Robocrop: world's first raspberry-picking robot set to work | Hacker News Reader

Robocrop: world's first raspberry-picking robot set to work

theguardian.com

96 pointsfinphil98 comments

I had a summer job picking raspberries on the Isle of Man. It was hard, boring, work. I left after only a few weeks, and got my first programming job (the island had a thriving offshore finance industry and no Manx coders).

So I'm ambivalent about this. On the one hand, awesome that yet another crappy job is being automated. On the other hand, that job literally saved my life - I was homeless and struggling until that opportunity came along, and it lead on to a career that has kept me secure ever since.

How do we replace the crappy low-skilled, low-paying jobs that we're automating, but that also allow people to stay alive while they sort themselves out?

redistribution of wealth.

What about doing it without aggression?

Redistributing wealth isn't any more inherently aggressive than any other change. If people object violently to being asked to share the bounties of our industrial economy, that's no different from if they objected violently to being asked to put up with their sister marrying a woman, or to stop smoking in the restaurant.

Isn't it the people insisting on taking a share who are the ones employing violence? If the people being asked to share say no then armed men show up.

I mean, no armed men come take your taxes (anymore). The threat is ultimately there if you withhold otherwise but it's the same if you smoke in a mcdondalds then refuse to leave. Or if you squat on someone else's land.

If you're sent to prison if you continue to not pay taxes, then yes, the system is still using force.

Taxes aren't aggression

What happens if you don't pay? The consequences are aggression, and the fact that threat is hanging over you makes taxes a form of aggression. Coercing is a form of aggression.

This is true for any enforcement of laws though?

That statement is precisely one of the tenets of anarchy. I think the philosophy as a whole has issues, but it's hard to argue with that premise.

I wasn't aware of that; I don't consider myself an anarchist really. I just think it's more like what Heinlein wrote, that violence is the supreme authority from which all other authorities are derived. I'm not sure I like it but I do think it's rather true. Almost everything can be traced back to some form of this.

Note the 'almost'. I'm sure some outliers exist but that's not what we're talking about.

Letting someone starve is aggression. If you have to work for minimum wage because the alternative is homelessness, isn't that also a form of coercion?

> What happens if you don't pay?

You get charged interest. Then, you might have your wages garnished. If that isn't sufficient, your bank accounts might get frozen. If that still isn't sufficient, liens.

Uber shows that it's possible to redistribute wealth from venture capitalists to customers without aggression.

Kalanick, the modern Lenin.

Every time we try that, the people who have concrentrated all the wealth say, "No thanks" and keep it all.

Would one way to redistribute wealth to ensure that everybody partakes in the wealth that is created? So you're always earning a stake in what you're doing.

In practice this would mean that all jobs would pay RSUs of some kind. If you rent then you are getting some of the equity of the house as well, so that when it is sold you benefit and not just the landlord.

The idea is basically that everyone should benefit from capitalism, not the just the lucky few with all the capital.

It isn't an uncommon thought. Unfortunately this isn't very popular, partly because it is actual socialism (which is even less popular than social systems). In general I think it is also too late to introduce such a system. You have to get ahead of the curve before people realize what they have got, sort of like in Norway with the oil fund.

Socialism is redistribution of wealth by the state. This is redistribution of wealth by capitalism?

Sounds more like a cooperative than socialism to me, which has been very successful in the past

https://en.wikipedia.org/wiki/Cooperative

farmers struggle to get people to take on those jobs in the UK. This is a talking point about Brexit. Farmers need citizens from Eastern Europe to do the work, UK politicians claim they should employ locals, locals refuse to do the jobs, citizens claim they "should just pay more", farmers claim their margins don't allow for greater wages for the work.

I believe we lost a bunch of last years crop to this issue and maybe this years. The drop in the pound since 2016 hasn't helped either.

Maybe we shouldn't be growing the stuff if the economics don't permit it. This is where the robots may help.

This job is actually very hard work, and would be called "skilled" if we allowed the use of that term for skills outside formal education. You need fitness and endurance, and skill to do it quickly. Trying to conscript random unemployed people to it doesn't work nearly as well. See the point about productivity going down under "austerity".

https://www.thegrocer.co.uk/brexit-and-the-workforce/why-are... : pretty sensible article on the subject, pointing out ways in which employers could improve their local hiring.

