[alancalverd]

I just googled my own name, not out of vanity but to find the date of a paper I wrote some years ago. My search engine helpfully compiled this biographical note:

Alan Calverd, a notable scientist, has made substantial contributions to the field of medical oncology. He holds the position of Professor of Medical Oncology and serves as the Director of the Cancer Research Unit at the University of Newcastle-upon-Tyne. His work has significantly impacted the drug treatment of cancer, earning him international recognition1. Additionally, Alan has delved into intriguing aspects of science history, including famous gold rings associated with physics and medicine.......

which has the unique merit of being completely untrue  in almost every aspect - ask anyone at the University of Newcastle, one of many institutions which I have never visited, let alone taught at. So I read to the end of the article and pushed the "learn more" button, which helpfully replied

Alan Calverd, a distinguished scientist in the field of medical oncology, has made significant contributions that have shaped cancer treatment. His work has had a profound impact on the understanding and management of cancer. While I don't have access to specific details about his individual contributions, I recommend exploring scientific literature and reputable sources to delve deeper into his research and accomplishments.

Everyone knows "garbage in, garbage out" but software that unaccountably generates random garbage with no factual input is almost as dangerous as a pope or president.

Would you let AI drive your car? Or even design it?

[Eternal Student]

Hi.

     There is a professor  Alan Calvert  and he does seem to have done those things.
https://acmedsci.ac.uk/fellows/fellows-directory/ordinary-fellows/fellow/Professor-Alan-Calvert-0005997

Would you let AI drive your car? Or even design it?

    It has got one letter wrong in the spelling,  or was programmed to recognise that users often mis-spell a word in the search bar and has automatically corrected for this  etc.

Best Wishes.   

[alancalverd]

But the stuff about gold rings actually refers (with consummate inaccuracy) to my appearance in the Naked Scientists Christmas radio show "The Twelve Days of Christmas" a few years ago.

And whilst your reference does indeed mention a professor Alan Calvert, entering my surname in the University of Newcastle's  own search panel  only offers a visiting female professor Hilary Calvert as the nearest fit.

So it seems that AI not only generates garbage but also infects other websites with it. And

a lie can travel half way around the world while the truth is putting on its shoes.

Even if you would let AI drive your car, how do you feel about "correcting" hypo to hyper? macro to  micro? m to M? In my world, machines do what you tell them, not what someone else thinks you might have told them, and the named driver/pilot/doctor/engineer is held liable for the consequences.

[paul cotter]

Good morning, Alan. It's a beautiful spring morning here in Donadea, Kildare. Spring has arrived!. Anyway on to more serious matters: these programs just hoover up what info they can find on the web and as i'm sure you know there is a near infinite amount of rubbish spread evenly throughout the web. I would not trust AI for anything(Hamdani will have a conniption).

[Kryptid]

Not with current AI, but as it advances I might. I'm not going to use a self-driving car until they get a proven safety record.

[Bored chemist]

Some humans believe that AI could drive a car.
Do you believe that humans should be allowed to drive?


Might there be some sort of issue extrapolating the ability?

[alancalverd]

I've just been sent two current news items concerning self-driving taxis.

Phoenix: two separate Waymo taxis crashed into the same truck within minutes of each other. Waymo blamed a "rare scenario" cause by the angle of the truck on the road which caused the taxis to "incorrectly predict the future motion of the vehicle".

San Francisco: Statewide ban on Cruise taxis after one dragged a pedestrian along the street.

Lawyers are oiling their cash registers. 

[Zer0]

The term ' Artificial Intelligence ' itself seems like a misnomer.

Not the Artificial thingy,
but the " Intelligence " part.

As of it's Current capabilities...
I won't allow it to drive Anything..
Coz the only road it knows of,
is the Highway to Hell!

ps - not against Technological Progress, but it's gotta be the Right kind of Change.

[Petrochemicals]

I would not want ai driving my car, there will be far better systems similar to thoes on planes. Would you want ai landing your plane for you.

[alancalverd]

Colllision avoidance in the sky remains the primary responsibility of the human pilot. This has already caused problems,  spectacularly where aircraft are flying on instruments (either auto or manual pilot)  in zero visibility and under direction from airways traffic control, and more recently where there is a conflict between manned aircraft and a drone. Even the simple rule "steam gives way to sail" can cause problems at low level: whilst a balloon, airship or kite is effectively a stationary hazard, a glider can fly as fast as a Cessna and both are  far more manouverable than a 747. Maintaining environmental awareness and responding appropriately requires communication with other humans and re-prioritising options: do I offer priority to a landing helicopter that has declared "alfa" (casualty on board)  and either risk landing in his turbulence or screw up the  approach controller's carefully planned sequence? Depends on the opinion of the paramedic in the chopper and the fuel state and experience of the trainee behind me.....   

