Joe

It’s not just about current LLMs, though. LLMs are being made to work 24/7 on the next generation of models (and not just LLMs), which may be quite novel or just more effective. The next generation might be the one to worry about… Or the generation after that.

LLMs can be used like an army of monkeys with typewriters, with evals guaranteeing progress. It’s inefficient, but effective.

Assuming it is possible to make progress (and I think it is), the logical conclusion is that progress will be made, and AGI and ASI will come

it gets integrated into automated killchains including a nuclear arsenal, and i hope noone is stupid enough to

Have you ever met a human? 🤣

The main issue was the motherboard. It’s too “new” and I ended up having to build a bunch of drivers to just get my computer to work exactly what Windows provided out the box.

Ok… To be fair, the drivers for Windows are probably all third party drivers. HW companies tend not to provide standalone drivers for Linux - either they contribute specs and/or patches that get incorporated to mainline, or do squat and eventually someone will reverse engineer it and create a driver.

There is so much hate for Windows, but you can’t beat their commitment to stability and backwards compatibility.

This isn’t the problem you just described, ftr. Linux often has a delay in supporting the newest hardware, but then supports it well and for a long time. OSS in general is good at that.

For example: my Wacom tablet is no longer officially supported on Windows (by Wacom), while it works out of the box on Linux.

Another example: Windows 11 refuses older hardware - not backwards compatible.

If you are talking about software APIs, that’s a different story. eg. There’s not much point in targeting Linux native APIs for games, because wine usually works better.

Mumble?

To be fair, protecting credentials and important data is the company and individual’s responsibility. The building blocks to restrict access are there, but are often not leveraged (even by large companies with the ability to invest)

Sandboxing is one of them: Both Codex & Claude’s sandboxing is reasonable (sandbox-exec, Linux cgroups & seccomp). Many others are lacking, sometimes deliberately.

I do most coding with Pi these days, and I have it heavily sandboxed. I expose sensitive services via a localhost network service with auth (typically for running scripts outside the sandbox). Reads are limited to the system binaries/libs, nad writes to the project dir & Pi’s own dirs. If I choose to give a particular session creds, then I have to be very deliberate. I also force egress traffic through a proxy (just logging for now, but I have plans)

$14000 in API pricing is not $14000 in costs, though. Costs are hard to calculate because of the huge capital outlays and unknowns about hardware lifecycles, various business deals, and limited public knowledge.

It’s likely that inference costs for good-enough models will go down over time. China’s API pricing tells us the direction already. Energy costs will be a driving factor in the west, I guess.

So… they are almost certainly subsidizing plans right now, but on average, it won’t be by sooo much. Your average ChatGPT user will hardly use Codex, for example. Your average developer is not token-maxxing either.

Why are they subsidizing plans? To build a sticky customer base … which means they want you to stick to their tools - their coding agents/harnesses, their integrations, etc. Models are/will be increasingly interchangeable, so they are building sticky ecosystems instead.

A couple of tricks I use:

• an apparmor profile tied to a shell script that wraps other commands … it restricts read & write access to a scratch directory … perfect for builds or one off scripts.

• iptables rules & cgroups to restrict network access… I have a setuid wrapper that drops privs again…

• bwrap and mounting only what’s necessary… quick to get going.

• custom landlock wrapper, similar to apparmor but allows for quick userspace wrapping.

They can be combined too.

I was recently reading about Talon Voice, which sounded quite interesting with lots of usability hacks. Unfortunately, not an option anymore if you are a Linux user: https://www.osnews.com/story/145162/accessibility-input-tool-removes-x11-support-doesnt-want-to-support-wayland-users-caught-in-the-middle/

AI is just the next excuse, and there will always be a next excuse.

checks lifetime jellyfin pass price … Yup, still $0

In many cases, yes. A difference now will be the long-term size and composition of the teams (smaller & more generalists, with PMs, POs & Architects just as likely to contribute code as engineers)

2 pizza teams can become 1 pizza teams who can manage an entire product/component, or more. And those 3+ pizza teams can strip the fat or split into more productive teams.