It's not skilled work. anyone who can work can pick raspberries.

The skill you are discussing is to pick them fast enough for profitability. Thats true of any job.

> It's not skilled work. anyone who can work can pick raspberries.

And yet, that is literally not the case.

Resisting boredom and maintain focus in grueling conditions are a skill.

Can you imagine going down to the river with a washboard and your laundry? No way we have robot washing machines. It would be ridiculous to wash our clothes in the river. Humans will have the same sentiment about many jobs that robots eventually take over. Why stand in front of hot grease cooking fries when a robot can do it? Simply so we can transfer some wealth from one person to another for doing something mindless? But also who will buy the fries if no one has jobs to make any money? Perhaps we will see deflation and the cost of things will be less and perhaps with less work available we can all strive for a 4 hour day. But ultimately I feel we are slowly approaching a post scarcity society. We need to change our views on capitalism if you ask me. Yes people deserve to make money but I think we can shift some services always from always trying to make a buck off someone.

>slowly approaching a post scarcity society.

Two things. Neither of these is meant to be hateful.

1. Which is more likely: That society as a whole will evolve into a benevolent post-scarcity society providing basic necessities for everyone in it, or that our new-age feudal lord-owners of everything will use automation to drive costs down far enough so that everyone can juuuuust afford to live?

2. Doesn't post-scarcity require energy inputs from a non-depleting resource?

I live in Europe so I expect the state to look after me in old age, when I'm ill, unemployed, etc. I'm not sure its such a leap to extend that further. Now I'm not sure if in the current political climate its the most likely, but certainly possible.

To answer your point 2. Solar and wind should keep us going for a few millenia?

I don't see the system changing without a war. It's not just already old people that adopt conservative or capitalism, it's people that get older. For every mega billionaire pledging to give away half their wealth, there are hundreds of thousands of millionaires that will try to prevent change and lock the system in with as much power as they can wield.

I have a feeling the future is more likely going to be a Bruce Willis or Will Smith movie.

How do we replace the crappy low-skilled, low-paying jobs that we're automating, but that also allow people to stay alive while they sort themselves out?

Take education more seriously and make it more accessible. I don't know how to do that, but I feel it needs doin'.

there's a lot of very educated people doing those crappy, low-skilled, low-paying jobs

You implement a Job Guarantee. The ability to sell your labour to the central bank at the living wage, who then allocates it to local democracy to use as local democracy sees fit working for the public good.

After all the 'interest rate targeting' function of the central bank, and the unelected wonks that run it, is supposed to guarantee full employment. If it isn't doing so, then those who screwed up should obviously pay the price of their failure.

That it acts as a powerful targeted automatic stabiliser injecting liquidity temporally and spatially on demand into local economies in a manner bank lending could never hope to match is just an added bonus.

For anyone unfamiliar, a job guarantee is a proposal of Modern Monetary Theory. Although the validity of MMT has neither been established nor disproven explicitly, it relies on, among other things, the ability of a nation to print its own currency, collect taxes in its own currency, and issue debt exclusively in its own currency. It is therefore not applicable in all nations.

> The ability to sell your labour to the central bank at the living wage, who then allocates it to local democracy to use as local democracy sees fit working for the public good.

How would a central bank determine allocations to various municipalities? This step necessarily requires discretion on the part of the central bank. Such a policy would distort the labor market, crowding out private sector employment and potentially producing malinvestment. It would also increase the power of the central bank, which is not necessarily desirable if the unelected nature of central bankers is a concern.

> the 'interest rate targeting' function of the central bank... is supposed to guarantee full employment

The Federal Reserve bank has a dual mandate of stable, low inflation and low unemployment. Central banks in many other countries are mandated only to maintain stable, low inflation.

> That it acts as a powerful targeted automatic stabiliser

There are other automatic stabilizers that can be implemented independent of the central bank. In the United States, for example, tax incentives for employer provided health insurance have a procyclical effect, as decreased employment during recession causing reduced healthcare expenditures.

> bank lending could never hope to match

Global lending has been distorted since 2008 due to the Federal Reserve's interest on excess reserves policy. The combination of increased reserve requirements, increased bank reserves due to the fed's enlarged balance sheet, and the availability of a risk free return on reserves has reduced the overall quantity of loans relative to prior trends. Even small, local banks, who would otherwise like to make loans and are subject to less stringent reserve requirements, have increased the reserves on hand due to the breakdown of the federal funds market caused by IOER. It's hard to definitively state that bank lending is less effective than other policy when it's been hampered by central bank policy for a decade.

