You can actually plot improvement as measured by benchmark vs power consumption to train, and you see the power consumption curve upwards while the benchmarks seem to be asympote trending towards a specific limit. I've got a theory on it too. An LLMs basic fitness criteria is "Output text as close as possible to the training data". Well one thing that is NOT in the training data is text generated by anything smarter than a human. Because we've never seen an entity thats smarter than us. So the whole thing seems to be converging on a limit thats essentially 'one extremely well read human", and it has no data to tell it what something smarter than a human would look like.

All training further does is push that curve closer to the limit line, but it can never go above it. Theres a reason a lot of AI researchers are pretty adamant that the transformer LLM is probably not the final form of AI that can do the whole AGI superintelligence thing.

Some of these things I suspect are old Steve Jobs dogmas. Jobs also hated two button mice. Thankfully you could always turn on the context right click, but even to this day the right-click seems to be something you have to turn on in settings (Not that I've set up a fresh mac in aeons. Jobs hated the ergonomics of touchscreens on laptops.

While I get the reverence for Jobs inside apple. Maybe its time they moved on from him. Well except for the customer service thing. Customer service from apple was *excellent* under Jobs. When the iphone sdk first came out, I wrote an angry email to sjobs @ apple dot com about the waiting time for approvals and how a client was threatening to sue me for delaying the project launch. That night at like 2am I got a phone call from his personal assistant telling me that jobs had flipped his lid and was on a rampage wanting to know why my app had been sitting in a queue for 2 months. Gotta respect that commitment to customer service. Crazy bastard actually read his emails. Even I dont do that lol

Nah, its all MacOS. Apple have been pushing the UIX for MacOS and IPad roughly in the same direction. I cant say I'm a fan of that (I'd rather the IPad be more Maclike than the other way around) but people seem to like it, so what do I know?

Oh, MacOS has been able to run IOS apps for a while now. Essentially IOS and MacOS have always shared the same XNU kernel (basically mashes together mach and FreeBSD kernel components with a custom API. And it was the userland libraries that differed. After porting the IOS userland over to MacOS that's allowed IOS apps to run unmodified on ARM based macs.

Personally I'd like to see Apple take away from the success of the Neo not that the mac should be made more IOSlike, but rather that the technology is ready for the next step where a user can carry around a phone, then place it into a dock somewhere and the phone becomes a mac. The only attempts I've seen at a similar idea where some of the Ubuntu-phone concepts, but I think the technology is now at a level where Apple, or Microsoft if they got off their arse and put the engineering hard work into it, could actually achieved it

Yeah earlier this year I brough my girlfriend a pretty base model Macbook M1 Air w/ 8gb, and about a week later my cat managed to damange the keyboard on my usual machine leaping onto it from a height, so I ended up using the M1 air for a couple of weeks as my work laptop while I waited for repairs and... it worked flawlessly? Keeping Jetbrains IntelliJ, microsoft word, and various terminals for logging into servers open, it ran it smoothly and even felt quite snappy. For sure the 8gb posed a few problems with large workloads, but for its intended use, my GF being able to read the net and use office suite for work. its great.

I'll probably buy her daughter one of the Neos for university, since she's been bugging me for a computer.

We are talking about an administration that has been busy paying companies NOT to do wind farms, and just tried to appropriate billions of dollars to give as gifts to the terrorists behind the J6 attacks.

We are not operating from a rational perspective here, its just pure malice.

For a while the Russians where making some really great blockbuster type action films, that looked like they where shot on 200mil budgets, but on investigation where 2-5 mil. The problem is towards the end of that period before the Ukraine war broke out, those films started turning more and more into military propaganda and it just got too obnoxious to keep seeking them out.

Maybe they'll come to their senses one day. They are quite good at doing what hollywood do, but on a budget.

If a purchase of a digital product looks like a sale, it is a sale,

Its how most of the western world sees it, and with natural-lifetime warranties for fitness of purpose, Its highly likely that they legally are supposed to refund purchases of these older products if they render them unfit for purpose.

