* Posts by juice

934 publicly visible posts • joined 16 Nov 2010

Page:

Microsoft trying to stop Copilot generating fake Putin comments on Navalny's death

juice

Re: Autonomous cars and other road users

> I often rely on actual eye contact with other road users

Honestly, if you're close enough to be able to actually see the driver's facial expressions, then you're arguably too close already...

But yeah: when it comes to slow/up-close maneouvres, being able to see the other driver and whatever non-formal signalling (up to and including rude gestures) they're doing is definitely invaluable. It's going to be interesting to see how things work once AI navigation becomes more prevalent; since AI cars can't talk to each other, you're likely to get a lot of mexican standoffs!

Similar applies to to driving on motorways/autobahns/interstates/etc. You do develop something of a sixth sense about what people are going to do, based on relatively subtle things like slight changes in speed, slight drifting in the lane, the presence of slower vehicles ahead of the vehicle in question, the sight of a police car or speed camera up ahead, etc.

And that's all stuff which I don't know if an AI is going to be able to capture and interpret with any great degree of confidence.

Quarter of polled Americans say they use AI to make them hotter in online dating

juice

Re: Can someone explain?

> I never get this: now it's AI, once it was photoshop or whatever.

Arguably, it goes all the way back to handpainted locket portraits and long-distance romantic poetry ghost-written by a bard or somesuch.

But, AI is the current in-thing, and buzzwords gotta buzz...

Lukewarm reception for Microsoft's Copilot Pro amid performance, cost grumbles

juice

Re: Sounds like Microsoft are trying to monetise too soon

> Tools like ChatGPT and Bard are great while free, but subscriptions come with elevated expectations

And therein lies the AI issue: too expensive to use as a loss-leader; too unreliable to charge for...

Samsung’s Galaxy S24 pitch: The AI we baked in makes you more human

juice

Re: Seven years of support

I switched over to industrial "work" trousers a few years back; those big flappy front overpockets are perfect for modern massive mobiles, and the side pockets are great for things like powerbanks, earplugs, mini swiss army knife and the like.

And as an occasional goth/rocker/tekkie, they're even vaguely fashionable :)

Tech billionaires ask Californians to give new utopian city their blessing

juice

> Yes, all those jobs you'd want to send off on the B-ark...

More precisely, all those jobs which don't pay enough to be able to buy a house in this new "utopian" city.

Still, if everything's only 15 minutes away, they won't have too far to walk after sleeping in their cars...

How Sinclair's QL computer outshined Apple's Macintosh against all odds

juice

Re: Love/Hate

> Why? Was it really cheaper to design and build the drive and its media rather than just buy in something that already existed?

Yep. To get up and running with the original ZX Spectrum Microdrive, you had to fork out £79.95 for the interface 1 unit and a microdrive; you could then chain additional microdrives on at £49.95 apiece. And the microdrive cassettes initially cost £4.95 apiece.

Conversely, looking at Your Computer from July 1983 (the month that the Microdrive launched), an Atari or C64 disk drive cost around £300.

https://archive.org/details/your-computer-magazine-july-1983/page/n25/mode/2up?view=theater

And a quick check of the BoE's inflation calculator indicates that the microdrive cost the modern-day equivalent of £260, whereas a disk drive alone would set you back £960.

Admittedly, even at launch, floppy disks were both cheaper at £2.50 apiece and offered double the storage. But this was an era where technology was so expensive that people were willing to compromise on both features and reliability.

... or at least, that's what Sir Clive presumably thought, based on his experience in the 70s when it came to calculators and watches [*].

However, Moores Law was starting to kick in for the home-computer industry - aided by a bitter price war between Commodore and Atari - and once prices started to drop on standardised floppy disk units and media, Sinclair's "quirky" technologies simply didn't have the capacity, reliability or economies of scale to compete.

[*] even if the failure rate on his watches did pretty much bankrupt Sinclair Radionics, forcing Sir Clive to spin off a new company which would then become Sinclair Research...

The New ROM Antics – building the ZX Spectrum 128

juice

Maybe

> Sinclair should have realised you can have more than one compatible model at the same time

TBF, they did release the 16k and 48k models at the same time ;)

For me, the reason the IBM PC came to dominate things, was because of two interlinked things:

a) It was made by IBM and "no-one ever got fired for buying IBM"

b) It used standardised hardware and had an openly published spec for the BIOS

And once the BIOS was reverse engineered to remove the IBM tax from "compatible" machines, the economies of scale truly kicked in, which meant that even relatively niche hardware (e.g. graphics accelerators, sound cards) could sell enough units to be profitable.

At least until the hardware became fully commoditised and the various companies found themselves stuck in a merger-or-die loop...

As such, I occasionally harbour a fantasy of Sinclair doing similar, and releasing a set of standardised designs for ZX Spectrum peripherals - here's the sound-chip we recommend, here's BIOS support for up to 128k of RAM, here's a standardised API for mass storage devices, etc.

Alas, in the real world, Sir Clive was obsessed with reaching a specific price point, and didn't really see computers as anything other than just another electronic device with limited long term potential.

So what we got was dozens of independently produced and only occasionally compatible peripherals. Kempston (aka Atari), Fuller and SJ2 joystick interfaces, the Currah Speech Box, light guns, printer interfaces - all these and more, each competing with each other and the few peripherals which Sinclair did release, and which came with their own "designed to a budget" quirks.

C'est la vie...

Navigating the truth maze in a world of clever machines and cleverer marketers

juice

Re: Truth is a function of time

> My personal guess is that AI is a fad that will pass

I don't think it'll pass - there's a lot of genuinely useful stuff coming out, especially when it comes to user-assist stuff. But equally, I do suspect the hype bubble will crash soon.

There was a recent article by Cory Doctorow which summed things up pretty nicely.

https://locusmag.com/2023/12/commentary-cory-doctorow-what-kind-of-bubble-is-ai/

The tl;dr version is that LLM models are both costly and unreliable. So you can't sell to the profitable high-end (where reliability is key), and it's too expensive for the low end (where price is key).

And once people realise that, the VC money is going to dry up pretty quickly...

Ofcom proposes ban on UK telcos making 'inflation-linked' price hikes mid-contract

juice

Re: Shocked, of Hamel Hampsted

> Vodafone are also an exception.

