Ubuntu_philosophy

rltjr.bsky.social•9 hours ago

Not sure to go with, “What could go wrong?” or, “Hello Sky Net.” This is maximizing AI for maximum short term profit modeling. This means your emotions, preferences, biases, habits, demographics, etc will be modeled to make your life more profitable FOR SOMEONE ELSE. You’re an opportunity or a risk.

cerberus25.bsky.social•9 hours ago

I think it’s time people started saying no.

cerberus25.bsky.social•9 hours ago

By that I mean saying no to this administration, not to a subscription to Wired.

amethystjones1.bsky.social•6 hours ago

It is most unhelpful that none of those jerks understand big words, and why some big words sound like other words, but have nothing to do with the dreaded wokeness they fear. To quote Gomer: goll-lee.

therese-ekstrom.bsky.social•8 hours ago

This is REALY dangerous!

jaredlipof.bsky.social•3 hours ago

In this context, “ideological bias” = “not conservative bias”

wayno62.bsky.social•9 hours ago

He hates science and education. Doesnt read

carolannete.bsky.social•9 hours ago

To the Trumper mind, "ideological bias" means "disagrees with me."
The people giving Trump his agenda believe that the federal government should promote & enforce their own biases.

mvde1975.bsky.social•6 hours ago

Sigh, will this nightmare ever end?😔

thetasalmon.bsky.social•5 hours ago

The only way you can impose a preferred political bias on AI is by forcing it to ignore reality and fail to provide the best answer… so you get your preferred answer.

Watch China eat your AI lunch forever from this point on.

This is why IVF Space Karen’s Grok AI is inevitably doomed… thankfully.

boxomcfoxo.bsky.social•3 hours ago

LLMs never provide the best answer anyway. They provide the most likely predicted answer.

mitchrm.bsky.social•9 hours ago

This interview with Maria Ressa discusses the effects of generative AI and propaganda.
https://bsky.app/profile/mitchrm.bsky.social/post/3lk4slwiyfc2v

leahfeiger.bsky.social•7 hours ago

thanks timothy <3

drvickyharris.bsky.social•9 hours ago

Really. With Trump on the cover and NO unequivocal condemnation of his crimes? Nah.

charjg.bsky.social•9 hours ago

already did and got some cool stickers

a-climate-refugee.bsky.social•9 hours ago

As if there’s not already plenty of misinformation?

jae-bond-007.bsky.social•9 hours ago

No msm tv coverage with protests nationwide, but we'll take care of those folks later! Busy now in NYC
https://www.youtube.com/live/D75u1uvSLfk?si=dBVx_R-Pack_JMBi

southernlibbing.bsky.social•9 hours ago

AI for propaganda scares the shit out of me.

dropthet.bsky.social•2 hours ago

Well it already surrounds you. Of the 1,600 people you follow at least 1/4 of them are bots. Any of those accounts that follow 50,000 people & have 49,000 followers, 99% of those are bots. And many others. All just to influence you.

dropthet.bsky.social•2 hours ago

And glancing at the first 5 people you follow I would place a large bet on it being at least half of them, actually.

For some reason, we are only scared of the "AI" we disagree with.

therese-ekstrom.bsky.social•8 hours ago

The conclusion should be not to use AI other when it's absolutely necessary, but that will not happen.

basalticintrusion.bsky.social•2 hours ago

Glad I didn't delete my X account. I've been playing with Musk's "GROK" AI (R. Heinlein rolls over in his grave every time I write those last 3 words.)
Grok appears to lack ideological bias right now. Perhaps Trump wants Musk to "fix" it, and then together they'll go after the other major AI apps.

vasya-b.bsky.social•1 hour ago

used car seller is talking about AI?

jake200.bsky.social•9 hours ago

This is totalitarianism.

cameroncanada.bsky.social•9 hours ago

I did yesterday. Lots of great reporting.

casssax.bsky.social•8 hours ago

They don’t want all ideological bias removed. They want theirs to be imposed

keporkak.bsky.social•2 hours ago

👍

gobosox54.bsky.social•9 hours ago

I couldn’t pass up that offer! Thx

margiem.bsky.social•9 hours ago

For the next month any donations you were going to send to SCHUMER, Gillabrand and pals donate to the @ACLU. Unlike Chuck they’re fighting for our rights.

https://action.aclu.org/give/now

pauluspeters.bsky.social•9 hours ago

As if predatory neoliberalism isn't ideological? You'll get into the murky world of Bannon, Dugin and such, unfortunately, but deliberately erasing corrective oversight IS ideological, just not of a benign kind.

rozreckons.bsky.social•9 hours ago

Great - the same AI that thinks kosher meat depends on the religion of the cow

sharoneis.bsky.social•8 hours ago

Done.

Comments

Posting Rules

Reply