"It is therefore not applicable in all nations."

It's applicable to all nations. You could have a currency that is specifically for this job. As long as you remember to tax it out of existence at the other end.

"This step necessarily requires discretion on the part of the central bank. "

The process in the UK is straightforward. It follows the existing Universal Credit system. The indidividual presents themselves at the local Job Centre, the Job Centre signs them up and allocates them as local democracy dictates to jobs in the area (whether working for local councils, or local charities as the municipality dictates). The job program signs off timesheets which are filed with the UC programme, and then the UC is paid - directly by the central bank. Rather than a 'Treasury' account, you use a 'Job Guarantee' account to charge the bill to. Other than than the process is the same.

"The Federal Reserve bank has a dual mandate of stable, low inflation and low unemployment. "

Which it is failing to keep. Central banks without employment targets have high unemployment - particularly if government is constrained.

"There are other automatic stabilizers that can be implemented independent of the central bank."

Not and allow a balanced budget provision at the same time.

"Global lending has been distorted since 2008 due to the Federal Reserve's interest on excess reserves policy"

Without that central interest rates would be zero. The natural rate of interest is zero in the interbank - as any banker will tell you. Central banks interfere in the market to raise rates.

Reserves have no control function over bank lending. At all. Ever.

And anyway that misses the point. Bank lending takes time, which as any student of dynamic systems knows introduces unstable harmonics into the response function. The reaction is precisely the same as trying to stabilise a skidding car with a steering wheel. Get too big a skid and you can't counter it because you can't respond fast enough. Then the car crashes.

It sounds nice. But isn't this literally the kind of top-down centrally-planned economy that was implemented in many post-WW2 socialist states? [0]

[0] https://en.wikipedia.org/wiki/Socialist_economics

What happens if the local democracy determines that your labor is not worth the living wage?

What if it really is worth the living wage, but the Allocator wants to cheat you?

I recommend this (well fundamented as usual) post from SSC:

'Basic Income, Not Basic Jobs: Against Hijacking Utopia.'

https://slatestarcodex.com/2018/05/16/basic-income-not-basic...

I think the main argument should be fundamented by charitable assumption, that most people actually have decent plans for employability and could use the available time much better than being stuck on a Job Guarantee (that could have very low marginal value to the employer anyway).

The problem with that is that other people won't accept it - because you are being paid twice.

Once with the output of other people who are using up their finite lives working for output from you, and once self consuming your own time.

Unsurprisingly those working resent this, as they'd rather have Friday off than work an extra day to produce output for people who are producing nothing for them in return.

The workers then agitate to have the income removed from those not contributing. Which is why every basic income 'experiment' in history has failed.

Once you look at it in real terms, the fallacy of composition shows itself. You can't expect others to give up hours of their lives for you if you don't demonstrably give up hours of your life for them in return.

Hence why you have a Job Guarantee (and frankly most of the private sector nonsense jobs out there). It is a public demonstration to those working the machines, that actually produce the output we need, that you are trying to give up hours in solidarity with them.

Then they won't feel they are being ripped off by working past Tuesday.

"that could have very low marginal value to the employer anyway"

It's the job of the employer to extract value. If they can't then they don't get paid. Capitalists only get paid if they transform labour hours into labour services effectively. If they can't, they go bust.

However the labour hours still need to be sold, because that is what keeps the labourer alive. You must give up hours of your life, if you expect others to do the same for you.

If value cannot be extracted, then it gets transferred from capitalists who have failed to do their job properly.

This is, I think, one of the most profound questions of this century. We will have to decide, as a society, what to do about the fact that robots are cutting off the bottom couple of rungs on the ladder and yet there are still humans down there who need to climb up.

I think the way you frame the question points to one answer: ensure that people are able to stay alive even if unemployed. We have enough wealth, at least in first-world countries, to give everyone a basic level of housing, food, and healthcare. We could engineer society such that money isn't required to attain the first two levels of Maslow's hierarchy of needs.

This doesn't necessarily mean some hypothesized feared Communist/Socialist welfare state where everyone lives on the dole and no one works. Working would still be required for a good, fulfilling life. But the safety net would catch you before you get so destitute that you die.

