As a person who used to work at MSFT:

I can almost guarantee you there are a whoooole lot of people who have made their careers basically championing the very old chat bot model, and they are probably now either directly in charge of the OpenAI stuff, or at the very least ‘stakeholders’.

They will do nonsense corporate bullshit to make them selves seem very important, never really wrong about anything, and this will result in extremely slow and gradual actual adoption of the GPT stuff, all the while stressing all the reasons their old stupid bullshit can’t be seriously modified because of reasons that have to do with synergizing with other MSFT products.

The process of the company gradually figuring out that none of that matters when it comes to producing something that is actually better will be slow, painful and incremental.

Itll probably take half a decade.

…

For reference, as an aside, I was doing a contract of DBA kinda stuff when they unveiled Windows 8. We had to dogfood it, ie, the MSFT process of everyone working at MSFT has to beta test everything else MSFT is making.

Well… Windows 8 initially broke basically everything we were using to actually do DBA.

I got angry and pointed out that Windows 8 had removed the ‘windows’ from Windows. The initial version was soley the tablet based design, only allowing a maximum of two ‘panes’ open at a time.

We had to wait about a month for the various problems with SQL Manager Studio to be ironed out, and for them to basically allow the option to just use the more or less Windows 7 desktop for you know actually working on our PCs.

Point of me mentioning this is: I saw how ludicrous this all was, and was frequently verbally abused by our team lead for pointing it out.

Youre not allowed to go against the grain at MSFT unless youre a big dog. And… you become a big dog by bullying people and vastly overstating the necessity of what your team is doing.

The culture there is downright psycho and sociopathic.

MSFT appears to still be using a fundamentally old chatbot model that they’ve just slapped a bunch of extra ‘features’ (namely, Wooow! It has APIs and works on other MSFT stuff!) to, much like Bethesda’s game engine.

Probably barely different from Tay in terms of broad conceptual design, just patched and upgraded to do what it does faster.

The core design is garbage, and just like Windows itself, its nearly certainly a giant fucking mess of layers upon layers of different versions of itself hiding under a trench coat, all standing on top of something 10 to 20 years old.

Absolutely amazing.

My guess is that at this point there are so many user prompts its received so far in its training set that bring up both Copilot and privacy concerns that it first interpreted the question, then searched for the most common topic associated with itself (privacy), then spit out a hardcoded MSFT override response for ‘inquiry’ + ‘privacy’.

https://en.m.wikipedia.org/wiki/Intelligence

Intelligence has been defined in many ways: the capacity for abstraction, logic, understanding, self-awareness, learning, emotional knowledge, reasoning, planning, creativity, critical thinking, and problem-solving.

LLMs are pretty capable of abstraction and understanding.

Though they obviously use logic in that they are constructed from/of it, they are not really capable of actual logical analysis, beyond emulating it.

They can’t really do any of the other attributes of intelligence at all, beyond basically decently to poorly emulating them.

Anyone who buys into this deserves the financial ruin that will so obviously happen to them.

Sats that beam data to other sats do not have to worry about the atmosphere, nor are they using anywhere near the kind of power involved to fry the other sats. Its orders of magnitude greater power for that, which means more more weight and thus launch cost.

Beam decoherence is a /huge/ problem when trying to go from ground to low earth orbit.

You would again end up needing a pretty significant power supply along with exceptionally precise tracking.

Im talking military grade equipment here, massive expensive and complex. Not something you could whip up in your garage, unless you worked at it for a decade, and if you did that, youd end up in jail.

I really want to stress how precise your tracking needs to be. Assuming you precisely know the orbital trajectory, your /exact/ location, the rotation of the earth… you would need to have a mechanical system capable of sustained tracking to… what like a few (roughly 3 by updated calculations) arc seconds, something like that, to hit something /and stay on target for probably 30 minutes/ that is 120 miles away, roughly the size of an SUV

EDIT: Fixed up my numbers, I was thinking in terms of the wrong unit.

Point is… this approach requires an astounding degree of tracking precision that is basically impossible unless you are a defense contractor.

Tracking a thing this accurately alone is practically impossible. And I mean that literally. There is no practical way you can do this, unless you consider starting up your own engineering firm to solve this, and you are allowed to use a whole bunch of tech with current security classifications, unless you consider that practical.

If you do, hi Elon Musk, didnt realize you were on lemmy.

Are you talking about launching your own satellite with the ability to aim a laser at another satellite while in orbit, or are you talking about attempting to point a ground based laser at something moving at roughly Mach 24 or faster?

Beam decoherence is a pretty big problem when you are lasering through the entire atmosphere, and both scenarios require an astounding degree of precision.

This person asked if they can make PopOS secure via TPM.

I am saying that while yes, you can, there isnt much point, because setting up LUKS to work with TPM is inconvenient, easy to fuck up, and basically offers no additional protection against all but extremely implausible security scenarios for basically everyone other than bladed server room admins worried about corporate espionage who are for some reason running bare metal PopOS on their server racks.

