>Presumably the tuning also voids the powertrain warranty,

Yes, and warranty is one reason VW's fee is somewhat justified. Upping the power places more stress on the driveline and suspension components. I don't know what their European warranty is, but in the US, it's 4 years. Probably not a big coincidence that the lifetime cost is very close to the cost for 4 individual years.

I don't know if they will succeed; but that's why I suspect that one major entry attempt will be the "empower paraprofessionals" line; and specifically avoiding being construed as a 'medical device'.

Obviously medical device vendors aren't going to just ignore the possibilities; there's already a fair amount of signal processing going on in some areas and if 'AI' is either trendy enough to merit a rebrand of what they are doing already or promising enough to be an addition to the processing pipeline they'll certainly do it; but getting tagged as dealing in medical devices has significant regulatory implications that medical device outfits are familiar with but bot generalists are not; and, realistically, a lot of tech 'innovation' is really a mix of enough tech to make it look like 'tech' along with enough regulatory end-run to provide a cost advantage vs. the incumbent.

Ideally(from the perspective of the vendor) you'd essentially pull an Uber or an AirBnB and, when the medical device regulators are in earshot, be selling a product that is merely a humble, and useful, personalized reference and self guided continuing education aid that should be regulated like a pile of flash card(not at all); while, when the haggling over what nursing home stuff you need a nurse for and what you can do with a nurse assistant or patient care technician is being done have a bunch of plucky, heartwarming, paraprofessionals advocating for their right to do more to drive patient outcomes thanks to the glorious future of advanced personalized learning. The less lucrative; but probably softest, target would be the various nurse-staffed telephone and video link telehealth services that do first-line medical questions and 'is it probably fine/should you really get it looked at' type questions; which can presumably be legally replaced with generic call center bots if you strike all references to 'nurse' and put enough disclaimers; but would see greater consumer acceptance if you could still market them as 'nurse' or something that sounds similar.

I assume that some companies will be at least indirectly involved in both; but I suspect that you are really looking at two fairly distinct product 'tracks', so to speak:

People aren't going to ignore medical devices and well formalized specialties like radiology; those are the ones where you'll actually need to put in the work and deal with medical device certifications and ongoing scrutiny of your system's machine vision behavior vs. radiologist readings; but there's absolutely enough money on the table(along with potentially just-plain-unavailable capabilities with keyhole surgery tentacle robots or whatnot); but that is too obviously a 'medical device' to really play fast and loose.

Trying to chip away at nurses vs. paraprofessionals, though, seems like more fruitful ground for what's ultimately a savings-oriented regulatory end run with enough tech to not be too blatantly visible as such.

Not all of them are; but "Devin the AI software engineer" has had me in a weirdly stubborn torrent of ad spend, so I know at least someone is doing it(and the place is rotten with "AI SOC Analysts", including ones that basically just seem to be the same EDR heuristics the company was selling last year dumping text into an LLM that has been told to apply an executive summary tone); and you can't swing a stick without hitting someone describing a service account as a 'virtual employee' in a thought-leadership-for-morons thinkpiece. Definitely some of it going on.

My impression has been that the ones trying to sell to individuals either primarily or as the first move to create demand for business sales the the vocabulary of tools; but the ones trying to skip directly to c-levels looking for a crash project are much more likely to ascribe job descriptions to their bots.

"may have a different interior structure or evolutionary history compared to the other giant planets."

I'm pretty sure we already expect that:
- axial tilt of about 98 degrees
- magnetic axis tilted 60 degrees from rotational axis
- magnetic axis offset from its center
- axial spin is retrograde ... all of which, to me, suggest a collision with a sizeable body OR capture but the orbital dynamics of that would seem unlikely.

... My company has an elaborate phishing training regime, but at the same time:

- sends out corporate news and hr internal things with "click this" shortened links

- connects essential/critical functions through outsourced partners (like fucking SharePoint) that demands you use your CORPORATE UBER SECRET ID AND PASSWORD on some password popup to a completely different domain

Even if you're conscientious and cautious, these corporate practices train you to obediently do things that are objectively unsafe.

Maybe I'm just unaware, but I'm astonished that the subscription seat heaters weren't instantly hacked just on principle.

The other reason to be deeply cautious of phishing training is that it tends to (when not just plain trivial either because nobody much cares to lovingly craft it to blend in with their specific environment or because they don't want awful result numbers) focus on the risks that are most amenable to technical solutions and waste the time you could be using for the actually dangerous stuff.

Even fairly middling mail filters get a lot of the really lazy stuff; and if you don't want people clicking on Important.doc.exe you just tell the mailserver not to give it to them; not try to train them out of double extensions. If they keep falling for fake login pages; well, that's what the FIDO2 requirement is for.

It's when an account gets compromised at a supplier and a nice looking email, legitimately coming from their infrastructure, body including knowledge of past interactions with them, asking accounts payable to please make a few updates that you have a problem you hope you actually spent time drilling people on proper procedure. Those ones are, at a technical level, impeccably legitimate; and a great way to send tends of thousands of dollars into the ether really fast.

The difference between medicine and IT is that that the titles are totally non-binding when it comes to IT.