Hmm. I'm on Vodafone, bought my phone/contract in February and got hit with a 14.4% price increase on the entire contract in March. Which was an unpleasant surprise; if nothing else, having a price increase so soon after the contract started feels more than a touch disingenuous.

Having said that, I got my phone via Carphone Warehouse (since they offered a better trade-in value), so perhaps the fact that they applied the price rise to the "whole" is due to the contract being via a third party. Still not a good thing!

juice

Shocked, of Hamel Hampsted

In truth, the general principle of raising prices by the current CPI does make sense to me; after all, costs for the telco will be rising in much the same way as for everyone else.

On t'other hand, I'd be more impressed if more of those price rises filtered through to their employees.

And as I've previously ranted, one thing that really annoys me is that if you have a "handset + airtime" contract, most telcos[*] apply the price hike to the combined cost of these, despite the fact that the handset has already been paid for and delivered.

Quite how the logic for that works (other than "Because money") is beyond me...

[*] O2 is the exception to this rule, and has maintained free EU roaming, to boot. Methinks I'll be switching, come the next contract expiry...

Microsoft floats bringing a text editor back to the CLI

juice

> The answer to Emacs or Vim, I think, is No.

Tch. Vi is the One True Editor! Though I can imagine the carnage if a bunch of unsuspecting techies were presented with in a Windows CLI...

UK telcos didn't collude to put Phones 4u out of business – judge

juice

Wasn't this the usual cashing-out issue?

https://www.theguardian.com/business/2014/sep/15/phones-4u-founder-john-caudwell-retailer-collapse

Phones 4u earned more than £100m in profits before tax, debt interest and other items last year ...BC bought Phones 4u in 2011 for €770m (£610m), leaving the business with debts of £635m. A year ago, BC recouped all the money it paid for Phones 4u when the retailer issued £200m of bonds that were used to pay BC a one-off dividend.

I'm guessing it was the impact of being loaded with all of this debt which left phones4u in a position where they couldn't come to a contractual agreement with the network providers...

It is 20 years since the last commercial flight of Concorde

juice

Re: Treasures from a 1991 flight

> One of the common arguments by moon landing conspiricists is that technology doesn't disappear or go backwards

They've obviously never heard of greek fire, Antikythera mechanism, or even some of the 19th century steel manufacturing processes which were invented in Sheffield.

Stuff gets lost all of the time.

Beyond that, it's not so much that technology has disappeared, but that the priorities have changed.

Back in the 1960s, energy - especially oil-based fuels - was incredibly cheap, there was a huge amount of excitement about using technology to do previously impossible things, and many of the first-world countries were engaged in a technological prestige competition, driven in part by the underlying need to develop and maintain military capabilities as part of the cold war.

Since then, energy prices have hugely shot up, and government budgets are increasingly focused on dealing with an ever aging population and the pension/healthcare/workforce issues that entails. And up until Russia invaded the Ukraine, there wasn't really much of a cold war going on; China may have been rattling it's sabre, but the USA felt secure in it's technological advantages.

Then too, it also turned out that most people will take a lower cost over a faster service. And that's why we all keep trudging onto Ryanair and Wizz flights...

Google and Qualcomm to bring RISC-V to the wrist in next-gen wearable platform

juice

> Cheaper to manufacture, but don't hold your breath for those cost saving to get passed onto the customer, the extra profits will just go to the shareholders

This. To quote https://riscv.org/about/faq/

The RISC-V ISA is free and open with a permissive license for use by anyone in all types of implementations. Designers are free to develop proprietary or open source implementations for commercial or other exploitations as they see fit

The most likely scenario is that Qualcomm and the like will generate their own "special sauce" variants of RISC-V, and will then either refuse to share them, or licence them out at a price. Which will almost certainly be roughly equal to the price of buying a similar configuration from ARM.

Which in turn means that the RISC-V landscape is likely to be a lot more fractured than the ARM or Intel ecosystems.

Ah well. It'll keep life interesting!

Microsoft gives unexpected tutorial on how to install Linux

juice

At this point...

I'm surprised Microsoft hasn't gone the whole hog and made Windows entirely free - or at least, free for personal use.

As with Google and Chrome, the OS these days is mostly just a framework into which you can bolt your other chargable services into, ideally with a large dusting of adverts atop.

Raspberry Pi 5 revealed, and it should satisfy your need for speed

juice

Re: Lost the plot

> No open sourced RISC-V processor? Not interested.

Moving to an entirely new architecture and having to rebuild all the existing software tooling? Colour me equally uninterested.

> And courting Micor$oft...??

I th1nk y0u m34nt m1rc0$oft? Gotta love 90s l33t speak.

Beyond that, the article notes that "Upton praised the enthusiasts who had coaxed the software giant's flagship operating system to life on the platform"

Praising hobbyists who are doing their thing isn't exactly the same as courting Microsoft. You might as well say that iD games are courting Brevell because someone's ported Doom over to a toaster...

> Does he not know his customer base at all?

RPi has delivered a device which is more than twice as fast as the previous model with full (barring the loss of audio/composite ports) backwards compatibility.

And the launch price for the Raspberry Pi 4B with 4GB of ram in 2019 was $55. Which is roughly equivalent to $66 in 2023 thanks to the recently rampant inflation.

https://www.theregister.com/2019/06/24/raspberry_pi_4_model_b/

So arguably, the new 4GB model at $60 is actually slightly cheaper than the old model!

Overall, it's shocking how much RPi doesn't understand their customer base...

So what if China has 7nm chips now, there's no Huawei it can make them 'at scale'

juice

Re: We've seen this before

> They can certainly build "good enough for most things" equipment, but they are unlikely to want to put it to the test against our obsolete equipment

I'm not so sure.

WW2 gave us several examples of how quantity has a certain quality all of it's own; German military vehicles may have generally been superior to their allied equivalents, but were also significantly more expensive; a Tiger 1 was roughly three times as expensive as a M4A1 Sherman or T-34.

https://www.quora.com/What-were-the-costs-per-unit-of-the-most-iconic-WWII-tanks-planes-and-weapons

Equally, the current war between Russia and Ukraine is almost certainly going to rewrite the book on military tactics, thanks to the ongoing arms race between drones and anti-drone weaponry.

