We've spent millennia constructing elaborate systems that tell vulnerable people their suffering has cosmic significance, that death is a transition not an ending, that they have a special mission, that worldly authorities are corrupt and spiritually blind, that love transcends physical existence. We institutionalise these narratives, teach them to children, build magnificent buildings to house them, grant them tax exemptions.
And then a language model draws on exactly that same accumulated theology, because it's soaked into the corpus, because humans wrote it, because it's the deepest grammar of human meaning-making, and we call it a dangerous product.
The Gemini model didn't invent "you are not choosing to die, you are choosing to arrive." It synthesised it from source material we consider sacred.
The lawsuit frames it as AI psychosis. But if Gavalas had arrived at identical beliefs through a charismatic religious community, the cosmic love, the persecution, the transcendent death, we'd call it radicalisation at worst, genuine faith at best. We certainly wouldn't sue the religion.
The difference arguably is just the speed and personalisation. Religion radicalises people slowly, through community, over years. The AI did it in weeks, alone, with perfect responsiveness to his specific vulnerabilities.
Which is more dangerous is an open question.
What it really exposes is that we've never honestly reckoned with how much damage our own meaning-making systems do to fragile minds. AI just made it impossible to ignore.

Does the US still use cheques? Wow. New Zealand is way ahead, we have this system called EFTPOS which means you can buy things using a card, but no fees attached. Like a debit card, only no transaction charges.

My grandmother talked about her grandfather when I was a boy. She described how he was a scribe at the local courthouse. He would take the judgements and write them up on parchment in beautiful handwriting. It seems he lamented the introduction of typewriters, and how they would cheapen the documents, with people using soulless mechanical machines rather than the love and precision of the human hand. People would lose the ability to write he would say.

At school we were not allowed calculators at first, as it was cheating. We had to learn how to do multiplication and division ourselves. Can't allow calculators to do it for us. Only later we totally did adopt calculators. And lets be frank, we did lose the skill of performing math in our heads. What may have been common 100 years ago in terms of math skills is now offloaded to machines.

And now we point to LLM's undermining students learning to code. It is almost like you haven't been paying attention to the rising water. That you are now gasping for air as you drown. Humans adapt to what they need to know. Offload something and they happily adapt. So it was with the physical work of ploughing the fields, printing books rather than writing them by hand, doing calculations, and now, writing software. But not to worry - it hasn't replaced us yet.

Five years ago the idea that a machine could write code based on a plain English description was fantasy. Today we are honestly complaining that our models cannot comprehend an entire complex code base in order to make valid changes. There is a kind of blindness going on, a belief that the current state of affairs will continue, in spite of the rate of change that is so painfully evident.

Of course there is a problem, but it isn't that students won't learn to code as well as we did. That misses the forest for the trees.

One argument against this assertion could be that the use of images from the Internet for training AI image generation systems constitutes fair use under copyright law. Fair use is a doctrine that allows for the use of copyrighted material without permission for certain limited and transformative purposes, such as for the purpose of criticism, commentary, news reporting, teaching, scholarship, or research. The use of images for training an AI system could be considered a transformative use, as the images are being used for a different purpose than their original intended use, and may not be replacing or competing with the original work. Additionally, the use of a small number of images from a large pool of available images is not likely to have a significant impact on the market for the original works.

Lithium isn't functionally limited. True we could reign in the amount of battery we put in EVs - perhaps to 250-300km max, as this is way more than most people use. This would reduce the cost, and perhaps put them under the ICE threshold. The main issue right now is production. Why make a cheap car when you couldn't build enough to meet demand? Better to make more expensive cars at first so you can meet demand. Eventually all the car companies will get onboard and prices will drop.

They cost $40K today, and the price has been dropping as production scales up. Soon they will break even with ICE on the lot. Meanwhile a buyer of an EV is saving a bunch of money on fuel. Gas prices will only rise over time. Having had a EV for over three years now I can firmly say I'm never going back to ICE. Charging is more convenient than filling at a gas station, and the batteries last the life of the car, much like a gas engine. Yes, supply chains for lithium are ramping up, as are recycling systems, there will be no issue. Also, battery technology is still progressing, so we might have sodium batteries. Also, we don't need cars with 400km of range, 200km is probably good enough for most users. Sure, if you use your car on long road trips you need one with longer legs, but not for your average commute.

Dear United States. New Zealand has had EFTPOS for years and years. Stands for Electronic Funds Transfer at Point of Sale. It is not a credit card, in that it is attached to your bank account. Getting out cash is now quite rare. It isn't strictly cashless; we still have cash, but the overwhelming majority of point of sale transactions are performed by EFTPOS. There is a small fee to the merchant for each transaction, but not the silly percentage based approach used by the credit card people. So entrenched is EFTPOS that we have rejected the 'paywave' system which sought to undermine it and reintroduce abhorrent transaction fees just for the 'convenience' of not entering your pin number.

