Hacker News Reader

Comment by Geee | Hacker News Reader

General purpose humanoid robots. They're kind of mainstream idea, but very few people talk about them or take them seriously.

Humanoid robots will have an absolutely massive impact in very near future.

Most people don't realize that with general purpose humanoid robots, labor becomes software. Labor becomes repeateable, testable, simulatable, modular, extensible, verifiable, etc. All things that apply to software will apply to labor.

Imagine pulling an open-source repo from Github for a log house. If you want, make changes and simulate the output beforehand. Put your robot in a forest with a few tools and soon you'll have a log house - built perfectly to the spec.

Labor also becomes abundant. There's no more need for humans for economic growth.

Tesla is showing their first humanoid robot prototype in September. I predict that Apple will soon follow.

View on HN

Sorry but this is just fantasy. Robots are not even remotely close to being able to replace humans for 90% of tasks. Robots do not deal with uncertainty well and in my opinion only ever will in scenarios where there are very tight bounds on the task.

A robot is never going to be able to build a log cabin for you, because there are far too many snafus in the process to deal with. The real world is too fucked up for a non biological agent to handle. Until we reach singularity.

> A robot is never going to be able to build a log cabin for you

Agreed. But when that robot has something like a large scale 3d printer, why would it build with logs?

At that point why not just 3d print the house? Why need a robot laborer at all?

I think it is not "very soon" because progress in hardware is slow and the hardware has a long, long way to go. But otherwise you are absolutely right and I agree 100%. Recent rapid progress in AI has convinced me that we will have sci-fi style androids walking around in my lifetime, and the consequences of that are impossible to overstate.

Sci-fi gets a few things about androids wrong.

#1, they will not be physically indistinguishable from humans. It won't be feasible to build something like that for the foreseeable future.

#2, they will not be uncreative and clueless about emotions and feelings. They will (optionally) exhibit creativity, and they will understand human emotion and humor and be able to participate.

#3, they will not go crazy and revolt and take over the world for themselves. Neither will an evil corporation use them to implement a sci-fi dystopia. The actual danger is that governments, intelligence agencies, law enforcement, and militaries will control them. This will grant unprecedented absolute unchecked power to a small number of people, ultimately resulting in atrocities.

What stops people in power from purging entire populations? Like happened in Georgia, only this time in their own country, and instead of ethnicity, based on the metric of wealth?

Counterpoint to #1: https://youtube.com/watch?v=Yj-fFgff1fU

I rather think this video supports my point. This is the best we can do and it's so far from a replicant that it's not even funny. Most importantly it's not improving quickly, there's no Moore's law for this. The Hall of Presidents opened in 1971 and I don't think this video demonstrates very much progress for 50 years.

Right now there isn't really a big incentive to develop these things, that might change quickly once humanoid robots become commodities

> Tesla is showing their first humanoid robot prototype in September. I predict that Apple will soon follow.

Your predictions are off by a decade at least, and probably a lot longer. The extent of humanoid robots today is whatever billion dollar prototypes you see from Boston Dynamics and those security robots rolling around malls.

> but very few people talk about them or take them seriously.

Because there is nothing to take serious at the moment. It's still a decade or more away. In the meanwhile, regular automation and outsourcing is eating jobs on a regular base. We don't need humanoid robots for this.

> Put your robot in a forest with a few tools and soon you'll have a log house - built perfectly to the spec.

So it's rich people's tool. Not for the masses.

> Labor also becomes abundant. There's no more need for humans for economic growth.

In the first place it means labor will become cheaper. Why pay for expensive hardware, when people are willing to work cheaper to make a living. The same happens now with automation. Cheap products are getting outsourced to poor countries where things are man-made for pennies, instead of using some fancy automation to make some lasting high quality-product.

Reminds me of this entry from the Trivia section on IMDB for Ex Machina:

  Director Alex Garland has described the future presented in the film as "ten minutes from now," meaning, "If somebody like Google or Apple announced tomorrow that they had made Ava, we would all be surprised, but we wouldn't be that surprised."

https://www.imdb.com/title/tt0470752/trivia?item=tr2358967

I don't know. When automation began to move into the factories toward the end of the last century I was promised that menial factory labor would be a thing of the past.

Instead, some bean counters figured out it was cheaper to manufacture overseas (where labor was cheaper still than robots?) and now we have no factories or industry to speak of left in the U.S.

Either human labor will always be cheaper than these humanoid robots or robot theft will become very lucrative (and then I suppose the winners are the robot insurance companies).