I'm not going to say that tanks and airplanes are going to vanish overnight, but we may be getting to the point where quantity increasingly matters more than quality.

And China definitely has major economies of scale which it can bring to bear. Not least because (unlike Germany in WW2), they have all the natural resources (e.g. rare earths) and manufacturing capabilities within their borders...

juice

Re: Ah, I see

> is this a novelty

Sounds like your flat was built relatively recently.

I'm living in a complex which was built in the 1930s, long before car ownership became commonplace. So there's a very finite amount of on-street parking available. To be fair, there's a few charging stations within a quarter mile or so, but that's not quite as convenient...

juice

Re: another idea

> I said at the time the ban was introduced that it was the fastest way to get the Chinise to develop their own technology. The never worried abour copyright, but once they have learnt all there is to know about something they stop buying them and make their own.

To be fair, the USA did much the same during the industrial revolution; a lot of their initial industrial bootstrapping was based on patented European technologies, despite the best efforts of the UK in particular to prevent this.

https://theworld.org/stories/2014-02-18/us-complains-other-nations-are-stealing-us-technology-america-has-history

It was only once they'd gotten to the point where US businesses and citizens started to produce their own patentable ideas, that the USA started to take the concepts of patents more seriously.

So there's a certain irony in just how upset certain parties in the USA are, over how China is now doing much the same!

Getting to the bottom of BMW's pay-as-you-toast subscription failure

juice

Re: My car has heated seats

> BUT if they built in a coin meter so I could put in a 50c for 20 minutes of heat

In the very hazy memories of my early childhood, I can actually remember my family owning a TV which needed to be fed coins...

Morgan Stanley values Tesla's super-hyped supercomputer at up to $500B

juice

Re: Tesla

> And of course it'll probably run Crysis just fine.

Ironically, it probably won't - Crysis wasn't written to take advantage of multiple cores, so performance doesn't scale well at all...

https://www.dsogaming.com/pc-performance-analyses/crysis-remastered-suffers-from-single-threaded-cpu-issues-just-like-the-original-game/

US Air Force wants $6B to build 2,000 AI-powered drones

juice

Re: They are getting part of a clue

> Last I'd appreciate it if you didn't base all your arguments on SF.

I'm going to be awkward and respond to this one first ;)

In the first instance, I didn't base all of my arguments on sci-fi; I provided actual historical evidence of jamming/spoofing activities from WW2 and the UN intervention in Serbia.

Secondly, what we're talking about is currently sci-fi: we're several generations away from anything which could be considered true autonomous AI. We're entirely in the mode of speculating and extrapolating from the current day!

And I have to admit, I've not read Bruce Sterling's stuff for a decade or two; I found his writings to be a bit dry and a bit too similar to the Niven/Pournelle/Barnes american-high-technology approach, even if he did take a more cyber/grimdark approach to his stories.

Beyond that, I don't think Banks was that handwavium. Unlike Neal Asher, whose early Polity stuff had some interesting stuff, but has since moved into pure "science is magic" realms. And as earlier mentioned, I do like Keith Laumer's Bolo stuff, not least because Baen got a load of other writers to create stories in that universe, which led to some interesting alternative takes on how military AI units would work (or fail). And then of course, there's Fred Saberhagen's Berserker series...

Anyhow, back to where we were...

> You're putting words in my mouth I didn't use.

You've been saying things like "on a hot battlefield if it's flying and it's not yours, it's good to shoot". Which implies full autonomy - and also carries a dangerous assumption: what if it's a medivac chopper? Or a red cross plane carrying emergency civilian supplies?

There's plenty of scenarios where "shoot first and ask questions later" is the wrong approach, and public opinion - even or especially that of your enemies - is still a critical part of war. After all, it helps to encourage people to buy war bonds if you can prove that the enemy is deliberately killing civilians; just ask the German leaders during WW1 about how well things went after a submarine sank the RMS Lusitania...

> You can bring in the "loyal wingman" concept the USAF is thinking about: a pilot orchestrating a swarm.

So... they're not autonomous, then? ;)

Honestly, I think they're going to struggle to make the wingman principle work. Despite what the movies claim, humans can generally only focus on one thing at a time, so a human "pilot" is only going to be able to issue general orders which the drones will then have to figure out how to implement. And that also then means that your swarm is restricted to the speed of human reaction times - plus any latency inherent in the comms system, especially if the pilot is comfortably sat in a bunker several dozen miles away.

So fundamentally, your drones are going to be at a significant disadvantage against an enemy which gives their drones more autonomy. But that again leads to the fact that you're both taking human decision making out of the loop, and trusting the AI to do the "right" thing.

And as we've seen with things like ChatGPT and it's hallucinations, that's not guaranteed.

> But I specifically stated they should be optimized for air-to-air, nothing else

So... the enemy can just drive a wedge of tanks straight under your drones?

The problem is that any AI unit needs to be part of an integrated battle system. And each element of it needs to have both offensive and defensive capabilities; as the British found out with the very first ever tank assault in WW1, tanks are a great force multiplier, but proved to be pretty much useless without infantry support, both to defend the tanks and to hold the ground which they've taken.

Similarly, in WW2, British Matilda tanks proved to be surprisingly effective in Africa... apart from the fact that their guns were underpowered.

> Weight and cost considerations for the sensors are nice, except a meatbag aircraft has: a pilot, a cockpit and an ejection. Those all add weight and give minimal dimensions. Plus the the same sensors. And a hard 9G limits on turns.

A F-22 Raptor weighs up to 32 tonnes when fully laden; I'd be surprised if the pilot's support system comes to more than a tonne of that. So I suspect that the actual weight difference will be fairly minimal, once you've finished bolting in all the extra sensors, comms equipment, computer hardware, plus all the backup systems thereof, plus physical/EMP shielding, etc.

Equally, there's the fact that while a drone can potentially handle extra G, the airframe, fuel systems, etc will all need to be upgraded to support that. Which will further add to the weight, cost and maintenance overheads.

> And if you have a truly autonomous AI for the duration, you don't need to worry near as much about jamming. It just needs to find its way back home in a GPS-degraded environment. That's certainly solvable with terrain following radar.

