News for hackers. Stuff that matters.
Re:Physical Intelligence's first generalist policy AI can fi... by NoLinkToMe on Sunday November 10, 2024 @03:12AM (#42099052)
Agreed, my girlfriend has a biochemistry degree and works in such a lab. It involves some physical work like collecting, preparing and inputting samples, doing basic maintenance of the machines, some analysis and then some administrative/secretarial work in feeding the results into a system or by phonecall to the doctor.

All of which can indeed be automated in my view.

I would say it really depends on where you are though. In the US it probably makes sense quite quickly. But she lives in a small EU country where salaries aren't high, and this is very much a junior position with a lot of students looking for such a job. Her position costs about 25k a year.

The NPV of a $1m investment with a 25k cashflow is negative at normal discount rates. Once you get to replace a $120k salary with a $1m robot, it does make sense.

Further I do still expect there to be some jobs in overseeing the robot (e.g. your average factory manager). That makes sense for large centralised production locations, because you can have 1 human job overseeing many robots. It doesn't necessarily make sense with many small decentralised production locations. And that's the nature of most labs, I believe due to the time sensitivity of a lot of lab work they need to be everywhere and close to customers. But maybe that will change.

Re:Physical Intelligence's first generalist policy AI can fi... by llm_trw on Sunday November 10, 2024 @03:27AM (#42099088)
Keep in mind that workers in the EU work for around 6 hours a day, a robot works for 24, so it's replacing 4 of them, not 1.
Re:Physical Intelligence's first generalist policy AI can fi... by kaplun on Sunday November 10, 2024 @04:10AM (#42099213)
6? Not 8?
Re:Physical Intelligence's first generalist policy AI can fi... by llm_trw on Sunday November 10, 2024 @04:23AM (#42099260)
Lunch and breakfast are serious business.
Re:Physical Intelligence's first generalist policy AI can fi... by muspimerol on Sunday November 10, 2024 @04:32AM (#42099288)
Huh? The average working week in the EU is 37.5 hours a week, or 7.5 hours a day: https://ec.europa.eu/eurostat/web/products-eurostat-news/w/d...

Breakfast and lunch are not paid breaks in any EU country as far as I know.

Re:Physical Intelligence's first generalist policy AI can fi... by Ekaros on Sunday November 10, 2024 @06:24AM (#42099636)
It really depends. For Finland lunch is unpaid unless you are forced to eat at premises. But then you usually have 2 refreshment breaks (10-15min) which are paid...
Re:Physical Intelligence's first generalist policy AI can fi... by Gare on Sunday November 10, 2024 @04:50AM (#42099341)
> Breakfast and lunch are not paid breaks in any EU country as far as I know.

One 30 minute break (lunch break) is paid in Croatia and Slovenia

Re:Physical Intelligence's first generalist policy AI can fi... by bottom999mottob on Sunday November 10, 2024 @04:24AM (#42099262)
Assuming a lunch break is not included as part of 8 hours, do you really think most exmployees are productive for a full 8 hours?

Most office workers spend an exorbitant amount of time twiddling their thumbs, reading emails, going to the restroom. On average, I'd say knowledge workers are typically productive for about 2 to 3 hours per day. I'd estimate physical laborers range up to 4 or 5 hours.

Assuming 8 hours of full productivity is a strange number to focus on

Re:Physical Intelligence's first generalist policy AI can fi... by Krssst on Sunday November 10, 2024 @10:52AM (#42100785)
I was surprised too but I guess OP included the weekends too since robots don't need those off. So 40 hours divided by 7 days.
Re:Physical Intelligence's first generalist policy AI can fi... by NoLinkToMe on Sunday November 10, 2024 @05:33AM (#42099475)
Good point, although it probably won't be 4 for most labs, perhaps 2. My gf does 1 nightshift a week and she is alone in the lab for the 3-4 emergency samples that come in (versus a team of people processing hundreds during the day), there simply isn't much demand at night. Most blood sampling has a human pipeline of people going to their doctor or hospital with an issue, getting blood taken, which is sent to a lab, the majority of that during the day.

Again for large scale centralised labs that get sent samples throughout the 24 hour period from all across the world, where timing isn't an big issue, this could work. But most labs are small, close to the customer, time-sensitive, and work with couriers to bring samples because samples need to be transported and stored in specific ways not to go bad.

She works quite long shifts btw (10-12h), but that's more a function of her country/company culture rather than the norm. Probably 6-8 is more common indeed.

Re:Physical Intelligence's first generalist policy AI can fi... by vidarh on Sunday November 10, 2024 @06:29AM (#42099652)
How much of the timing on samples is down to lab availability, though? E.g. my gp wants samples in by 1pm because otherwise they'll got to the lab the following day. If the lab would still process things that came in at the end of the day, I could very much see larger doctors offices sending off samples more than once a day, with the last one coming in after the end of their clinics.

I could see demand spreading out more - though I agree it might be closer to 2 than 4 - if the availability of human labor wasn't an issue.

Re:Physical Intelligence's first generalist policy AI can fi... by bagels on Sunday November 10, 2024 @03:33AM (#42099099)
Production version of these robot arms would not be anywhere near $1m
Re:Physical Intelligence's first generalist policy AI can fi... by Falimonda on Sunday November 10, 2024 @03:49AM (#42099145)
That seems to only account for the hardware. How much do you reckon a software license will cost for your fleet of cheap arms?
Re:Physical Intelligence's first generalist policy AI can fi... by thanksgiving on Sunday November 10, 2024 @04:44AM (#42099324)
I got really excited for a second for people who need to work on a lab thinking they will be able to work from home now...
Re:Physical Intelligence's first generalist policy AI can fi... by csomar on Sunday November 10, 2024 @05:00AM (#42099378)
> That seems to only account for the hardware. How much do you reckon a software license will cost for your fleet of cheap arms?

Just a bit lower than what a human would cost.

Re:Physical Intelligence's first generalist policy AI can fi... by amelius on Sunday November 10, 2024 @06:27AM (#42099645)
Remember though that AI is a winner takes all space. It will probably be won by a company with lots of cash behind it (like this one which has support from many companies, including OpenAI).
Re:Physical Intelligence's first generalist policy AI can fi... by notatoad on Sunday November 10, 2024 @12:13PM (#42101281)
significantly lower than what a human already doing the job would cost, until the VC money runs out. And then a whole bunch more than what a human would cost, after they've got you locked in and you've let your staff go.
Re:Physical Intelligence's first generalist policy AI can fi... by ValentinA23 on Sunday November 10, 2024 @08:22AM (#42100068)
It wouldn't cost 1M. The BOM[1] of Stanford's Mobile Aloha[2] (very similar to physical intelligence's robot) is around $30,000.

[1]https://docs.google.com/document/d/1_3yhWjodSNNYlpxkRCPIlvIA...

[2]https://mobile-aloha.github.io/

$7000 for the wheels/navigation module $3500 x2 for the control arms $6000 x2 for the controlled arms

I don't think you need a $7000 navigation module and the arms are also probably overkill. To which extent do we need closed-loop Dynamixel servos when the policy diffusion algo already relies on computer vision driven closed-loop control ?