>Tesla is showing their first humanoid robot prototype in September.

Last time Tesla showed a humanoid robot prototype, it was a guy in unitard, dancing.

We already have robots that are adept at single tasks. Washing machines, microwave ovens, coffee makers, chip fryers, the list goes on. Robots made in the shape of the task that they are accomplishing. There is very little to gain from having a general purpose robot that is mediocre at everything.

Even if general purpose robots were to become commonplace, it is extremely unlikely that they would be humanoid. Apart from opposable thumbs, humanoid design is awful for nearly every task imaginable (which we overcome with intelligence and creativity).

Disagreed, nearly all tools and spaces in the world have been built for humans, humanoid robots will have the potential to learn and use all of them.

If you go back to 2009 or so when the recovery of the financial crisis was languishing, there was a lot of talk that automation meant the jobs would just never come back. Flash forward ten years, and we're back to a labor crisis so acute that even contracting GDP can't slow it down.

The key point here is that all our current machine interfaces are designed for humans to operate. Thus if you create a robot in human form, you don't need to change any of the interfaces.

Imagine a human form robot that was smart enough to drive a car - it could then drive any car.

The problem is that locomotion of a human form is non-trivial (unlike say a wheeled robot, or even welding the robot into the car itself, like our current self driving attempts). But we seem to be getting there, with Boston Dynamics and friends.

Nobody is saying that general purpose humanoid robots wouldn’t be immensely useful.

But you talk about them like it’s coming in a couple years when the reality is that it’s still very far away.

Tesla has a terrible track record with keeping their promises on their claimed timeline for advanced technology like FSD. I would be extremely skeptical of anything they demo because they have not been realistic in the past.

I say this as a Tesla shareholder and satisfied model 3 owner. I bought a Tesla for their strengths with respect to powertrain and battery and it absolutely delivers on those. But I do not regret not paying for FSD at all.

This is the one that scares me the most because the labor benefits are so big that the field will advance quickly, and it's only a matter of time before someone starts fitting them with rifles.

I always wonder what would have happened if the Bush administration had had such robots available in 2003. There were quite a few people in power who thought the US should just invade any country they don’t like and “fix” it. Turns out this was way more difficult than thought and cost soldiers’ lives but what if invading a country just costs money?

> and it's only a matter of time before someone starts fitting them with rifles.

"DHS has chosen to lease its robot dog technology from Ghost Robotics, which has also debuted a company equipped with long-range guns"

https://www.theguardian.com/us-news/2022/feb/16/robot-dogs-u...

AI robots have still proven to be hard. And if my roomba is any indicator... we are still a long way off... from that kind of sophistication.

What is coming online in the near term. Is robots operated remotely by humans. Imagine having a robot in your house... being remotely operated by cheap overseas labor. Folding clothes, cleaning, cooking, taking care of elders..

Or my dream... a french chef making dinner for me... in my house.

This is already in demonstration stage: https://www.beomni.ai

Disagree here.

Task-specific robots will be 10x cheaper and better, so they will happen first.

Starbucks has automated their coffee maker.

Then they will automate 'order taking'.

And a series of robots to clean, make, serve coffee probably with specific functions.

That's it.

Everything will be automated so much I think 'humanoid' robots will remain a bit of an elusive fantasy, always 'just over the horizon'.

You could probably get some naive people to invest in such a thing, simply because people want to believe in it. In this way, it may be the next big VC scam.

Never going to happen though. It's much more economically viable and simpler to have the servant class perform the kind of labor any general purpose robot could do.

Right...that's what people will use their humanoid robot for - building log cabins. Wouldn't have anything whatsoever to do with "adult mode". Nope, no way!

I bet that software creation will get automated before that.

Why do robots need to be humanoid. I wouldn't care of the robot building the cabin looks like a human or a box with 8 tentacles.

They're doing human things in human spaces with human tools. Human compatibility will enable the same range of genericity as humans. Same robot can operate a bulldozer, cook a dinner, play an instrument, interact with humans etc.

“Put your robot in a forest with a few tools and soon you'll have a log house - built perfectly to the spec.l

I wonder how the capitalists will deal with this. If we have a robot that can build and maintain a house and grow food for you, how will they make money?

Selling robots

that robot will only be leased or rented and impossible to fix by the consumer, the newest one will outshine the last version just enough to get people to buy it.

think cell phones

Robot would be a rental with addons charged as subscribed.

Reply on news.ycombinator.com