The US are special snowflakes however and seem to have a different way of doing things however. But don't be suriprised of the europeans or australians get diabolical on microsofts ass if they try and impose this in the EU or Australia.

Being that its a mac, Apples Pages and Numbers apps are surprisingly functional software and I havent found many Doc or XLS files it cant open and work with.

Obv not going to be great for folks who use the more powerful features of office. But I *believe* they are free.

And yeah Libreoffice is reasonably functional too, even if it does feel a little arcane at times.

The problem with better DNA error checking is you risk choking off evolutions necessary source of entropy, and if that life can't evolve, it wont do well protecting against OTHER threats.

The conclusion would be that while too much radiation precludes evolution, (as it requires too heavy "error checking"), too little radiation also precludes evolution (no errors means no changes) and therefore there has to be a 'just the right amount', making that yet another constraint , in this case magnetic, on what sort of environment qualifies as the "goldilocks zone" for life.

Most of those complaints are from people too lazy or too stubborn to just type in the word into google and see what comes out. Worse, there is also a variant of people who say "I dont know what is", in a transparently bizzare attempt at trying to sound knowleagable, by feigning ignorance.

Don't this just make them chase never-ending profit to the detriment of all?

Anthropic where always a for-profit company, meaning that chasing a profit is a fiduciary legal requirement.

Tbh, of all the AI bro companies, Anthropic are the least obnoxious, and I do wonder how their stance against weaponization survives when board is stacked with venture capitalists instead of TESCRAL doomers.

Mindyou Nvidia may well be skewing young with its headcount. Prior to the AI boom NVIDIA had a very generous vested share program for its engineers, and suffered a rather unique problem when the AI boom shot their shares through the stratosphere when suddenly all their senior engineers where sitting on, in some cases, upwards of 20 million USD worth of shares each. And like normal people instead of wall street suits, they pretty much collectively said "Well, fuck this working shit" and cashed their chips and retired with their millions, gutting their ranks of senior engineers.

We probably would do well to shake the conception that Intelligence agencies are all-seeing/all-knowing fountains of competence. In reality they are filled with paranoid people of various levels of competence with a whole range of dispositions, including occasionally criminal.Intelligence agencies need to be a little criminal at times to get the job done. The idea that one of them might have been doing shady shit, in an agency that specializes in shady shit shouldn't surprise anyone. Hell, it was probably why they hired him.

Honestly, its almost a good thing.

A friend recently asked me to look into their kids laptop who had gotten around some age restriction stuff, and I was mystified. I asked the kid how they did it, and they laid out all the registry keys and the line of reasoning they followed to find them, and all I could think was this 10yo kid was as good a security professional as I am (he's not, but well, it was absolutely impressive work. ). I see this as a positive. The world of computers I grew in involved 10yos teaching themselves assembly to make C64 and Amstrad (I'm australian, so we got the same computers the UK folks did. BBCs, Sinclairs, and eventually C64s and Amstrads. Tandys where esoteric american imports) games. We started with basic, supplemented it with assembly we learned from library books and magazines (no internet in the 1980s!) and taught each other at school, then when we got to highschool and the fancy new IBMs we then learned Turbo pascal and got hooked on the possibilities. 40 years later I'm still at it.

But kids now, its internet slop, fortnight, and computers that dont require you to learn to LOAD "*",8,1 and tempt you with a manual promising infinite games if you learn to code.

But kids are resourceful, and if hacking their way around idiot-boomer designed age locks force them to actually learn how their computers work, then great.

I told my friend to download for their kid a copy of the Godot game engine, and look up the GDQuest lessons, because clearly this kids got a future, if he can be inspired to chase it.

Also, fuck AI, let kids learn.

I saw a pretty funny video on youtube (something like "creating the worst AI music") and they are just pumping more and more absurd prompts into it, and getting some pretty funny (and occasionally disturbing) results