I think we’ll also see increased demand for platform/deployment standardization and concentrated/novel support structures, as teams start biting off more than they can chew, along the the desire for out-of-the-box guardrails around AI code & tools.

Very true.

Given that AI is particularly useful at increasing alignment (when applied smartly), and that this is often a role delegated to middle managers, it is quite likely that flatter orgs will happen.

The need for top-tier technical, product, and business judgement and problem engagement will increase, while the need for muddle-through managers and similar roles will decrease.

We’ll see more initiatives organized end-to-end by small groups of smart people, with virtual teams/coalitions forming to bypass “archaic” processes and deliver meaningful results. We’ll see a lot of sloppy failures along the way too, but the overall trend seems clear.

spits on ground, squints reeeeaaal slow

Now listen here sonny… we don’t take kindly to none of that AI business around these parts.

We is simple, God fearin’ folk, raised on sweat, dirt, and good honest labor, not all of that fancy machine learnin’ contraption nonsense.

Ain’t no place for thinkin’ machines where a man’s meant to use his own two hands. So I reckon I’ll mosey on over and downvote this here post myself, nice and proper.

This was similar to a trick that a few smaller (less serious) hobby-ISPs did back in the days of 14.4k/28.8k modems to take advantage of the “reasonably priced” business plans for ISDN. They’d register multiple businesses at a single address to qualify for the plans, then balance new egress connections across the pool using squid and other magic. Fun times…

How is it being enforced in Australia?

A webcam photo by the website or a specific third party service, ID verification through a “trusted third party” process, or a checkbox to confirm age?

How much information does the website get over and above “user is over 16 years old”, and how much does the government get, if any?

Explain your reasoning, please.

If it helps to accurately fill in the details correctly in the backend system, which are then properly validated or escalated for human review/intervention (and let the human requester choose the escalation path too, as opposed to blindly submitting), then it sounds great.

Guided experiences, leading to the desired outcome, with less need for confused humans to talk to confused humans.

I want the same for most financial approvals in my company. Finance folks speak a different language to most employees, but they have an oversized impact on defining business processes, slowing down innovation, frustrating employees, and often driving costs UP.

Judges usually don’t know this stuff, but they primarily work with systems and software supplied by the state…whose experts should know what they are doing.

My bet is that this guy decided to work on personal equipment, probably in violation of the rules. Being a judge, he’s unlikely to be sanctioned for it, and will certainly learn from the experience. If anything, there may be some internal discussions which we’ll never hear about.

Law is an area where AI can add value, though… searching through past rulings and legal opinions is tedious, and anything that can assist to find needles in haystacks would be welcome. It shouldn’t be used to write legal judgements or arguments though…

I don’t think you need separate laptops, but a separate router may be useful.

If you use Linux, you can have apps isolated to their own lightweight network namespaces (like containers), using different VPNs. Otherwise VMs can serve a similar purpose on Windows and Macs.

Iptables can also be used to block traffic, and force it through proxies (which can be whitelisted by uid/gid) or VPNs.

If you want a more secure VPN setup, I’d even recommend having the VPN(s) running on the router (eg. portable OpenWRT setup) so your laptop never gets offered a public IP / connects directly to network. Put a proxy on it for special (eg. DNS based) routing exceptions, like banking from real IP, reddit via the US, etc.

Disable IPv6 on your router or primary interface, and enable it on your VPN. If anything can discover an Internet IP on your PC, the link can be formed. Worst case, you are not using the VPN for IPv6 at all.

They could. The protocol also supports IP spoofing, so doxing could also be a thing.

For individuals, it is a time consuming and costly legal process, whether justified or not. For the law firm, it costs a few cents per letter, but they get a few hundred (or more) euros when some sucker pays.