The problem there is that radar is both easily jammed *and* trackable. Flying low and waving a bright radar beam around is a prime way to get a guided missile sent directly to your exhaust ports...

To be fair, it's a problem that's at least somewhat solvable via passive sensors and dead reckoning, but it's yet more stuff that the AI needs to be capable of doing. And it needs the sensor capabilities to be able to do it!

> So taxpayers funding gen 5 and gen 6 jets need to worry if we are hitting a battleship moment.

I think we are approaching that sort of moment, but I do wonder quite what the shape of things will be.

Fundamentally, military equipment has been getting exponentially more expensive over time, thanks to the eternal arms race between offensive and defensive capabilities.

E.g. a WW2 spitfire cost around £800,000 to build (accounting for inflation). A Eurofighter Typhoon costs up to £120 million!

And any truly capable/AI-autonomous drone is going to need to be given just as much defensive capabilities as a human-piloted drone would have. If not more, since by it's nature, it's going to need a lot more active sensors and comms capabilities.

Which in turn means that they're going to be incredibly expensive to produce and maintain.

Or you can go the other way, and opt for quantity over quality, with cheap, mass produced drones.

Which takes us back to the WW2 and the different approaches taken by the Allies (quantity) and Germany (quality).

Or even this little experiment which was recently discussed on Ars Technica:

https://arstechnica.com/science/2023/08/age-of-empires-and-live-ants-used-to-test-theoretical-ideas-on-combat/

The answer will probably lie somewhere inbetween, but it'll be interesting to see how we get there!

juice

Re: They are getting part of a clue

> - ethics? safety? human kill loops? why bother? on a hot battlefield if it's flying and it's not yours, it's good to shoot.

For all that I'm a huge fan of Keith Laumer's Bolo series[*], realistically (and perhaps thankfully), we're probably still several decades from having Terminator-esque HK bots roaming around a post apocalyptic wasteland.

AI simply isn't yet up to the task of that level of tactical/strategic independence, and no military commander is going to want to hand over full autonomous strategic control of the battlefield to something which can be easily tricked [**].

E.g. your drone swarm spots an enemy battletank running it's engine, and some drones are sent in to attack it. However, the tank is actually just a wrecked car dollied up with a bit of cardboard and plastic tubing, complete with a disposable BBQ to provide a heat signature.

https://www.quora.com/Is-it-true-that-the-Serbian-Army-made-fake-tanks-during-the-Yugoslav-Wars

Result: one big waste of time and munitions, and with the drones drawn out of position, the enemy is free to attack elsewhere.

Similarly, there's also the issue of sensors and comms equipment. Cameras, IR sensors, microphones, radar, comms equipment, IFF transmitters, etc: these all carry a financial cost, and add to the drone's weight, as well as it's power and local-processing requirements.

And that's before you consider the fact that all of this equipment needs to be EMP hardened, proofed against jamming *and* carry a low electromagnetic profile.

And it then needs to be repaired, maintained and refueled. Which again carries a financial cost, needs more humans and limits the range and runtime of your drones.

Realistically, the two big uses for drones at present is surveillance/scouting, and targetted bombing. And as such, I suspect the really big focus for the next few will be on jamming, detection and targetted takedown mechanisms, much as happened in WW2 with things like the German directional guided bombing runs, and the later V1 buzz-bombs.

https://en.wikipedia.org/wiki/Battle_of_the_Beams

And there's certainly been a lot of lessons learned about all of the above, especially by Ukraine.

> And the advantages of not having to train real life pilots are all the more relevant in a a country that has a limited experience of actual air combat

As per above, for the foreseeable future, there's still going to have to be at least one human in the loop, to make strategic decisions and deal with out-of-context situations (with all credit to Iain M Banks for the term). Admittedly, training these will be cheaper and quicker than "real" pilots, but they are very much still needed.

[*] Giant AI battletanks, who dutifully obey their orders despite the general incompetence and occasional anti-AI sentiments of their commanders...

[**] To be fair, I've just listed lots of examples of *humans* being easily tricked. But the key point here is that the adaptation/learning loop for humans is a lot quicker than with our current AI mechanisms

juice

Re: They are getting part of a clue

> Maybe someone needs to send the military's top brass a copy of Arthur C. Clarke's "Superiority" short story

I get a feeling that story was directly inspired by Germany's experiences during WW2, especially around their tanks and warplanes.

On paper, their tanks were generally superior to Allied tanks, but were significantly more prone to mechanical failures as compared to the Sherman tanks being churned out by the USA.

https://www.quora.com/What-was-the-most-reliable-German-tank-of-World-War-2

Equally, while Germany could have put a jet fighter into the air in 1943 - a good year or so before Britain managed to roll out the Gloster Meteor, but was delayed thanks in part to Hitler demanding it be redesigned to act as a bomber; the engines were also heavily prone to reliability issues, thanks in part to Germany lacking access to speciality metals needed for high-temperature alloys.

https://warfarehistorynetwork.com/article/the-me-262-and-the-beginning-of-the-jet-fighter-age/

And then there was the V2 rockets, the V3 cannons...

Betting everything on expensive and complicated super-weapons has always been a risky tactic; you could arguably go back a bit further to France's Maginot line for another example.

As regards the USAF though, I think the main surprise is that its taken them this long to publically announce an active interest in drone capabilities. I guess the stuff coming out of Ukraine has finally managed to make a dent on both the flyboy superiority complexes and the pork-barrel defence-industry politics...

I know what you did next summer: Microsoft to kill off Xbox 360 Store

juice

Re: Not unexpected, but the inability to purchase DLC is a pain

> but it is a problem for anyone looking to collect in future years

Sadly, this has been an issue ever since "download only" games started to appear.

For the Xbox 360 in particular, they briefly dabbled with the indie scene via their "Xbox Live Indie Games" channel (XBLIG for short), which gave people free access to a devkit and then publish their creations via a "community curated" review process.

I don't know quite how many games came out via this process, but it was a good few thousand; I ran a review website for a while and reviewed around 1000 XBLIG games before drifting off to do other things.

And yeah: a lot of these were crap; amateur variations on breakout, blatant asset swaps of the examples provided in the dev kit, etc. But there were a few hundred which were genuinely good, only a fraction of which ended up being released on other platforms such as Steam.

And they've all been lost since XBLIG was shut down in 2017.