"We have enough wealth"

The current system has wealth. If we changed to that system, all definitions would be less of what they are now, so you'd have to bake in the idea that more people are now building cheap housing in random places, and that grocery stores are given food non-stop to feed those people, and that doctors will want to live near them.

It doesn't account for the droves of people that may stop working because we don't _think_ like the people that work for Chipotle that no longer want to do so. We're all fucking programmers (well a lot of us are).

How long until we can automate the crappiness out of coding for a company?

Be careful what you wish for...

>> How do we replace the crappy low-skilled, low-paying jobs that we're automating, but that also allow people to stay alive while they sort themselves out?

Universal basic income?

Seems like anything else is just a roundabout way with additional layers of beaurocracy of doing universal basic income.

I hate to be a downer but ill believe it when I see it. That robot is big, bulky, and does not look nimble or fast. A human could have picked a couple boxes worth of berries in the time it took that robot to decide to pick one berry. Anybody who has ever picked berries of any kind, be it strawberries or blueberries or blackberries will tell you, speed is the name of the game and even the average human versus an experienced picker isn't even a competition. An experienced picker is going to be grabbing multiple berries at a time with all of their fingers, not just one at a time. An experienced picker will have decided which berries to pick next before they even got done putting the ones they just picked into a basket which only takes a fraction of a second.

You will get far better returns breeding and genetically modifying the plants to be easier to pick or shake the berries loose than any robot ive ever seen. Sure, it is technically feasible, but nobody is going to spend a million dollars on a robot that doesn't pick faster than an experienced human, it would take multiple decades to pay itself off even if you assumed the power and maintenance was free.

Maybe in another 10 years it will be worth seriously looking into berry picking robots again, but berries are literally the hardest and most difficult fruit for a robot to pick. Uneven ripening, very delicate fruits, tangled messes of plant vines, very little maneuvering room, ect. It is like trying to automate a rally car before you are able to automate an oval track race.

They did start with the hardest berries so that it's easier to work backwards for easier fruit. Also the plan is to hire them out to farmers at a price per kilo picked,which is aimed at less than the cost of a human.

I genuinely expected Raspberry Pi inside.

I'm sure one of the end products will include pie.

Hearing my dad talk in the past about machine designs to automate both tomato and strawberry machines. He recommends cutting the stalk with either a laser or a scissor. This is significantly faster than trying to pick a vulnerable fruit.

Laser might cause a fire which is why they don't use it. I think best method is to simply mimic the human hand motion. You grab the berry but you put your two fingers over the stem and then pull down on the berry until the stem disconnects. This method might be even superior to the scissors method since you won't need to replace the scissors due to dullness.

Yes, my thoughts exactly. I'd imagine a sharp razors and a tube that sucks air (or a net) would be much cheaper than a robot that simulates a human hand to pick tomatoes.

I wonder if someone experiment with some kind of scaffolding that would make fruits more exposed and easer to pick.

If you do that with raspberries then you need to have people process the fruit again as they're sold without stems. Tomatoes and strawberries are sold with stems as they don't detach without cutting. Don't know what the effect would be, commercially and in terms of preservation, of selling raspberries with stems.

I think if robotically trimmed fruits with stems sold for 10% cheaper than slow picked stemless fruits, there'd be no market left for stemless by the time the season is over.

If selling with stems works fine, no doubt. But raspberries detach from stems as they mature so you may end up with a messy box after a while. Or maybe this is just done because it's easy to pick manually by grabbing the fruit and picking with stems works as well. I'll have to try picking some with stems and see what happens.

>raspberries detach from stems as they mature

If the problem can be reduced to separating objects of different characteristics from a process flow, existing food processing techniques like vibrating belts and air guns can handle it, no need for robot arms.

Ripe raspberries are very delicate. Some of these small fruits are sometimes picked directly into the final retail boxes to minimize handling. I don't know if you can apply those methods if you're not then processing the fruit further into jams or juices.

Shipping them like that would probably make it more expensive, lots of wasted space and added weight.

I think this is a much better idea, because it decomposes the problem into easier things for robots. People are talking about stems, but hold up! You can just make a second robot that picks up raspberries from a bucket and removes the stems. This sounds much easier to me.

This would only work for autumn fruiting varieties. Summer fruiting varieties fruit on 1 year old stems, so unless you identified the new stems and left them...

