Septimaeus

Should’ve just used IP location so users could simply pick different VPN servers to simulate either a quick flap across town or a brutal multi-week transcontinental migratory path

Fair enough, though I’m not sure personal and psychological analysis of your life is something AI should ever be entrusted with.

While I don’t use LLMs, IMHO unreviewed code is unreviewed code. If it’s unreviewed, that means no one has taken responsibility for it, and so it doesn’t belong in any codebase without that giant disclaimer front and center.

Edit: lol @ downvoters who “vibe-read” this comment 🤦 the toxic element of lemmy who think we can’t see who they are

Atm the meta for local inference is unified memory and “routed” local agents (multiple smaller role-specific agents in a trench coat)

The former is standout for cost efficiency (e.g., 4x RDMA 48gb Minis for a 192gb cluster @ $43/gb vs a $45k b200 alone)

The latter is standout for many things, including resource efficiency on smaller machines

I mean, you can have that today, in some form or other, since that’s just how local AI works.

Hi @LordCrom@lemmy.world

Hi @pHr34kY@lemmy.world

I was talking about the 802.1X/MDM profiles of enterprise networks since at least that would indicate a baseline attempt at appropriate equipment and configuration given the footprint and number of active stations In hotels.

lol same at Union Station WDC and all Amtrak trains (if the wifi even works). Finally switched to a gl.inet (openwrt) hotspot for work travel and haven’t looked back.

Hotel can see every site you connect to and you’re exposed to SNI fingerprinting by their ISP. Metadata leaks over both LAN and WAN (traffic volume, timing patterns, dest IPs, connection frequency) and both netadmin and ISP can infer your browsing habits without actually seeing the content. Other guests with wireshark and too much time on their hands.

Moreover, many types of traffic aren’t HTTPS, including NTP, DHCP, SNMP, FTP, SSH (without HTTPS), most IoT devices, VoIP, gaming traffic, many application APIs, apps with embedded DNS overrides (really apps in general, especially mobile apps, especially meta and alphabet apps), even email unless TLS is explicitly configured. And many websites simply don’t serve HTTPS and will attempt to redirect to HTTP.

This is without getting into encryption strength and post-quantum standards. And I was referring to 802.1X/MDM enterprise profiles not CA.

Seriously, this is the lowest hanging fruit. Never trust hotel WiFi.

E: infants and their alts

Then you don’t lol but probably will some day

Typically both, where the business is more like (a) their legitimization of the collecting to family and friends and (b) a way to help themselves let go of books slowly, knowing the point of books is not to sit on shelves.

So an over large order of their books that won’t be read by a person is exactly the worst possible customer, rather than a boon to their business.

It’s pcgamer.com no ones expecting Pulitzer work

Depends. 1 for many vocational degrees, 2 for most academic or engineering. Also usually only the 1s accommodate completion by working professionals. 2s tend to assume full-time study.

Haha I like it

All true. To your question, however: until software can be named as a legal fiduciary, CEOs can only be functionally replaced. That is, a real person who can be summoned by a court has to take responsibility for whatever shit the LLM gets the company into.

E: way more replies than I’m used to, most re: same issue, so I’m clarifying here.

Preface: totally not my area, IANAL.

To clarify, I specifically meant CEOs can be named in a lawsuit by the shareholders for failure in fiduciary capacity. I didn’t mean a summons to answer for the sins of the corporation. TLDR: math models can’t be sued.

(But if I’m wrong I’m wrong.)

Then clearly you’re talking about deontological rights and they’re talking about legal rights and both wasting your breath

He’s saying that right was taken away by George W Bush.