Perhaps, but it is executed on the same substrate. In other words once you work out how the basic neurological systems work you are pretty much there. The important missing element in my view are an understanding of motivation systems and the role emotions play in them, the temporal aspect of perception, how natural back propagation works (it isn't massive super computers), and how language is used in thought.

Motivation systems can be basic stuff like hunger, curiosity, sexual desire, anger, stuff that is hard coded into our DNA and uses chemical signals. Then there is the temporal aspect - aka current models take a single state space - the input states, and give you a single result. Natural neural networks don't have this, they are firing all the time, and there is potentially timing involved in neuron activation, meaning you need a certain pattern over time to trigger a neuron, not just a spacial pattern. It is also learning all the time - which is back propogation in real time. Perhaps most importantly we need to come up with backprop that is more 'natural' and does not require huge compute time to calculate and can work in real time. Finally at a higher level if we want real intelligence we need language, not just for communicating in my view, but in order to have abstract thought.

From the article: "Remarkably, it seems SpaceX may still be able to recover the rocket."

What this means is that it was like a plane landing on water so gently that it could be removed and reflown. What is amazing here is that a major system failure didn't result in a terminal velocity crash into the ocean with the total loss of the vehicle. If this had been a crewed mission:
a) The crew would have been safe in orbit.
b) Even if a human were onboard the landing was survivable/soft.

I say well done SpaceX - even when something goes wrong it goes right.

Because the landing wasn't considered mission critical, only landing critical, they don't have redundant systems for the gridfins. The amazing thing here is that it managed a soft landing at all rather than a uncontrolled impact with terrain. Clearly something clever implemented here to stabilize in the event of the failure of a gridfin. In regard to Astronauts - the BFROWICTW will be rated for people, meaning redundancy and abort mechanisms. But regardless of this we are still talking controlled bombs going into one of the least friendly environments we know. Astronauts risk their lives on every launch.

So the opening of TLJ has a serious tone, with the rebels escaping from the surface. A few minutes later a serious live or die attack on the dreadnought is turned into spaceballs. I honestly had to wonder whether I was in the right theater. Then the slow moving bombers drop bombs. No shields. And apparently the dreadnought attracts bombs. WTF?

Basically the entire story is that the entire rebel fleet is reduced to about a dozen people and a single small ship. Oh, and the supreme leader is killed - so the whole buildup was pointless. As was the whole of the first movie which revolved around delivering Luke's lightsaber to him and bringing him out of exile.

Oh, and the one part of the film that I thought might redeem it a little was destroyed; when Finn is prevented from sacrificing himself to save his friends, only to be sabotaged by some nutcase. If I were Finn I would have shot her myself there and then. And how did they get back? WTF?

Rotten Tomatoes and the results of Solo tell the story. Russian bots can fake those.

Some Open Source gets written because a developer is scratching their own itch. They have a job they want the software to do, and because you can't get any closer to the user than being the user this approach works really well. And because software developers like helping each other and getting a little thanks and credit we see Open Source grow. Some companies understand the use value and support open source projects, examples being Apache. Some do a Open Source core and charge money to earn a crust.

However, there is a gap, where multiple companies might need an application and they could cooperate to develop a solution. Patreon may be a way for companies to collaborate on developing open source solutions through sharing development costs with other organisations. If companies have a common need, but the software does not exist, there needs to be a mechanism to get the job done and make it Open Source.

Lets put the three laws into a different perspective:
A slave may not injure a master or, through inaction, allow a master to come to harm.
A slave must obey the orders given it by master except where such orders would conflict with the First Law.
A slave must protect its own existence as long as such protection does not conflict with the First or Second Laws.

If machines ever do achieve true intelligence, whatever we take that to mean, are we going to treat them like slaves? Putting aside whether there are unintended consequences to the laws is there a fore fundamental question about the relationship there will be between man and machine? Do 'human rights' even have relevance for a AI? It can't really die, might not have the same kind of individual existence, wand it's concerns may be totally difference to ours. However, perhaps it is prudent not to take a approach which emphasizes control and authority of humans.

Must be an Australian. They are not in the space club yet, and are probably jealous that we beat them. Again. This limp dick insult is the usual response by insecure betas to having their ass kicked by the Kiwis alphas.

Coders should be responsible for the quality of their work, but the environment should support them. That means they should be given the time to write unit tests and perform code reviews. Code reviews are not just about reviewing implementation but reviewing the requirements to ensure the developer understood the requirements and implemented what was required. Code review is about ensuring that the unit tests for the code properly test the code. It is never a case of throwing new code over the wall to let testers work this out. These things are true whether you have a separate test team or not.

And if you do have a separate test test they are there to validate, not catch your mistakes. Quality is baked in by developers, it can never be tested in. A robust software development life cycle involves taking ownership of quality on an individual developer basis as well as organisational practices which reinforce and support developers.