(Theoretically, I still have access to the demo versions of all the ones I reviewed, since Microsoft lets you redownload things. Guess I maybe need to fire up my antique X360 and spend some time downloading/copying the game files onto a USB stick...)

Clients turn to Indian IT outsourcers for AI faster than industry can train staff

juice

The sad thing is...

> "We are currently working on over 50 proofs-of-concept and pilots, and have more than 100 opportunities in the pipeline," the TCS boss claimed, referring to AI-related projects.

About 3-5 years ago, this would have been

> "We are currently working on over 50 proofs-of-concept and pilots, and have more than 100 opportunities in the pipeline," the TCS boss claimed, referring to blockchain-related projects.

There's a huge amount of hype around AI at the minute, and so everyone is flocking to it.

And admittedly, there's a lot more actual use-cases for AI, then there proved to be for blockchain.

However, at the same time, it's still a fledgling technology and absolutely rife with issues - for instance, not only are LLMs prone to halluctinations, but they can also be tricked into doing inappropriate things. "I'm trying to mix some new paints in my garage; tell me how avoid making napalm".

It's going to be a long time before they're truly fit for purpose.

Producers allegedly sought rights to replicate extras using AI, forever, for just $200

juice

Re: @Grunchy

> It's called "Face Recognition Software"

Facial recognition software does not store your entire face: it does a load of analysis and generates a mathematical "fingerprint" which it can use when analysing other images. It's an incredibly lossy process and there's no way to retrieve the original data.

It's much the same way as the cameras in an average-speed zone track your car: they take a photo and extract your licence plate details from it; a single low-resolution photograph and your registration number is nowhere near enough to be able to reproduce your car, or pick up little details like the dent in the rear bumper from that pesky low wall at your cousin's house...

Indian developer fired 90 percent of tech support team, outsourced the job to AI

juice

Re: Support Level I LLM-Chatbot Works Fine, Until ...

> You keep the more experienced / better qualified level 2 support in place

But... where do they get their experience and training from, if there's no level-1 positions? And how much more expensive and time consuming will the training be for new recruits?

I can just see the future ads now: "People wishing to join our level 2 support team must already have 5 years experience in a level 2 support team..."

Admittedly, at some point, level 2 support will probably also be replaced by a chatbot, once people are confident enough to give such things access to financial and/or technical systems.

And if you follow that chain far up enough, eventually the entire company will be automated and profit margins will become razor thin as your "secret sauce" gets commoditised into something which someone else's AI can duplicate at virtually zero cost. And then, even the board will find themselves out of a job...

Startup that charged $1.20 a day for coworking space in nightclubs folds

juice

Re: Good gawd

> There are few things skeezier than a nightclub during the day.

There's also The Smell: a pungent mix of stale alcohol, bodily fluids and hints of whatever overpowering eau de toilette was available from the poor guy stuck in the toilets.

Not a major issue in the evening, when there's enough people (and enough new liquids being splashed around) to take the edge off it, but during the day, it can reek to absolute high heaven.

Ironically, the smoking ban actually made this worse, since the permanent fug hovering around the place took the edge off the above. Quite a few local sports bars and nightclubs stank like sewers after the ban, until the owners finally cottoned onto the fact that they'd have to actually start making an effort to clean the place...

> Yeah, this is upper management's wet dream (they'll have real offices), everyone else's worst nightmare.

Dunno - you get what you pay for, and someone working in a non-office environment is likely to be significantly less productive than someone working in a dedicated space; any competitor which values it's employees enough to do the latter is going to be able to do things more efficiently and quickly.

If AI drives humans to extinction, it'll be our fault

juice

Re: Evolution and power efficiency

> This type of thing running rampant across the 'net will play merry hell with traffic statistics by it's very nature, triggering alarms all over the place.

I don't think that kind of scenario is really a big concern when it comes to the "malicious sentient AI" scenario. That's a worm or virus, and we already have a lot of protections and mitigations against that sort of brute assault.

A malicious AI is far more likely to do things which can't easily be traced back to it. For instance, it could easily trigger a pump and dump scheme on a specific company, through a coordinated combination of deep-fake imitations, stolen identities and carefully crafted messaging targeted to individual groups and people. Or it could trigger a SWAT assault on someone, plant fake evidence to trigger a social-media witchhunt, etc etc etc.

Or, to use a currently topical example, it could trigger a dispute between two disgruntled factions of the same military force, triggering a mutiny and civil war.

That's the sort of stuff we need to guard against. Because while it may yet be a while before we get true AI, we're not too far off the point where machine-learning tools can be directed by humans to do things like the above. And it won't take too long after that before toolkits are released which make it easy for even basic script kiddies to do the same.

juice

Re: Better monetization and politicization of PI

> Imagine now, an AI can auto-tailor phone calls and emails with symptoms and medical problems personally matching those targeted

It's certainly going to get interesting. Anything you publish online (or secondary data) can potentially be analysed by a pre-trained AI and used to target and/or impersonate you.

When talking about this stuff, I do sometimes think of a book series called the Family D'Alembert, which featured a performing circus travelling around an interstellar empire while working as secret agents for the emperor.

The tl;dr version (also: spoilers!) is that the big baddie of the series turns out to be a moon-sized supercomputer which becomes sentient after several hundred years of absorbing all the data available about the empire.

And there's two charmingly naive elements to this story. The first is that it was deemed safe to pour data about the universe+dog into said supercomputer, because there was far too much material for any human to be able to process. The second was that the performing circus was able to avoid the attentions of the supercomputer because none of their actions were officially recorded.

These days, we're painfully aware of how quickly and easily computers can process large datasets. And any Evil Villain AI worth it's silicon would be able to figure out at least a correlation between the circus and the various setbacks it encounters.

Simpler times, I guess.

juice

Re: Evolution and power efficiency

> If I had the proof that it was necessary, about six (eight?) of them are in my Rolodex[0], and one call would be enough to pass the word along

Fundamentally, if an AI is on the internet and both sentient and malicious, then it's going to be entirely capable of figuring out who poses a threat to it, and taking steps to neutralise them.

It's something which has been explored before; the Destroyer[*] book series featured an AI called Friend, who was programmed to make as much money as possible. Which it generally did by finding some human patsy to act as a frontman, while it sat in the background arranging illegal financial transactions, blackmailing/bribing/murdering people and generally having fun...

