Issues with the spacecraft bus and flight software?
still.........
313 publicly visible posts • joined 20 May 2020
just a few weeks ago.
The amount of power datacenters are sucking up just begs for local installations of SMR tech.
Any 'extra' power can be sold off to the proles. Indeed, the installation of self-contained datacenters with SMR would be another 'selling point' in areas that are struggling to get similar tech through the political bureaucracy.
It takes someone with the security and authority of working in the C suite to challenge anything regarding The Cloud (hallowed be it's name).
At any other level, individuals re potentially committing Career Suicide if/when they point problems with spend, pace, and not-so-recent 'discovery' of not just repatriation costs, but storage 'per-click' access costs for data a company once believed it owned.
.....and back them up to spinning metal every two weeks or so. The flash array stays at home office while the spinning metal gets exchanged to a closet shelf at the bro in laws house.
Primary backup software are the community editions of Veeam running a set of three backup media plus the primary/on-line storage. Checksums all around.
I 'trust' the longevity of spinning metal. You can pretty much put it on the shelf and walk away from it. The long-term shelf-stability of flash drives is yet to be proven to my satisfaction.
NB; on the conventional drives, I go out of my way to get non-SMR stuff. That's serious black magic there........
non-sequitur word-salad. Reads as though it's been copied from a badly translated Microsoft / Intel / Nvidia talking-points internal document that's been handed around globally the past 8 weeks. Unfortunately, that document was focused in AI creation of gameplay imagery, not so much on useful human-centric knowledge and decision making.
Look, I've been working in AI and LLMs from time to time over the past 30 years. The legitimate big players have very comprehensive research and incremental implementation programs they've been successfully executing on the past 30 years. Not what's now considered to be old-school algorithm analysis and rollout work, but legit AI learning and inference model implementation.
Those legit players have been very very quiet, protecting their considerable intellectual property and industry position / client base / reputation. One of their key design and rollout points has been accountability and accuracy. Those two key points have been pushed to the background by the current rush for blind funding and copy-paste of "AI-based" into marketing materials and web sites.
is in fact a 'pyramid' scheme being hyped by 2-3 industry players along with about 10 'partners' who are implementing/integrating their tech waaaaaaaaay before it and regulation/accountability is ready to absorb it. I actually don't care about the regulatory piece; the courts will take care of that. But accountability, as in who is liable for AI-induced 'hallucinations', etc.
But those players and their partners were rushed to market by the public 'release' of a cute conversation bot. I have a low opinion of the CTO/Product Managers in those partners who are playing along with this madness out of FOMO.
....that has a 30 year history of creating and integrating ever more sophisticated and accurate 'AI' into their product lines.
I believe that the sudden public access to OpenAI and ChatGPT.v.xx LLMs caught them off-guard and panicked them. Drove them crazy. In the space of 90 days, the firm has spent about 1.5 billion Euro on a few rapid-fire acquisitions, not so much to gain additional tech and competence, but to keep their competition from buying said smaller companies. That spending was top-Dollar/Euro/Yen and has nearly drained the M&A coffers.
How else are they paying for the needed integration? By clipping every non-AI tech group in the company by 10% and directing those headcount 'funds' into support for the new acquisitions.
I give them 3 chances in 5 of pulling all his off without causing serious harm to the overall health and reputation of the firm.
IBM Power has delivered on-board discrete vector processors since Power6 (2005 or so, finally implementing the AltiVec/VMS instruction set in hardware).......
ARM VPUs (using NEON) has had ever capable dedicated hardware available on-spec (implementation depending on the end-user and build foundry) for 7-8 years too.
The Intel solution using SSEwhatevergenerationyouwant has depended on and been a slave to various features (inconsistently) baked directly into the CPU.
In the mid-90s, around Boston, it was common for EMC to take CFOs, CEOs, just about anyone who wasn't in IT, on excursions and trips. Lavish stuff.
Then we found that our CFO had inked a really large purchase deal with EMC with no involvement from IT. None at all. Of course at the time, EMC has no connectivity to Novell or the very standard SCSI-2 ports that 3d and 3d generation RS/6000/AIX boxen used. Even their mainframe connectivity used non-standard channel adapters which required patches to the IBM source including modifying block-device byte counts.
Gear was delivered. And sat on the dock for six months (paid for of course) because of this 'problem'. Started looking for the door when in their death throws, the Pyramid Unix sales guy started buying large screen TV sets for people.
My last act going out the door was to out the CFO to the board of directors.
Company was absorbed into a smaller competitor (!!) a few months later and remained a Novell shop until Y2K and an AIX shop until around 2015. As far as I could tell, the EMC gear was never put into production.
make a determination about something critical
study it for a year, determine a course of action
shelf the determination as politically untenable
wait for a change in government
let the prior determination slip back out into the public awareness
bury the action needed in some other legacy program as risk reduction
implement 30% of the original determination, issue a report to the Bundestag, let attached and associated bureaucrats retire
................that was powered on once a year for 10 days to support end of year tax calculations and reporting.
No one could be bothered to actually extract the data, even to Excel. Was easier for them to continue to pay the JDE, then Oracle, support contract than to staff and manage a data capture/migration project.
And as could be predicted by everyone here at el Reg, one year the system refused to power on. Deadder than door nails. Not even clicks or heterodyne whines from the power supplies.
Oops, too late!