AmbiguousProps

joined 2 years ago
[–] AmbiguousProps@lemmy.today 4 points 9 hours ago

I use last.fm scrobbling so it can learn what I like and provide suggestions based on my locally hosted music.

[–] AmbiguousProps@lemmy.today 13 points 9 hours ago (2 children)

I pray that my Pixel 9 is my last Pixel.

[–] AmbiguousProps@lemmy.today 4 points 4 days ago

This is what this user does, for the record. They don't care about anything anyone else says, they must be 100% correct or they turn to insults. You may want to block them if you care about actual discussion.

[–] AmbiguousProps@lemmy.today 2 points 4 days ago* (last edited 4 days ago)

Holy shit, you're too far gone. Yes, I've operated my own local LLM front end for personal use and have run them frequently in the past at a low level with no front end and messed with parameters directly. I've modified models as well, and I have a server designed to run them. It's insane that you think those apply to his topic (billionaires lying to you). You're just throwing random terminology out to make yourself seem smart and to reinforce your stance, but all that does is make you seem insecure about your knowledge. It does quite the opposite of what you intend.

All LLMs do is hallucinate, and sometimes, by pure coincidence, get things correct. This is why it's impossible to get rid of hallucinations. They do not think, they do not have their own goals (refusing a request can be baked in, but that's no different than programming something with guardrails), and they certainly will not suddenly become sentient. LLMs cannot do that, by design. Perhaps something else could, but LLMs are not that.

You really had to go to insults for this? I think you need to touch grass and stop believing billionaire marketing. LLMs are not technologically capable of doing what you have been brainwashed to believe. They will crash the economy when the bubble bursts, because MBAs and billionaires have convinced you and rich VCs that they can do more than they actually can.

[–] AmbiguousProps@lemmy.today 2 points 4 days ago (2 children)

"what they are going to do" is exactly that: marketing. It's not technically feasible, they just want to ride the bubble a little longer and convince people otherwise (like they've convinced you)

[–] AmbiguousProps@lemmy.today 34 points 5 days ago (5 children)

We test in production, silly.

[–] AmbiguousProps@lemmy.today 1 points 5 days ago* (last edited 5 days ago)

Yes, but is that AI's personal goal, or is that the goal of the billionaire class? OP is saying that AI is "getting goals" as if current AI has the ability to have goals of its own.

[–] AmbiguousProps@lemmy.today 3 points 5 days ago (7 children)

Just because they want that doesn't mean they understand how it works. It's not currently feasible from a technical and resource perspective.

[–] AmbiguousProps@lemmy.today 3 points 5 days ago (4 children)

Are the goals in the room with us?

[–] AmbiguousProps@lemmy.today 49 points 5 days ago* (last edited 5 days ago) (1 children)

Depends on your definition of "smart' I guess. ZigBee stuff like buttons and the like probably won't become obsolete for a long time. I guess you could argue that ZigBee protocol updates could eventually brick them though. Good thing a lot of it is open source

[–] AmbiguousProps@lemmy.today 9 points 6 days ago (2 children)

I just hate that Mapillary is owned by Facebook.

[–] AmbiguousProps@lemmy.today 9 points 6 days ago* (last edited 6 days ago)

Unfortunately not all locations have that, such as Western WA at the moment. I say this as an EV owner, I wish we had that. I charge at night anyway, and home charging is still cheaper than gas.

 

Wildland fire veterans are seething at a claim made by federal officials that two crews raided by immigration agents at the scene of a wildfire in Washington state were “NOT firefighters.”

Many political figures and media outlets have repeated the claim, even though public documents show the crews have firefighting classifications and were assigned to key frontline roles battling the blaze.

“Everybody in the profession sees through it, but the public doesn’t and that’s concerning,” said Riva Duncan, a former wildland fire chief who served more than 30 years with the U.S. Forest Service. “It’s a lie. Everybody I’ve talked to is very upset about it. It does not just those two crews a disservice, but it does all firefighters a disservice.”

Duncan also serves as vice president of Grassroots Wildland Firefighters, a nonprofit that advocates on behalf of wildfire professionals.