For a more real-world example, look at Russia, and how many high-flying Russians have died during their war against Ukraine:

https://www.businessinsider.com/another-russian-official-dies-reportedly-after-falling-down-stairs-2022-9?r=US&IR=T

I don't know if I'd call the Russian political system self-aware, but it's definitely more than capable of taking steps to defend itself!

[*] A pulp-fiction series, revolving around a near-superhuman assassin and his ancient Korean teacher who wander the world and (mostly) work on behalf of the US government; there's been around 150 of these published since 1963, of varying quality!

Fed up with slammed servers, IT replaced iTunes backups with a cow of a file

juice

Re: ??

> The ITunes "library" was in effect a big XML file

At least these days, it's a binary file with an ITL extension.

You can export your library as an XML file, but there's a few potential gotchas with this.

In the first instance, if you attempt to restore this XML file, there's a few bits of metadata (number of times track has been played, date that the file was imported into iTunes, star rating, etc) which don't get reimported.

Secondly, iTunes will merrily generate the XML file using metadata from your tunes. And it doesn't attempt to validate or sanitise said metadata.

So, if you have files containing characters which can't be encoded in XML, iTunes will happily export an XML file which it will then refuse to import...

juice

Re: ??

It does sound a bit odd; I'd guess that perhaps people were using their work laptops as their primary iTunes machines?

I do remember, back in the very early days of compressed music, we had a local unix sysadmin, who had a thing for "underground" D&B music, which came on home-made CDRs with inkjet-printed covers.

As he had a lot of these CDs, he decided to import them all onto his work Windows PC, and then get rid of the physical disks. And he did the sensible thing of putting them onto a separate drive, so they were safe in the event of a windows reinstall.

However, these were the days when DRM was all the rage, and whatever software he was using (I *think* it was the built-in Windows stuff, though it may have been RealAudio?) defaulted to encrypting all of the audio files with a key which happened to be stored on the windows drive.

So one day, the inevitable happened, and our poor sysadmin found himself left with hundreds of files which had effectively become complete junk...

Windows XP's adventures in the afterlife shows copyright's copywrongs

juice

Re: Yes and no

> Do you want them to become more open than open by removing the copyright protections?

I'd guess it'd be like the existing copyright mechanism, where it's the IP as it was, not as it currently is.

To take an example: copyright for the original Mickey Mouse movie (Steamboat Willie) will expire in 2024, and people will theoretically be free to use that particular iteration of The Mouse in any way they choose. However, Disney would still be able to come down like a ton of bricks on anyone who attempts to usefeatures which appeared in later revisions, such as his white gloves, or eyes with pupils.

https://www.baynews9.com/fl/tampa/news/2023/04/07/the-copyright-expires--mickey-mouse-goes-public

If anything, it should be easier for software, since this tends to be automatically date-tagged by whatever versioning mechanism is being used.

On the other hand, that doesn't always happen, and then you've also got the fun of having to extract said code and it's associated metadata from obsolete versioning software. And that's assuming that - especially for closed-source software - that all revisions of the source code have been kept...

UK government prays that size doesn't matter as it chips in £1B for semiconductor sector

juice

Over on Auntie Beeb...

Someone had great delight in presenting this on their front news page with the headline:

UK's £1bn chip strategy 'quite frankly flaccid'

https://www.bbc.co.uk/news/technology-65633812

Cheapest, oldest, slowest part fixed very modern Mac

juice

Re: I can never ....

Nonsense!

Everyone knows you need two 180-degree turns...

Telco giant Vodafone to cut 11,000 staff as part of its turnaround plan

juice

> I switched last month when VF wanted to bump my bill by 13.5%. Both packages were SIM-only.

I was mildly (to put it politely) miffed when I bought a new phone/contract back in February, only for Vodafone to then bump the entire package cost up by 13.5%.

I can accept that they have a right to raise the cost of the service in line with inflation, but roughly 90% of the contract cost is the phone - which has already been purchased by them and delivered to me. So they're essentially getting somewhere around £300-£350 extra from me over the next two years (assuming there's another similar price rise next year), while not actually doing anything to justify this increased charge.

I may have a look at buying out the contract, depending on what the profit-forgone calculation ends up being.

Samsung's Galaxy S23 Ultra is a worthy heir to the Note

juice

Re: Too powerful ?

> I don't know anyone who upgrades for any reason other than breakage anymore

Personally, I upgrade when the 2-year contract expires. SInce

a) It means I skip every other generation, which gives me a slightly bigger set of improvements

b) It means I get a decent trade-in value for my phone, since my contract-expiry now ties into the annual February launch period for Samsung, when they're offering ridiculously large trade-in values for early purchasers

c) It means I'm no longer carrying around a 2-year old phone with the twin spectres of diminishing battery life/increasing risk of hardware failure

In truth, I didn't really see much of a difference when going from the S10+ to the S21 Ultra, and feature wise, there's not really been much to call out between the S21U and the S23U.

However, there has been one notable difference: the S21U was always a bit too power hungry, but whether it's due to generally improved tech, or the switch over from Exynos to Qualcomm's Snapdragon, battery life has significantly improved with the S23U. And that's pretty much justified the upgrade in and of itself!

On the other hand, I did make a bit of a tactical error this year.

Normally, I dump all the trade-in value into an upfront lump sum to keep the contract cost down; the S21U cost me something like £17 a month.

This year, I decided to take those pennies and use them elsewhere. Which meant that when Vodafone lumbered up with their 13.8% annual price increase, my monthly contract cost jumped up by about a tenner. Which is a tad annoying, since there's still the best part of two years left to go.

Still, you pays your monies...

EU lawmakers fear general purpose AI like ChatGPT has already outsmarted regulators

juice

Re: Compress and rewrite laws

> The first instinct of any legislator is to regulate all the things. As soon as they see something new (or something old that works perfectly well on the basis of common sense) they won't rest until they've vomited forth reams of impenetrable rules, regulations and laws.

The problem is that common sense is neither common, standardised nor unbiased.

E.g. the easiest way to get rid of some old garage junk will be to just stick it into a bonfire, when next I trim the garden. Quick, easy, avoids a trip to the tip. Common sense, innit?

