I can't believe it took this long for them to realize people didn't want a camera that they could not fully control within their own private homes possibly recording and displaying everything to random people at the company that built and maintains the device and potentially for anyone who found a security flaw in it.... I mean, really, what a smart item to stick into a TV that might go in say a bedroom, because nothing happens in there that shouldn't be broadcast across the world and saved forever....

What too many people do not seem to understand with LLMs is that everything it spits out is simply a probability matrix based on the input you gave it. It will first attempt to deconstruct the input you provided and use statistical analysis against it's trained knowledge base to then spit out letters, words, phrases and punctuation that statistically resembles the outputs it was trained to produce in it's training materials.

Until this version, ChatGPT obviously suffered from a lack of training materials within it's trained neural network to have it overcome the English language's typed grammar rules for it to be able to discern that em dashes are not typically used in everyday conversations and/or that the input to not use them needed to change it's underlying probability network to be able to ignore the English language's grammar rules and adopt it's output without the use of the em dash. This is a very difficult concept to train into a neural network as it needs to have been training on specifically this input/output case long enough to have that training override the base English grammar language model, which is a fundamental piece of knowledge a LLM requires to function and one of the very first things it is trained to handle.

It also exposes a flaw in how neural networks are typically working. There is a training/learning mode and then there is the functional mode of just using the trained network. In the functional mode, the neural network links, nodes, and function are effectively static. Without having built in-puts to the network so that it can flag certain functionality, it can not change it's underlying probability matrix to effectively forget something it was trained to do. Once that training has changed any of the underlying neural network, you can not effectively untrain it (without simply reverting to a previous backup copy of the network before it was trained). This is why it is so important to scrutinize every piece of data that is used to train the network. One you have added some piece of garbage input training, you are stuck with the changes it made to the probabilities of the output. Any model that is effectively training against the content of the internet itself is so full of bad information that the results can never really be trusted for anything other than probability of asking a random person for the answer because it will have trained on and included phases like "The earth is flat", "birds are not real", and "the moon landing was a hoax". It will have seen those things enough times that it will include them as higher and higher percentages of the proper response to questions about them....

of a slightly shittier country.

I have never heard of Vine, but it sounds like Tiktok in much much worse. And since it comes from Dorsey, it's safe to assume it's gonna be shit.

Dupe or not, worth leaving this here as a reminder that no, China EV sales are not imploding. Debunked rubbish.

You'd be surprised how difficult that actually is. Take the role of a DJ for instance - a good DJ will know what genre they're doing, play well know things from that genre but also introduce new music to keep it fresh. They might also step outside the genre a little - not too far so it's not dissonant with the rest of their set, but just far enough to give a break and a moment of "ah, that's nice/romantic/gnarly/metal/" for the listener.

It's a skill, and if you haven't got that ability to start with then you're unlikely to be able to give the correct prompts to create it. You might well get a lot of identical things, but a listenable varied set is more than that.

A worry might be SteamOS as a requirement, rather than as simple support. You could imagine kernel modules being developed for 'anti-cheat' and them running under SteamOS but not some other distro that may (justifiably) block them.

This is the most (in fact only) interesting thing about the announcements to me. Must say I'm not sure about it - can't see how mouse+keyboard style games, which the original Stream Controller was explicitly designed to work well with, would pan out.

I have hugely customised layouts for several games to the point where I can't imagine playing them without it - they tend to be RPG games like Skyrim and Elder Scrolls Online. It's that style of game I'm trying to imagine mapping to the new layout, and to be honest gen1 looks more amenable to me at first glance. Hope to be proven wrong though.

Also, it is bringing to the forefront a long running battle that is waging in academia. Students have all been taught from a very young age that if they work hard, get themselves prepared, strive for and achieve excellent grades that they will get into the best colleges (i.e. the concept of meritocracy, where those that put in all the work and have the gifts for excelling in academic study and life would be rewarded for their efforts by being able to get into the highest rated colleges).

Yes, a lot of the students who are able to achieve and excel academically may also be from a more "privileged" background once you start mixing in things like social-economic backgrounds, family backgrounds, and even location. These are things that are hard to dismiss, especially in a public education system which has guidelines and mandates to educate the whole population, not just those who have excelled. This is the biggest clash that then occurs because you have a limited resource (i.e. the number of spots for students), which must somehow be divided out in some reasonable way. Things like standardized tests tended to do a good job at predicting who was academically prepared, and was one of the reasons those tests were created. Comparing the grading curves of different schools, teachers, and even students against one another is almost impossible without lots of additional information that the college selection committees simply do not have and don't have the time to research. The standardized tests provided some of that information.

Not really. The UC system has way more applicants than it can accept, and it has been that way for decades. As such, they already know they have the "income stream" that is "nice and healthy".

What this really means is that the UC system is doing a much worse job then they previously did in "selecting" the students into their system that are ready to meet the requirements without needing remedial math and writing.

In other words, the UC system changed how they were selecting people for acceptance, and the metrics of tracking the need for these remedial courses by a much larger percentage of the incoming students are showing that their current selection criteria is doing a worse job of picking out students who are academically ready for the standards of the UC system at the time of their selection/acceptance.

People want helpful AI that doesn't hallucinate, code crap code, tell you to eat glue or commit suicide.

Sorry but from an outside perspective that just sounds nuts. So let's take your 'worst case' - $129M overall cost making it $434 per entry - you're saying there are only 297,235 (129M / 434) tax payers in the US? A quick search from me shows the number of filings to be at 145M+. If everyone could file for free, that $129M would be 88c per person.

And I'm speaking from experience. I'm in the UK. I've recently filed my annual self-assessment tax. I used the free service on the UK government web site and the thing that took the longest was working out how much to put as charity donation. Whole thing done in less than an hour and a half.

I seriously cannot comprehend the approach where you have to pay to be able to pay someone. It's...well...it's nuts.

Cloud is useful for a few very specific things. Specifically, one off requirements (i.e. I need to temporarily use a few thousand CPUs/GPUs to run this study/calculation and never need to do it again once it is done), and initial ramp up/expansion (i.e. we need to start working now and we can't wait for our datacenter to finish building before we start work, but have a specific timeline for when we use the "cloud" and bring the work back home as the datacenter(s) come online).

We don't want to be associated with a fascist regime when the wind changes.

I trust Windows Server 2003 more than I trust Windows 11. It's less stable, but Microsoft isn't in control of your machine.