Last week, federal immigration officials staged a raid at the site of the Bear Gulch fire in Washington, the largest active wildfire in the state. Agents with U.S. Customs and Border Protection questioned two crews, totaling 44 members, and arrested two firefighters who they said were in the country illegally.

But the facts clearly show that the crews were firefighters. In planning documents drafted by the management team overseeing the fire and posted to a public federal database, the crew from contracting company ASI Arden Solutions, Inc., is listed as a “CR2I” crew. That’s shorthand for a Type II Initial Attack wildland firefighting crew.

“They’re just one level below a hotshot crew,” Duncan said. “[Saying they’re not firefighters] is incredibly insulting to them.”

The other crew, from contracting company Table Rock Forestry, Inc., is listed as a “CRW2,” short for a Type II wildland firefighting hand crew. That means both crews were certified under National Wildfire Coordinating Group standards as firefighters who met rigorous qualifications and held “red cards” verifying their status to fight fire.

Additionally, the documents show that both crews were assigned to active firefighting roles in the days leading up to the raid. The crews were tasked with securing the fire edge, protecting structures, constructing fire lines and addressing hazards caused by the initial suppression work.

Many wildfire veterans who have served in similar roles privately expressed anger that the crew’s status was called into question because they had been assigned to cut firewood on the day of the raid. That frustration is heightened by the widespread belief, shared by many fire professionals, that the crews were given that assignment under false pretenses to lead them into contact with federal immigration agents.

“They were doing suppression work, and it was only when they were reassigned that day [that they were raided],” Duncan said. “To paint this picture that they would never do that to actual firefighters, it’s total spin.”

More in the article.

Archive link

 

Two people fighting the Bear Gulch fire on the Olympic Peninsula were arrested by federal law enforcement Wednesday, in a confrontation described by firefighters and depicted in photos and video.

Why the two firefighters were arrested is unclear. But a spokesperson for the Incident Management Team leading the firefighting response said the team was “aware of a Border Patrol operation on the fire,” that it was not interfering with the firefighting response and referred reporters to the Border Patrol station in Port Angeles.

Over three hours, federal agents demanded identification from the members of two private contractor crews. The crews were among the 400 people including firefighters deployed to fight the wildfire, the largest active blaze in Washington state.

It is unusual for federal border agents to make arrests during the fighting of an active fire, especially in a remote area.

While they were waiting for their division supervisor to arrive, federal law enforcement showed up around 9:30 a.m.

Federal law enforcement asked the crews to line up to check their IDs, according to the firefighters.

One of the firefighters said members of the crew were told not to take video of the incident.

“You risked your life out here to save the community,” the firefighter said. “This is how they treat us.”

More in the article.

Archive: https://archive.is/lcEnX

46
submitted 2 months ago* (last edited 2 months ago) by AmbiguousProps@lemmy.today to c/news@lemmy.world
 

The Mason County sheriff’s office ordered people near Olympic National Park to evacuate as the Bear Gulch fire grows.

The human-caused fire was first reported early this month near the park and has since grown to cover 2 square miles on the northern shores of Lake Cushman. It is less than 10% contained.

Anyone near the Dry Creek Trail, along the lake’s westernmost shores should evacuate immediately, the sheriff’s office said on social media. This is called a Level 3 evacuation order.

Archive link: https://archive.ph/ztkUn

Such a shame. The staircase area is (was) beautiful, and probably my favorite part of Olympic National Park. Tons of old growth rainforest is now gone, and the fire won't be out until it snows, according to officials.

Here's a link to the watchduty listing for the fire, has much more info: https://app.watchduty.org/i/54759

 

With the recent first light milestone for the Vera Rubin Observatory, it's only a matter of time before one of astronomy's most long-awaited surveys begins. The Legacy Survey of Space and Time (LSST) is set to start on November 5, and will scan the sky of billions of stars for at least ten years.

One of the most important things it aims to find is evidence (or lack thereof) of primordial black holes (PBHs), one of the primary candidates for dark matter. A new paper posted to the arXiv preprint server by researchers at Durham University and the University of New Mexico looks at the difficulties the LSST will have in finding those enigmatic objects, especially the statistical challenges, and how they might be overcome.