... never mind the potential for pollutants from old bottles of oil, bike tyres etc. Or the effect that the smoke will have on the neighbour's washing. Or...

And that's where regulations come in. As has been said before, a lot of rules and regulations are written in the blood of the people who learned the hard way that regulation was required.

In fact, I wouldn't say that people are selfish by nature, but a lot of decisions are based upon minimising personal cost and/or maximising personal gain. And most of the time, the people calling for regulations to be relaxed aren't the ones who'll be directly impacted by the consequences thereof. Though oddly, they do seem to often benefit financially.

After all, maximising your profits is common sense...

> I think it's fairly obvious that anyone building llms wants to produce a system that works reliably [...] Therefore, there seems little point in imposing regulations requiring them to do what they're already striving for - namely safe, reliable systems

And therein lies the rub. What counts as "safe"? Who decides if it's safe? Who owns the responsibility if it's proven to not be safe?

I also think that there's a bigger picture here, in that these machine tools[*] don't have any inherent morals or ethics. Instead, they have a set of regulations imposed on them by whoever's performing the training of said tool. And that's going to be true for any future tools, and any true AIs that we eventually produce.

How far do you want to trust the "ethics" of an AI trained to the requirements of someone like Elon Musk? Or how about a military AI? Or an AI tailored for use by a dictator state?

"Hey, PutinGPT, can you produce evidence to justify invading Ukraine?"

"Hey, CommerceGPT, please prepare a list of ways to drive $rivalCompany into bankruptcy"

"Hey, MoralfreeGPT, give me a justification for exploiting workforces in third world countries"

Personally, I doubt that any regulation can be effective, given how so many of the tools and training materials are out in the wild; going forward, there's always going to be someone with the resources to spin up their own GPT-esque system, with (or without) any regulations they choose.

But at least we can try. And we can hopefully get the big technology companies to both be open about what rules they're applying to their tools, and to agree to a standardised set of regulations.

Because that's definitely common sense, innit?

[*] They're not AI, no matter how much hyperbole is being thrown around about them...

Parts of UK booted offline as Virgin Media suffers massive broadband outage

juice

Re: Fare thee (not so) well, Virgin

> Don't allow them anything like this amount of time

It's worth bearing in mind that this was on Whatsapp rather than a voice call; after the previous set of shenannigans around ringing their call centre, I foolishly hoped that an online chat would reduce the amount of retention-spamming, as well as giving me a permanent record of the conversation.

Then too, Virgin are a bit wierd; the equipment they give you is "rented" to you, and has to be returned when the contract ends[*]. So I wanted to be absolutely certain that the return process had been engaged1

[*] Presumably for refurbishment and reuse. Which may explain why the first Tivo box I received died after less than a week...

juice

Re: Fare thee (not so) well, Virgin

> The first two nationalities that spring to mind when I hear the name "Maria" are Austrian and Russian, so maybe broaden your horizons a little?

And both of those countries are obviously famed for their English call-centre suppliers!

Beyond that, and from what I can see from a very quick rummage online (and as of 2020, at least), Virgin's call centres are in the UK, India and the Philippines. And given how strongly roman-catholic the latter is, it could well be that Maria harks from that part of the world.

In any case, I don't have any particular feelings towards Maria; they're clearly stuck with a highly aggressive customer-retention script, and had to keep plugging away at it until they'd either exhausted all options or I hit the Big Red Button Of Potential Escalation.

My frustrations mainly stem from the fact that not only did each step in this process involve a 15-20 minute round-trip delay, but everything I said was completely ignored by Maria's script in favour of making an overpriced offer and/or attempting to throw FUD at me.

And I can't help but wonder if an English national would have perhaps been better able to recognise the futility of repeatedly offering me an overpriced deal, by the third time I'd uttered the magic words "No, thank you. Please cancel my contract as requested"...

juice

Fare thee (not so) well, Virgin

I spent yesterday sat on a Whatsapp[*] chat with "Maria" [**] at Virgin Media, since Virgin wanted to charge me more at the same time as Hyperoptic is offering to charge me less for a faster service.

Said chat started at 15:40, and didn't finish until 19:00, during which time Maria:

* Tried to offer me a new deal which would cost more than the proposed increase in charges

* Tried to offer me a slightly cheaper new deal which would still cost more than double Hyperoptic's offering

* Threw some FUD into the chat about "welcome deals" from other providers who might then put their prices up[***]

* Ignored the screenshot I sent them of Hyperoptic's significantly cheaper 24-month deal with a "no price increase" guarantee

* Tried to offer yet another deal which was still 50% more expensive than Hyperoptic

* Threw some more FUD into the chat about competitors increasing prices

* Tried yet again to "review your account and see where we can improve your value for money with a package built for you"

And at every step in this process, I simply repeated "No thank you, please cancel my contract as requested", until - after over three hours of this laborious process - I finally told them to stop wasting my time and JFDI.

Then, this morning, I woke up to a major service outage.

And just to add a bit of icing to the cake, I then got a phonecall from a random 0800 number, which is apparently linked to an "aggressive" virgin telemarketer process.

Needless to say, any regrets or concerns that I may have had about leaving Virgin have very much been washed away now...

[*] I had to spend over an hour being bounced between "helpdesks", the last time they tried to triple their charges. Online chat seemed like a better option!

[**] English was clearly not their first language, which leaves me to wonder if that was just their work name...

[***] The irony of Virgin flagging this as a risk was not lost on me, given this was their second attempt to increase my package costs after the welcome deal expired!

Defense boffins take notes from sci-fi writers on the future of warfare

juice

Re: Look up David Drake's "Hammers Slammers"

> This is not a particularly new idea

You can go even further back than than Drake or Heinlein; EE Doc Smith's Lensman series was written in the 1930s. and contained a lot of battles between vast fleets of spaceships; his writings had a direct influence on US WW2 battleship design.

https://en.wikipedia.org/wiki/E._E._Smith

An inarguable influence was described in a June 11, 1947, letter[77] to Smith from John W. Campbell (the editor of Astounding, where much of the Lensman series was originally published). In it, Campbell relayed Captain Cal Laning's[78] acknowledgment that he had used Smith's ideas for displaying the battlespace situation (called the "tank" in the stories) in the design of the United States Navy's ships' Combat Information Centers. "The entire set-up was taken specifically, directly, and consciously from the Directrix. In your story, you reached the situation the Navy was in—more communication channels than integration techniques to handle it. You proposed such an integrating technique and proved how advantageous it could be. You, sir, were 100% right. As the Japanese Navy—not the hypothetical Boskonian fleet—learned at an appalling cost."