Once we are on final instrument approach, things actually get a lot clearer: approaching aircraft "cleared to land" has priority, and the old Trident could indeed autoland in zero vis, using a radar altimeter full of thermionic valves, and a mechanical computer with gears and bicycle chains.  But recent events show how we still need awareness of other people's conversations, e.g. between the tower and a small aircraft holding to enter the runway for an unscheduled emergency flight. Been there: if I can't hear the other aircraft confirm or broadcast "holding", I assume he hasn't seen or heard me, so I "land long" (not a good idea in a passenger jet, and even a Cessna causes problems for the ground traffic controller) or abort the approach.

Driverless trains work because (1) all the other trains are driverless and under positive central control, and (2) they are solving a 1-dimensional problem. On the road, at sea, or in the air, there are 2Nm or 3Nm dimensions, where N is the number of other vehicles within collision distance at any given speed -- two more numbers we don't know - each with m different unknown priorities.

[evan_au]

I heard an interview with a BBC reporter who recently tried out an AI-driven car. Apparently, UK authorities have taken the approach that cars can be tested and approved to drive on certain specific highways. This test drive was on the major highway circling London. WHen the car enters a highway on which it is approved, it advises the driver that they can take their hands off the wheel (but the driver is still responsible if anything goes wrong).

My car has a milder form of AI - the computer nudges the steering wheel if it thinks you are drifting away from where it thinks the centre of the lane is. But the driver must keep their hands on the wheel (I suspect that it occasionally nudges the steering wheel to ensure I still have a grip??).

[alancalverd]

Not a problem - and probably a help - in the outback, where falling asleep is a secondary hazard (IIRC kangaroos being the primary one) . But a real pain in the backside driving on British roads where suicidal cyclists and arrogant BMWs move at random. Instinct makes you cross the lane marking, and the  car fights back. Even worse is the "intelligent" forward radar that warns you of rapid closure by flashing the entire instrument panel in red and honking a siren when you are trying to concentrate on and avoid the surrounding traffic. I recently rented a VW Golf and spent a couple of minutes switching off all the "intelligent" systems every time I started the engine, in order to proceed safely thereafter.   

[Zer0]

https://www.theguardian.com/us-news/2024/feb/12/waymo-car-fire-san-francisco

Imagine if You were inside one of them when the Unfortunate incident took place.

Uncertainty isn't only at Quantum scales, Emergency arrives without a Warning.

Fully automated sounds pretty cool & efficient, but a Manual mode is a Must!

ps - & wat bout all those humans who are employed in the transportation sector?
Universal Basic Income is a
Farce!

[alancalverd]

Time was that before taxying for departure, you checked that (a) the autopilot disengaged (and told you it had done so) when you pushed the red button and (b) even with autopilot engaged, you could manually overpower it.  Then Boeing decided that pilots were to stupid to follow or even learn simple departure procedures, installed an auto trim that couldn't be overpowered, and didn't tell anyone how to switch it off. 300 dead to date.

The great thing about human drivers is a strong  selfpreservation instinct, coupled with a learned "if that doesn't work, try something else" process, which solves many problems that the designers and programmers haven't considered.

[KaydenceBradshaw]

I?ve always been a bit skeptical about AI taking over driving, especially with cars that I actually care about.

[Eternal Student]

I?ve always been a bit skeptical about AI taking over driving, especially with cars that I actually care about.

       That's quite an amusing one-liner, thank you.   The other road users and pedestrians don't matter as much.

[Bored chemist]

The great thing about human drivers is a strong  selfpreservation instinct

This is less comforting if you are a pedestrian...
It's one of the things that will be hard to sell about an AI car.

Salesman to buyer "And here's where you sign to say that you accept that, in some circumstances the AI will kill you if that's what's required to save a group of people".

(Also; the AI can not reliably distinguish between a group of people and a bunch of ducks and some litter or whatever.)

[nicephotog]

Not the final problem as I see it, AI vehicles have limitations to sensor operation parameters, and no ability to genuinely determine anomaly or misperceiveable misrepresentations.
However, if the article was generated AI, I suppose it's is comparative. If not, it may been a clip board or broken file join error.

[alancalverd]

And a new one! I rented an all-singing all-dancing electric MG4 last week (they didn't have any small diesels left, and nobody wants the electrics, so I got it for the price of a mini). 

Easy to drive, excellent acceleration, unimpressive fuel consumption (4 kW at 30 mph - same as any other medium car, but this one is burning fossil fuel several miles away, not at the point of use). Buried under a heap of unintelligible menus on a touch screen that doesn't like bass players' fingers, is a "speed limit sign sensor" that they don't tell you about. So I'm bowling along a fairly empty motorway at 70 mph when the idiot machine spots a temporary 30 mph sign on the adjacent slip road and starts yelling at me to slow down. Thank goodness I'd managed to switch off the speed limiter - a truck up the arse can really spoil your day. 

And the aircon was rubbish. The advantage of a local combustion engine is that you have several kilowatts of waste heat that can be pumped into the cabin on a cold day at no extra cost. Trying to keep warm in an electric car can double your fuel consumption and drown someone's island paradise.

[paul cotter]

Yep, on a cold day more of the battery capacity might go on keeping you warm rather than propelling you. The much quoted range figures assume no use of ancillary equipment.