This is the routing of parts of a task to cheaper models dressed up as some kind of innovation in efficiency. Why anyone takes anything this guy has to say at face value is beyond me.
I have watched multiple times now where those most excited are the least capable. They prompt something out of an LLM and feel the thrill of vicarious creativity. They have never been able to do something like this before, and it feels good.
They show their barely working demo to management who is also excited. It kind of works if you don't look too closely, but includes features and screens no one asked for or can explain, tables and graphs which turn out to be static artifacts using made up data (in one case I saw a pie graph which when summed, exceeded 100%), is a security nightmare, and simple changes must be made across the handful of functions which are hundreds or thousands of lines long, because there is no separation of concerns.
Eventually they turn this steaming pile of "prototype" over to an engineer who is asked to get it ready to ship because "the product team was having trouble adding the few extra features, but it should be easy, just use AI".
I can't tell if this is stupidity, an elaborate joke, or maybe I am already dead and this is some kind of purgatory.
No one getting rich off data centers wants to live next to one, which kind of tells you everything you need to know. They are a gigantic externality, plunked down in the midst of people who do not benefit from them, but have to bear the costs of the noise, pollution, and increased energy and water costs.
Regarding what is available, imagine a system with reports and dashboards showing a timeline of which application was in focus and for how long, metrics on "activity" like keypresses and mouse clicks, periods of inactivity, lists of websites visited, whether you are joining scheduled zoom meetings, whether your camera was on, when you badged into and out of the office, periodic photos being taken from your webcam, geolocation on where you sign in from, and I could go on.
Most of these things are available bundled with most of the business Microsoft subscriptions while other telemetry comes from other tools or homegrown sources and is available to managers and IT staff on demand. Now, most of the time no one was really looking at most of this unless they had a reason to, and while I am no longer in this end of things since LLMs have reached this stage of maturity, I can imagine they are now being tasked with constantly watching for patterns in worker activity which deviate from the expected norm and are fully capable of notifying your manager automatically along with a detailed analysis of your activity.
The thing to understand is that the modern office is a veritable panopticon.
Considering Richard Dawkins has recently succumbed to the same delusion it is a reminder that no matter how intelligent someone may otherwise be, we are all human and have certain tendencies and blind spots; anthropomorphizing non-entities being one of those.
Mythos is another grand slam by the Anthropic marketing team.
They issue a provocative public statement dressed up as a safety advisory and then sit back while the media and various industries (who for some reason take this all at face value) generate them tens of millions of dollars in free publicity through their own overreactions.
It needs to go well every single day, and only needs to go very poorly once. Not to conflate LLMs with actual super intelligence, but for this (and many other reasons related to basic human dignity), this is not a technology that a responsible society should be attempting to build. We need our very own Butlerian Jihad