 

Japan on Sunday successfully launched a climate change monitoring satellite on its mainstay H-2A rocket, which made its final flight before it is replaced by a new flagship model designed to be more cost competitive in the global space market.

The H-2A rocket lifted off from the Tanegashima Space Center in southwestern Japan, carrying the GOSAT-GW satellite as part of Tokyo's effort to mitigate climate change. The satellite was safely separated from the rocket and released into a planned orbit about 16 minutes later.

Scientists and space officials at the control room exchanged hugs and handshakes to celebrate the successful launch, which was delayed by several days due to a malfunctioning of the rocket's electrical systems.

Keiji Suzuki, a Mitsubishi Heavy Industries official in charge of rocket launch operations, said he was more nervous than ever for the final mission of the rocket, which has been his career work. "I've spent my entire life at work not to drop H-2A rocket ... All I can say is I'm so relieved."

 

For years now, U.S. police departments have employed officers who are trained to be experts in detecting "drugged driving." The problem is, however, that the methods those officers use are not based on science, according to a new editorial in the Journal of Studies on Alcohol and Drugs (JSAD).

With marijuana now legal in many U.S. states, the need for reliable tests for marijuana impairment is more pressing than ever. Police can evaluate alcohol-intoxicated drivers by using an objective measure of breath alcohol results. But there is no "breathalyzer" equivalent for marijuana. The drug is metabolized differently from alcohol, and a person's blood levels of THC (the main intoxicating chemical in marijuana) do not correlate with impairment.

So law enforcement relies on subjective tactics—roadside tests and additional evaluations by police officers specially trained to be so-called drug recognition experts (DREs). These officers follow a standardized protocol that is said to detect drug impairment and is said to even determine the specific drug type, including marijuana.

The process involves numerous steps, including tests of physical coordination; checking the driver's blood pressure and pulse; squeezing the driver's limbs to determine if the muscle tone is "normal" or not; and examining pupil size and eye movements.

But while the protocol has the trappings of a scientific approach, it is not actually based on evidence that it works, said perspective author William J. McNichol, J.D., an adjunct professor at Rutgers University Camden School of Law.

 

cross-posted from: https://lemmy.ca/post/46641802

 

Used a 12 inch bit. It's a great workout, but really sucked when we encountered tree roots with it. Tomorrow, I'm going to set some posts in concrete using the holes.

 
 

The number and diversity of insects is declining worldwide. Some studies suggest that their biomass has almost halved since the 1970s. Among the main reasons for this are habitat loss—for example through agriculture or urbanization—and climate change.

These threats have long been known. What is less well-known is how these global change drivers interact and how their effects can become even more severe that way. For example, insects that have been deprived of their natural habitat could be even more affected by higher temperatures in a new environment.

Researchers at Julius-Maximilians-Universität Würzburg (JMU) have investigated precisely this serious interaction at 179 locations throughout Bavaria. The study is part of the LandKlif research cluster, coordinated by Professor Ingolf Steffan-Dewenter within the Bavarian Climate Research Network bayklif.

They published their results in the journal Proceedings of the Royal Society B: Biological Sciences.

 

In 2019, Tesla set out to lower insurance rates for owners of its electric cars. The goal was simple, at least in theory: fix the broken cost of car insurance. Instead, Tesla may have broken its own calculator trying to make sense of repair costs.

See, Musk's vision of Tesla's insurance product was that traditional companies just didn't "get it." Tesla's data claims that its Full Self-Driving software has fewer accidents than a human driver. Plus, its cars are rolling computers that can collect copious amounts of data on its drivers and adjust risk based on their driving. So why wouldn't drivers get a lower rate for putting around with FSD enabled if they also happen to be a safe driver? Tesla quickly found out that despite these assumptions, it's still taking a bath on claim-related losses.

The data comes from S&P Global and shows that the automaker's insurance subsidiary took a loss ratio of 103.3 in 2024. The loss ratio, for those who don't know, is the amount of money that Tesla pays out per claim versus the money it takes in from premiums. The lower the number, the better, and break-even is a flat 100. In 2024, the rest of the industry averaged 66.1.

Archive link: https://archive.is/G4Kvj

view more: next ›