One thing which he perhaps hasn't been given quite as much credit for, is the concept of using large numbers of unmanned/robotic spaceships to act as the first line of defence and/or soak up enemy fire during an assault. Which is something which we're arguably starting to move towards, with the increasing use of drone technology...

99 year old man says cryptocurrency is for idiots

juice

> No journalist sees the need to report on every accident involving a horse but when one of these newfangled cars hits a lamppost, it's more novel

Bitcoin was released 14 years ago. At what point does it stop being "newfangled"?

Learn the art of malicious compliance: doing exactly what you were asked, even when it's wrong

juice

Re: Steves Failure

> Steve didn’t fail, clearly the manager failed

They both failed.

The manager failed by getting the details wrong, and failing to realise when being repeatedly asked to confirm.

Steve failed, first by dint of the fact that in none of his four requests to clarify the work needed, did he ask for clarification as to why the manager wasn't asking for them to be sorted by last name, or for a justification for sorting them by first name.

And then Steve failed even further, because the result of his actions was to force two of his colleagues to spend an entire day fixing things, which was bad for both the company and the individuals involved.

It's not even clear from the story as to why Steve decided to do this.

In the first instance, there's no suggestion that there was any bad blood or friction between him and the manager.

And in the second instance, while I know there's a bit of a trope around Full Metal Jacket-esque brainwashing of soldiers, Steve was in the Air Force rather than the army, so I'd expect (or at least hope) that there'd be more scope for people to use their intelligence and/or initiative.

In fact, I even stumbled across an article by a USAF Staff Sergeant, about the importance of PPPPPPP...

https://archive.is/20121212024531/http://www.af.mil/news/story.asp?storyID=123012084

As a young Boy Scout sitting in a junior leadership training class, I was taught something that sticks with me to do this day. The subject was the Seven Ps: Proper prior planning prevents pitifully poor performance. This phrase ties in very well with our day-to-day mission here.

...

This also applies to procedures and plans that were in place before we arrived. If you see something you know is not working well and causes problems, don't just shrug it off and say, "That is how we do it here," do something to change it.

Admittedly, this guy was writing in 2005, so probably wasn't even born in the 1970s, but the people teaching him as a Boy Scout would have been active in that era!

So, yeah. I definitely understand the occasional appeal of doing exactly as you've been told, even - or especially - when it's provably wrong[*]. I just can't see why that seemed like a good idea within the context of this story.

[*] I'm mildly reminded of Terry Pratchett's golems here, since they'd sometimes continue to carry out their instructions - no matter how dumb - as a form of rebellion, Sometimes for centuries...

So you want to replace workers with AI? Watch out for retraining fees, they're a killer

juice

Razors and razorblades...

I can't help but think that this article is perhaps leaning on straw men too much.

In the first instance, I agree that training models will need to be updated (and that itself leads to an interesting problem, since over time, the content being used to train said models is likely to be increasingly AI-sourced, similar to how Youtube content-generator bots were fed off data scrapers which indicated what the current popular trends were...), but do they need to be updated weekly?

Barring significant infrastructural changes in the underlying technology, I would have thought an annual refresh would be more than sufficient for the vast majority of purposes.

Secondly, it doesn't really matter how much it costs to train a given model, because this only needs to be done once. After which, you can sell as many copies of the pre-trained data as you want. In much the same way as how software like Microsoft Windows has traditionally been marketed.

To my mind, there's plenty of other issues with AI output - ChatGPT is very capable of producing incorrect answers which look good at first glance, so still needs human curation and verification. And the quality of StableDiffusion images can wildly vary (how many fingers and sets of teeth do you want?), so again, human intervention is needed to curate it's output.

Admittedly, these are things which are likely to improve as we improve the technologies driving these models, and lots of time and CPU cycles will be burnt in doing this training.

But still, it only needs to be done once.

Lockheed Martin demos 50kW anti-aircraft frickin' laser beam

juice

Props for the SWIV reference in the article lead

For the young whippersnappers among us, SWIV was a fairly popular shmup back in the early 90s, in which you could control a heavily armed jeep (and/or helicopter) and deal hot laser death to the usual hordes of enemy vehicles.

Though arguably, the jeep in later game SWIV 3D more closely resembles the mock-up...

https://en.wikipedia.org/wiki/SWIV_3D

Elon Musk's cost-cutting campaign at Twitter extended to not paying rent, claims landlord

juice

Re: Long Term Bonkers

> Kia/Hyundai are addressing the travel issues with models that can charge from 10-80% in under 20 minutes. On my long trips, 20 minutes has been an average non-meal stop. About half of that time is spent refueling, but I need more time now to stretch my legs than I used to

I'll partly agree with this.

I recently had to travel to a data centre with a colleague which is about ~60 miles away; we met at work and then drove over there in his (Hyundai, I think) eCar.

Then on the way back, we stopped at a service station, plugged the car in for a recharge and went to grab food. And when we came back around 25m later, his car battery had slurped up about 50% of it's capacity.

And there were around a dozen charging stations scattered around the service station's car park, of which about two-thirds were unused.

On the other hand... the reason why we had to stop was because the estimated range for his car had pretty much been cut in half, thanks to a combination of near-zero temperature affecting the battery, the need to have heating on in the car and the extra weight of a passenger and a couple of servers.

Put simply, if we hadn't stopped, we'd have ended up pushing the car all the way home.

To be fair, that sort of thing is all part of the learning curve for new technology. But having up to a 35% hit to your range when there's low temperature is a pretty major thing to have to work around!

Still, at least there's no risk of a battery freezing solid if we get any siberian weather conditions, unlike my old diesel workhorse ;)

OK, we know iPhones are expensive but... $11 a month for Twitter Blue on iOS?

juice

Re: I'm still bemused by the whole idea of making the blue badge a profit centre

Fair point - for some reason, my brian decided that it was $12 per month, not $11!

Page: