Re: So, Musk is digging the hole ever deeper
I don't think they would use a French word
20711 publicly visible posts • joined 31 Dec 2009
>mine has endured significantly more abuse than a modern X series could stand up to.
I think you might be misunderstanding the nature of capitalism. I have a storeroom full of Dell XPS that are out of official support and so corporate abandoned them.
There was also a bunch of Alienware that somebody managed to specify for 'graphics intensive CAD workloads' don't know what happened to them !
Not necessarily, most consumers aren't optimising for reduced install footprint.
We used to install the Windows 'N' builds - the anti-trust version with no bundled codes - to save a few Gb on VMs
But for a customer buying a preinstalled machine with a Tb of disk - it's optimal for them to have drivers for every bit of weird hardware and fonts for every language, without having to go online to download it.
But it is a union effectively making government policy. Suppose the US longshoreman decide they aren't unloading Volvos (officially cos of their Chinese owner and uighurs) or no Porsches (cos he was a Nazi)
No, of course the USA hasn't banned the import of European cars, it's just legitimate worker's rights.
>but there was this one guy who was some kind of kernel developer, and he was an absolute cock. He had to stick his nose into everything.
That's the nice thing about working in commercial corporate software development, everybody pulls together for the good of the shareholders. The efficiency of the market means that anyone not contributing surplus value is weeded out by ruthless capitalism
>We all remember the earlier Ariane "mishaps" so they are not our of the woods yet for sure.
Ariane 5 has a pretty good record. The first one blew, the first one of the improved version 10 years later fell into a swamp - but the others stayed up !
they had a couple of 2nd stage partial failures but on a scale of massive rockets they have an enviable record.
They were also much more precise than alternatives at the time, so the payloads didn't need as much manoeuvring fuel themselves - don't know how SpaceX's craft compare
I meant Germany specifically.
In defence spending Germany plans to spend pretty much in line with their huge throbbing GDP, but every time they actually choose a system the contract is immediately protested by every other company in the business. That's why they are such good international partners - the only way to get a contract out is to make challenging it a diplomatic incident.
The French are the opposite, they join every collaborative defence project to sabotage it and then release their national competitor - but I didn't say that !
Always works well - there's nothing like a blanket dictat from on high to improve efficiency
1, They will just hire the people as contractors / PPI deals so they end up spending 10X as much money to employ civil servants through a consultancy
2, They reclassify a bunch of roles as not civil service by moving them to other quangos and hiring replacements while keeping "civil service" headcount the same
3, They freeze a bunch of projects cos they can't hire some specialist, while keeping 1000s of existing staff on payroll doing nothing
4, All of the above
OpenAI were in a tricky situation. They weren't a company. They were explicitly setup because a bunch of distinctly non-fluffy tech billionaires ,led by Musk, decided that if AI was just a bunch of internal projects inside Google and Facebook it would be
bad for business limiting humanity's growth.
The idea is that they would all not be tech-billionaires if the internet had been developed inside IBM, Burroughs and Bell and kept for their own internal use and what AI needed wasthe equivalent of an IETF.
So they all chipped in a few quid to OpenAI while also running their own internal programs hoping to be the winner.
Non-competes have little chance in Ca and especially against a guy you just fired !
They might have a chance for non-solicitation if Altman explicitly offered MSFT jobs directly to the 700 engineers, but if they quit 1st and were careful not to wait for an individual job offer.
Plus, once these 700 go OpenAI is going to be a going concern for about a Planck time, it's very difficult to claim they are competing with a company that doesn't exist anymore
Actually rather the opposite.
Altman is the former boss of arch-capitalist, famous silicon-valley VC YCombinator.
He was saying that they need to find commercial uses and partners to raise the $Bajillion it takes to train these models.
The rest of the board of the (sort-of) non-profit were installed to keep AI pure and out of the hands of people like Microsoft
Then companies would simply pay the same amount as consultant fee to a 3rd party cyber security expert to ensure the data was never leaked.
In the same way it's illegal to pay bribes but you can pay consulting fees to the family of the president for their technical expertise
>You can trust me: I'm a blackmailer.
Ironically they have to be trustworthy or nobody would ever pay them.
The best way for a law enforcement agency to stop this sort of crime would be to steal some data from a very public organisation, have them pay up and then release the data anyway with the message "Suckers!" then nobody would ever pay again and so there would be market and no thefts
OpenAI isn't really costing Microsoft $$$ - they are paying in Azure time and can schedule when nodes aren't busy.
In return they are getting the only upgrade to O365 that people will pay for.
"Want to upgrade to Office365-pro? Only $49/month and you get 3 new Powerpoint bullet styles" vs "Office365-AI for $99/month, and it writes your Powerpoints for you"
Copilot is seriously impressive for doing a lot of the boilerplate setup and skeleton code for complicated stuff. Try writing Vulkan code just from the specs without it .
A perfect diffraction limited system (ignoring laser beam coherence and beam waist stuff) the angle it spreads out by only depends on the size of the transmitting telescope and the wavelength.
So call it 1um infrared wavelength and a 1m telescope (7llinguini = about the biggest you could reasonably put on a space probe and point accurately)
That gives you an angle of 1 in 1million, so a 1km wide spot at 1M km (or should that be 1Gm ?) or 16km at 10M mi
Make the telescope a compact 1 Linguni ( 14cm ) and you have a beam 6-7km wide per million km
Of course that's the best case, you can always make it wider by buying a cheaper mirror or being a bit more cack-handed during assembly. Interestingly the atmosphere has a much smaller affect looking down than looking up at stars. Since you only get the wibbly-wobbly atmospheric turbulence when the light is almost at its destination it doesn't have time to bend out of the way much.