Imagining if robots can pick raspberries at leak ripeness, in the near future one machine can do the same for all crops, plus weeding. This seems more economically viable than the current monoculture and chemical based approaches.

And if one field can support a wider array of crops, they can produce more for local consumption, and lower transportation distances and costs.

The future is already here: https://farm.bot/

This kind of robot would take million gallons of water to build a piece.

You can not mix a bunch of random crops in one field. This is such a HN comment.

Well you can, permaculture advocates do that but it's not viable for industrial agriculture yet, if ever.

Of course, by random I mean specific crops chosen for their compability to each other...

[0] https://deepgreenpermaculture.com/permaculture/permaculture-... (I really like that image of corns, beans, and squash together from traditional Mesoamerican farming)

You absolutely can, they just can't be serviced with tractors or combines. Huge monoculture fields are a recent invention, and an unsustainable one.

Replace "random" with "local" and "seasonal." Biodiversity and crop rotation are not whimsical HN topics. They have been around for a while.

It works in my backyard garden and neighborhood urban CSA. The CSA could increase their production significantly if/when technology like this matures. Most of my local, seasonal produce needs can already be supported within a few block radius of my home.

I too appreciate the diversity of thought on HN... lots of experts to share their knowledge.

I dunno, seems like a good way of hedging ;)

https://en.wikipedia.org/wiki/Multiple_cropping

https://en.wikipedia.org/wiki/Polyculture

https://en.wikipedia.org/wiki/Nurse_crop

https://en.wikipedia.org/wiki/Companion_planting

Why not?

Theres no reason why you couldn't do a 4 field crop rotation in one field.

Market gardens quite often have rows of different crops right next to each other.

If you're talking properly mixed up, 3 sisters planting did OK for Native North Americans. And many gardeners will sowing a row of radishes in with your parsnip rows because they germinate so slowly.

And that's without getting into forest gardening and permaculture.

> "Analysts attribute this lack of economic efficiency to a shift towards more low-skilled jobs since the financial crisis, a lack of business investment and a decade of austerity."

Stopping austerity would help then?

"worlds first" seems to be a misnomer here. There are machines for tractors that can pick raspberries. The machine works by shaking the entire raspberry bush and collecting the berries that fall.

https://www.youtube.com/watch?v=dOG85YtZhII

I suspect that damages a very large proportion of fruit, so either yield ends up being terrifically low, or its only suitable for jam.

I suppose you could be pedantic and point out it isn't actually picking.

The whole time I expected it to just smash the berry. This would fit into the “one day this will be neat” vibe.

dynamixel servos are a ton of fun to play with, if you ever have the chance.

(the robot featured seems to be using MX-64T's on the wrist)

No Raspberry Pi inside?

> The world's first raspberry-picking robot

The majority of raspberries are harvested by machine and have been for decades. Handpicking is only necessary for high quality fresh berries.

3D printed parts and dynamixel servos do not inspire confidence.

It picks ~700 per day at most, not counting acquisition and repositioning time for finding a new fruit. Article is intentionally misleading, video is a waste of time.

For the cost and maintenance of the machine over a year, you could probably exploit backpackers for--in my expert estimation--approximately one billion years.

I'll believe it when I see a video of the “final version”.

Until then it's just like any other flaky and slow robot grasper demos out there trained using various flaky deep reinforcement learning algorithms.

As mentioned, the "25k/day" is for a future model in 2020. According to a linked video in the article, a human does about 10k/day.

I believe that right now we're at the very start of this. Robots were supposed to come for all our manual labor jobs years ago, but they never did, because if it's not a pure assembly line environment, robots can't cope with the uncertainty. Research in robotics is increasing and getting places fast, largely due to advances in reinforcement learning.

I give it another 2 or 3 years until we have a robot hand with a camera that acts as dexterous and capable as a human hand. Once we have that, tons of opportunities like this raspberry picking are open for automation.

"I give it another 2 or 3 years until we have a robot hand with a camera that acts as dexterous and capable as a human hand."

-some guy, circa 1975

The original 1958 AI proposal was expecting to crack it over a summer hackathon.

Also, that's 25k/24h for the robot vs 10k/8h for a human.

Taking into account that the time of day it's picked has a dramatic effect on a raspberry's taste, these robots may only be useful with artificially illuminated crops... or to pick those "raspberry shaped" flavorless things.

