An average fill up runs at about 34 megawatts.
Most of that energy is lost because ICE are very inefficients. Still, impressive.
Also, it’s not possible to refill at home, and it’s expensive when there’s a war near an oil-producing country at the other side of the world.
Electrifying has this downside of slow recharge, but quite a lot of benefits as well.
And I guess engineers would be held responsible for the code produced by the AI agent’s they’re pressured to use.
So management can blame and fire more engineers when things go wrong.
Good. This may reduce the amount of sloppy code being created. And prevent prices from increasing for everyone.
That’s a hard pivot. And building more AI infrastructure is a bad idea. But they might make some short term money given the AI hype.
Low to average reliability is fine if the service is cheap, and if that avoid the need for backup diesel generators in datacenters.
I doubt this applied to Github:
Microsoft to use diesel-fired generators as backup power for data centers
That’s a good way to represent LLMs. Very bad and very prolific consultants.
It shows LLMs can do significant harm without the capabilities of an AGI.
Overhyping LLMs and overinflating their capabilities makes things worse, as people are less skeptical of LLM output.
According to Clayton, the AI agent involved didn’t take any technical action itself, beyond posting inaccurate technical advice, something a human could have also done.
Producing innaccurate technical advice, with a confident tone, at scale.
If that LLM were an employee it would get a formal blame, and then demoted or fired as it continues.
Rumors; didn’t read.
Anyway the last few versions of Windows already convinced it’s best to keep avoiding it.
There’s still someone at Microsoft with common sense. That’s probably too little too late.
Using an ultrasonic dog trainer in public may piss off dogs, and other domesticated or wild animals in the vicinity. I woudln’t recommend this, except maybe as a last resort on rare occasions.
Politely asking the person to use a headphone or earpiece may be more effective in many case.
The only justification you could possibly have would be that if we don’t do it, our adversaries will do it. And we will be subject to their rule of law.…
Quick, undermine democratic values and rule of law before someone else does!
The platform that works closest to this is https://www.liberapay.com/
It used to allow making 1 monthly payment and spreading it between projects. But they had to change to comply with banking/payment platform rules so each money transfer is directed to one beneficiary from the get go. Meaning you setup a recurring payment per project.
There’s also the nlnet foundation https://nlnet.nl/
It’s possible to give them 1 recurring payment to support OSS projects. But they manage how the money is split between projects and their own overhead. You don’t pick the projects.
Separating data structure from implementation has benefits.
In languages with classic OOP classes and objects, it’s often necessary to write wrappers or adapters to allow new operations on existing objects. This adds overhead and require more code.
We don’t need another centralized messaging service in the EU. We need a secure and decentralized one. Such protocol already exist such as XMPP, Matrix, Briar, Ricocher, RCS, …
Yes, this is fucked.
I doubt a sane society would aim to replace as many jobs as possible with automation, or necessarily be happy with it. Making this a goal mean trying to remove humans and unions out of the equations.
A sane society may seek to decrease workplace injuries and be more efficient, ie wasting less resources while producing stuff. That could involve better workplace conditions, better product design, and maybe automation. Automation may incidentally replace some tasks, even though it’s not the end goal.
The headline and/or article is confusing. It says “machine learning library for the Linux kernel” but goes on to say it’s going to do ML in userspace.
It sounds like it’s going to be a process that lives outside the kernel and tunes the kernel via some interface. In which case there would be no need for any ML-specific bit in the kernel. Rather a generic interface to collect stats, tune thing. Processes would use those however they want, using ML or astrology or classic algorithms for their internal logic.
They’re requesting mostly wrong solutions for real problems.
Age verification doesn’t address social media’s problems, but does increase data collection and decrease privacy. Same for decrypting private messages.
A guardian account does seem reasonable.
They could also completely turn off user seach for minors, so they would have to add contacts by username or email, and couldn’t reach or be reached easily by online strangers.
Minors could circumvent this if there’s no age verification. But today’s age verification methods are neither privacy-friendly nor hard to circumvent. Until they are, it’s not worth requiring age verification.