Like the only actual use case I can see for this is /maybe/ having a LUKS encrypted portable backup drive, but even then you can still base the encryption key in the actual main pc’s harddrive without using tpm, though at /that and only that point/ are we approaching parity between the difficulty of using or not using tpm to accomplish this.

Oh ok so the use case here is if this casual linux user asking this question has only their harddrive stolen from their pc or their laptop in their home or apartment or workplace, not their whole pc.

Mhm that seems likely.

I guess this maybe makes sense if youre running like a server room, but chances are low thats the actual context of this question.

Why would you run PopOS on a large operation’s servers?

Again, you dont need to use TPM to have a LUKS encrypted partition unlock automatically on boot.

You can just do this via the standard drive management included with PopOS.

Its not though, it requires a ton of extra work to set up, isn’t necessary, doesn’t allow you to do anything you can’t do without it.

You also do not need TPM to automatically decrypt drives on boot, I have also done this on PopOS for 3 years, with TPM disabled.

Ok… so… if you have TPM… and LUKS…

You still have a scenario where the encryption key is still on your physical device, LUKS with or without TPM, or … some kind of TPM based Linux encryption solution I have never heard of?

Does Windows Secure Boot work on Linux via the TPM?

No…

Am I missing something?

Theres no point in involving TPM in securing a linux computer.

In a scenario where you’ve physically lost your computer, using TPM or not it wont matter if your pc gets into the hands of someone who can attempt to brute force the keys.

If your pc is remotely compromised to the point it has something on it that can grab your keys, it also will not matter if you are using TPM in some way.

The only practical use of full disk encryption is if your linux pc and or laptop gets stolen and falls into the hands of a non tech savvy person, and in that scenario, going through the trouble of correctly binding LUKS to TPM will have just been a waste of time.

Thus, you should probably just use LUKS and not bother routing it through TPM.

Sure but you dont need to use TPM at all to use LUKS.

You can store the encryption key on the harddrive, in the LUKS partition layer.

Like thats the default of how LUKS works.

Im really confused why people think TPM needs to be involved in anyway when using LUKS.

Generally speaking you have to go out of your way to correctly cajole TPM v1 or v2 to actually correctly interface with LUKS.

Not necessarily?

Im pretty sure I used PopOS for 3 years with LUKS encryption with TPM disabled.

Why would you do this when PopOS offers LUKS1/2 disk encryption?

(nevermind)

You forgot the perfectly uncanny valley eyes on the forward display of the thing.

Digitized, slightly too large, and the ambient lighting conditions dont act on the goggles the same way they do on human flesh.

I find the effect roughly comparable to a real human face being replaced with that particular style achieved by just having the sort of news paper cut outs of eyes and mouths in some old animations, or that creepy instagram filter.

Hah, Ive gone uh, full tilt, and actually am working on making a game myself that will hopefully /actually have meaningfully innovative and compelling gameplay/, and i dont plan on or seem to have any real need to fall into the kickstarter/early access trap.

From a developer standpoint, both those approaches mean deadlines and managing expectations, which is basically maddeningly stressful and soul crushing.

From a gamer perspective, more often than not that means throwing money at a promise that at best will not live up to the hyped experience you have generated with the fandom, and at worst is just a total bust, failure, or scam.

So yep, my plan is tinker away for a year or two until the fundamentals are technologically sound and the actual gameplay is unique and compelling.

Then, only then, would i maybe release a demo or in depth teasers or testing session footage.

Yes thats right. Testing. Remember when games used to actually be playtested, not just for bugs, but for actual gameplay experience?

Many of at least my favorite games and mods were hugely shaped by tester feedback that radically reworked certain game elements to solve unexpected gameplay problems, or to further an idea that the testers found fun or useful that tje devs didnt even realize was really possible in the world theyd constructed.

Anyway… woo video games, shit sucks mostly these days but there are some notable basically niche exceptions, and hopefully i can make something thats at least niche successful.

In the words of a person i truly do think is an actual genius of game design:

These things, they take time.

Time where no one has any real clue wtf youre actually doing, haha.

I dont know the technical terms, but, is it possible that due to either general wear and tear, perhaps being physically dropped, or perhaps some kind of power surge… that basically some kind of actual mechanism or electric motor responsible for actually moving the … lever thingy (arm?) with the read head… maybe its actually just broken?

EDIT: I cant make it out from the image, but youve got a sticker with a barcode on there, and some kind of id number. Presumably as thats inside the HDD, it might have some relevance to some internal product designation code, and there might be more serial numbers or model numbers etched into the thing if you look at it closely with a flashlight, and maybe glasses or a magnifying glass depending on your eyes.

If youre lucky, net searching the specific numbers may reveal more information, and you /may/ be able to find an actual technical manual.

Good luck though, in my experience that kind of shit can easily take a whole day of rabbit hole diving.