It does add a certain amount of faff and confusion to trying to compare titles between organizations; but there's just not much urgency or cogency to trying to nail down titles when there are no requirements beyond confidence, misplaced or otherwise, to do anything. "Developer" vs. "Software Engineer" can sometimes tell you something about how a person thinks about what they are doing; but it's not like only one of the two can sign off on a production release.

I suspect that they (likely correctly) are worried about attempts to put bots in the chain delegating to "unlicensed assistive personnel", who are currently only supposed to be doing patient-facing stuff under more or less close direction by doctors or nurses. Unless you've got a really fancy robot or are specifically looking to do machine vision on MRIs, the logical point of entry would be a bright, cheerful, logo and some stuff about how AI is 'empowering paraprofessionals to drive patient outcomes' or something; with the selling point being that you run more healthcare assistants with fewer doctors and nurses if you just buy in to the notion that the clinical support chatbot is capable of doing the task delegation normally handled by nurses or doctors.

Given the significant financial incentive people would absolutely try that if they had the chance(in some of the less-supervised corners they already do the informal version where you just understaff and de-facto leave more of the job than you are supposed to to the cheapest hires; presumably the bot guys would range from fairly cynical whitewashing of that situation so you can bring it to more visible areas to sincere, possibly even evidence driven to some degree, attempts to do clinical expert systems).

Script kiddies never really went away; they just got a bit more professional, so (while it is possible that 'AI' is or will enable script kiddies at some point) the difference is much less visible.

The quality of some of the attack toolkits is a lot higher, especially with ransomware there's real money on the table to actually pay for competent work, though the low cost of mounting attacks mean that opportunists using cracked copies of older tools, lightly modified proof of concept releases, etc. are still banging on the doors as well.

Especially with it being hyped so hard for 'legitimate' business and marketing communication I suspect that 'AI' is genuinely helpful to the phishing guys whose grasp of the target's language is less than idiomatic; but the effect is vastly less dramatic in terms of what malware and discovery tools people are using.

I suspect that the medical ethics of this would be...fraught; but it would be very interesting to know whether that's pretty much entirely the mechanism; or if there's some combination of "reduction in ADHD severity is good for you" and "psychostimulants are useful for certain tasks" at play.

'Transport accidents' jumps out as the one that probably isn't unrelated to ADHD, attention is pretty critical on the road; but where there's a fairly obvious group of tired but otherwise normal drivers who would probably be a lot safer if they had a little adderall XR in them at the time.

I suspect that the ethics, and even the interest, in trying to tease some of those boundaries out by slicing up a bunch of additional test subgroups and medicating or not medicating them differently would be a hard sell; but you could certainly come up with some comparisons that would be interesting if the implementation cost were low enough to be justified by idle curiosity: would an ADHD group not medicated for ADHD; but on modafinil or similar, see substantial traffic safety benefits related to the wakefulness effects despite the very limited psychostimulant effects? How close would their change, if any, be to just taking a baseline test group and doing the same with them?

Unless the people running the study were total lunatics or exceptionally hard up for data(which seems unlikely for a fairly common condition); I suspect that they'd just ignore self-diagnoses.

Since you can't get the required prescription otherwise the 'medication' group is going to be essentially entirely people with formal diagnoses(almost certainly of varying quality; a GP mostly being willing to take your word for it and trying to get you out in under 30 minutes counts as much as 20+ hours over a period of time with a psychiatrist who falls on the somewhat skeptical side of the discipline in terms of ADHD diagnosis both count as a diagnosis and as a prescribing physician; but you will not be getting a schedule II/Class B just by answering some tiktok quiz); so allowing self-diagnoses into the study would mean that the 'medication'/'non-medication' comparison is now between groups with even more potential confounding variables, which is going to send you into methods hell for little obvious benefit.

There's probably already enough hassle with confounding factors: pharmacy availability for scheduled drugs is always a huge pain vs. unscheduled ones(allowed prescription durations, number of allowed refills without a fresh prescription, ID requirements, limits on delivery or 3rd party pickup, sometimes additional restrictions on transfer of a prescription between pharmacies); so even if it's all NHS and they can omit some of the US insurance-related stuff; so they would already have to be considering how to avoid the possibility that their study is actually "People who can, reliably and repeatedly, go through the hassle of maintaining access to medication and remember to take it more or less regularly are more pulled together than people who fail at one or more of these tasks" rather than "medication has these effects"; and adding a bunch of randos to the study would not help at all.

Social media already gives me too much access to too many people's inner monologues that should have been kept "inner", thanks.

Weirdly written summary.

Let's be clear: the summary repeatedly talks about the drugs being the thing that's giving all these beneficial results.

It's the suppression of ADHD that is giving these results -the drugs are apparently successful in their job.

"He added that he thinks it's "insane" that some AI startups with "three people and an idea" are receiving funding at such high valuations."

Because the insane thing is definitely the valuation of the random little guys with costs barely distinguishable from zero; rather than serious professionals who need to set some billions of dollars a year on fire just to keep the lights on.

I'm sure that this view of who is and who isn't bubble is entirely unrelated to which kind of AI startup Mr. Altman is running.