Same goes for grapes and several other fruits.

I don't think that its that robots can't do these jobs but more that underpaid illegal migrants can pick strawberrys for cheaper than a robot setup and maintaince can.

25,000 a day is one berry picked every 3.46 seconds, including overhead to switch plants.

That sounds amazing and impossible, for something that needs to move a gripper around the stems and leaves of an actual ("unguided", natural) plant. But I'm no process automation engineer or roboticist, so I'm just looking forward to the amazement. :)

Think: multiple grippers. With two grippers on opposite sides on a robot then it could harvest both sides of a row of plants at once.

Or four grippers: two on each side for different heights.

FTA: The final robot version, expected to go into production next year, will have four grippers, all picking simultaneously.

>Robots were supposed to come for all our manual labor jobs years ago

It depends, on one hand we have people trying to make Robot act like human, and could easily adopt to many of our job, I think we are still at least another 10 - 15 years away from that. But Robotic Automation for a few specific job is much much closer ( if not already here ) , I don't think it really ever was a "technical" challenges, it is the cost of these development has come down, from manufacturing them in China to Software development tools that make them work. The cost of the final product vs the possible economical advantage has finally reached a point where many of these make sense.

More often, if not in all cases, it is always the economical side of things that makes the call, not the technical side of it.

The picking rate is meaningless because you can always increase it by buying more robots. In this case, they plan to have 4 slow grippers to achieve that 25k/day.

The performance needs to be given as a ratio with some other important variable like cost so it can meaningfully be compared to humans or other robots.

blogspam for https://www.theguardian.com/technology/2019/may/26/world-fir...

Also the title is bombastic, 25k/day is what they expect to be able to build by 2020

FYI: The Guardian article is cited as a source.

Yes. This is an academic prototype. And it's really, really slow.[1]

There are much better robotic fruit pickers.[2][3]. And the commercial ones that do this entirely by mechanical means.[4]

[1] https://www.youtube.com/watch?v=KFemWAXx-3I

[2] https://www.youtube.com/watch?v=UaL3UxUclKY

[3] https://www.youtube.com/watch?v=M3SGScaShhw

[4] https://www.youtube.com/watch?v=bt73GOk4JRY

1 = Raspberries outdoor. Camera and two-finger gripper + pull.

2 = Apples outdoor. Camera and three-finger gripper + twist + pull

3 = Strawberries indoor/outdoor. Camera and two-finger gripper + pull.

4 = Blueberries + raspberries outdoor. Multiple general agitator designs + fall. (see https://youtu.be/bt73GOk4JRY?t=100)

I think the fourth is the only honestly commercially viable option. However, it's never going to work with most fruit, because it relies upon general agitation to disconnect the fruit.

Additionally, all of the solutions require neatly cropped rows. Whereas, the promise of robotics is that it will allow intercropping which creates more productive land with a stronger and more natural ecosystem, reducing outlays on pesticide, herbicide, fertilizer and related labour and increasing effective crop density and land use. None of these solutions even touch on these very real opportunities.

FWIW, I am personally confident that my R&D team (who normally work on food production and packaging robotics) could produce a functionally superior prototype to the OP video in a matter of weeks, for a fraction of the budget.

So much for the HNer who confidently claimed that robots wouldn't be doing this kind of work any time soon, and used that as the basis for flaming me.

With the current state of deep learning, this kind of application is so easy to make, it could be a science fair project.

We used to call this "vaporware".

Certainly with my own robots I can pick raspberrys much faster and better than those guys can currently do.

This thing is extremely slow. I am wondering what kind of engineering those people have used for it to be so slow. Probably they are using java or python or mathlab code, instead of using things like C, shaders and FPGA code, expecting the creation of this low level code to be the "easy" part(in the future of course).

From my experience, your greatest advantage is the ability to master those low level technologies. It has to be designed around this technologies and not as an after thought like this project looks.

How can I reach you? Your design philosophy is very interesting!

As someone working in automation, I don't think the real bottleneck here would be python. It seems more like it's waiting for accurate sensor input and trying to properly position itself. Slight mechanical imperfections with a delicate fruit like a raspberry means disaster, while a machine like this could probably recognize and tug apples off a tree without any delay.

Reply on news.ycombinator.com