What I wonder is: how are they going to call in all the AI workers for meetings?

And what doe RTO look like for an LLM?

There will be consultants. Most will be charlatans with an LLM, but there will always be a small number of people who learn the craft because they’re interested… and they will command high salaries by those who understand quality engineering.

It’s not like this is the first time our society has done this. We did it to textiles, then we did it to farming in general.

The difference here is that automation of thought is what’s being promised, but that’s not what’s being delivered. But then, for many of its applications, real thought was never needed in the first place.

Anyway, back to my actual point: manual software development will become a niche hobby like using a hand loom. The skills will survive, but more as a curiosity than a common career path.

I hope I’m wrong, but it all depends on how long it takes the bubble to burst. If the LLM companies get a critical mass of dependence before it does, this will be the result.

Yeah; my point is that it could be as useful as a button board, but for most people it’ll be no different than something that detects ANY audio from a pet and plays a a random sound from a collection. And once the novelty wears off, they’ll ignore it the same way they were already ignoring their pet.

But for those few people that would already invest the time and effort in a button board, this could be useful as a mobile equivalent.

Hmm. The article appears to conflate multiple things?

One of them is “viewing” the RF spectrum to build up an image. The other is reading the unencrypted beamforming data from a router. That second one depends on people carrying a WiFi-capable piece of electronics with them doesn’t it? There has to be something for the beam to focus on, some sort of beacon signal.

Although I guess all it really needs is for the person to step between the router and a device connected to the router; that should enable analysis of the disruption patterns.

It’s definitely an interesting way to shrink those button boards. But it’ll only be useful if the owner gets trained to pay attention to their pet.

Yes. You could experiment with DNS on a DOS laptop with an i286 processor.

The hardware doesn’t really matter as long as it can run a TCP stack and a DNS server, which your reference machine should be able to do handily.

Anyone is free to se up their own DNS, which they can tweak however they want.

An interesting thing about DNS is that there is absolutely no validation done that requires the server you talk to to have a chain of authority back to the official root servers.

You can use DNSSec to add validation and encryption layers, but nobody’s forcing you to use that at this point.

So go ahead… set up your own virtual lab and play around. Get a feel for how it all works, and what breaks/works unexpectedly when you tweak a component.

Just don’t mess with an authoritative server.

Despite how the information is presented to you, LLMs aren’t reasoning engines. They don’t come to conclusions based on the information fed to them; they don’t learn. They’re complex probability sieves that try to predict the expected output based on a set of input tokens.

So they have no concept of anything, including time. They look at a prompt, compare it to the prompts they were trained on, and with a bit of random path selection thrown in, try to find the best path to the goal.

My guess is that up until now, this hasn’t required taking time into consideration. It would definitely be possible to create a model that valued time context, but apparently that hasn’t been an important variable to this point in time, at least for frontier models.

Indeed. And how do you make that quantitative? Because otherwise, such laws will be weaponized by the hubris factory.

“Biden won the election.”

“Lie! Arrest that man! We’ve all seen the truth in the media!”

As AI is leaned on more and more to patch and defend against vulnerabilities, the value in sneaking back door logic into the models or finding vulnerabilities in the models themselves will increase.

This is one of the reasons for the sharp increase in CI/CD attacks currently; instead of targeting the deployed software, attackers are now targeting the development pipeline where they know there’s minimal human oversight and a lot of security protections are intentionally disabled.

Fruit flies have been studied very thoroughly in this manner. A lot of their brains are hardwired to sensory inputs and motor control, but the interesting thing is that their thoughts aren’t just made up of the connections between nervous tissue; the connective tissue itself is also used to store and modulate information.

What they don’t have is a highly developed cerebral cortex; so while they can feel pain and anxiety and probably anger and frustration, they don’t have much ability to reflect on any of it after the fact.

Animals that can do that include cats, whales, elephants, octopuses, parrots, ravens and jumping spiders.

On that theme, MacGyver too.

I watched the plane crash into the building, then witnessed through media reports the cleanup and demolition?

There was definite structural damage, as I said, but the impact didn’t bring the towers down.

What Wikipedia says doesn’t disagree with that.

What level of accuracy are you looking for? The towers weren’t destroyed by the jets; they punched holes in the side and damaged the overall structural integrity while starting massive fires and causing glass to shatter.

The towers were demolished after the fact via controlled explosive charges because they were no longer safe to enter.

I already answered this in the one about the two guys. Drowning is the cause of death, but usually something else causes them to keep their face submerged long enough for them to lose consciousness; either slipping in the mud and hitting their head, or getting disoriented in the water and panicking, not knowing which way is up, or intentionally keeping their head down, or having their head held down, etc.

So drowning is the cause of death, but many things could be the cause of drowning, most of them quite accidental.

It’s why you need a spotter for people who are swimming; people have a propensity for drowning in any body of water large enough to submerge their face… including bathtubs.

I haven’t touched anything that used Adobe DRM in over a decade….

Sounds good… who do I pay $130/year to to get to drive an EV?

This is a good point. Con men target people who have been conned in the past, because the person’s personality hasn’t changed — they’re still an easy mark with the same levers that worked last time.

I do indeed :D

Although I still use XMPP.

It was?

I started streaming video in around 2001. Over SMB. Using VLC as the client.

Then I switched to using iTunes as a server.

Then came XMPP/Kodi.

And eventually, Jellyfin.

Every few years, I’ve tried Plex, and it’s never done quite what I wanted, and required security/privacy compromises. About the only thing it has going for it is that the client and the server will run on just about anything.