Will SpaceX using technology from Open AI

[cro-magnon gramps]

I just watched a video from August of last year where Open AI was matched against a game player and won. Then the video went into a discussion of the Alphago defeat of the GO player. It got me thinking about whether or not SpaceX (or any of Musk's companies) was using this technology. Even in it's most rudimentary or primitive roll out, it would be a very powerful tool...

This is more a place holder for future discussions about this topic, as relates to the developments in the near future.
perhaps I should have made this an update thread, Chris or Lar you may want to consider that... I have realized in the last 2 months just how imminent Primitive AI is, in real world applications and think it is important we begin to think about it...

[mlindner]

No. SpaceX's problems are well constrained and don't require AI to solve them. AI is reserved for problems that are otherwise impossible to solve without AI. SpaceX's problems are mostly control system problems (automated vehicle landing) that AI is poorly suited towards.

[Roy_H]

No. SpaceX's problems are well constrained and don't require AI to solve them. AI is reserved for problems that are otherwise impossible to solve without AI. SpaceX's problems are mostly control system problems (automated vehicle landing) that AI is poorly suited towards.

I wouldn't be so sure about that. AI has been around for a long time now. I remember reading many years ago about NASA using AI to design a new antenna for one of their outer planet space probes. It was far removed from anything I have seen, looked like a bent coat hanger, one single rod with about 3 fairly sharp bends more or less pointed at earth. This was apparently superior to anything else from a performance, mass, size viewpoint.

[mlindner]

No. SpaceX's problems are well constrained and don't require AI to solve them. AI is reserved for problems that are otherwise impossible to solve without AI. SpaceX's problems are mostly control system problems (automated vehicle landing) that AI is poorly suited towards.

I wouldn't be so sure about that. AI has been around for a long time now. I remember reading many years ago about NASA using AI to design a new antenna for one of their outer planet space probes. It was far removed from anything I have seen, looked like a bent coat hanger, one single rod with about 3 fairly sharp bends more or less pointed at earth. This was apparently superior to anything else from a performance, mass, size viewpoint.

Yes, that's using genetic algorithms to solve for an antenna design problem. However these are quite slow and you aren't guaranteed to find the best solution or even a good solution. Good description of the issues with them: https://www.quora.com/What-are-the-disadvantage-of-genetic-algorithm

I haven't read NASA's paper but I expect they ran many many simulation runs until they found one that happened to make an antenna that was better than what humans had designed.

There's one other telling sign. If it was so much better, why aren't they using it? Those experiments were over a decade ago. There were likely drawbacks not included in the news reporting about the antenna.

[su27k]

Just to put my words in before this gets locked by mods (a similar thread in Advanced Concept already disappeared into the ether): OpenAI is mostly pure research, they don't focus on practical applications (except maybe AI safety). Their results are built on others in the field, and they publish their research in papers so that others can built on it. So there wouldn't exactly be "technology from Open AI", a better question is whether SpaceX will use machine learning technology (Deep Learning and Reinforcement Learning in particular) in their work.

The game player AI is based on Reinforcement Learning, its practical application for SpaceX may be robotics that can learn by following human demonstration, instead of being programmed, although this application is still in its infancy. More mature machine learning technology would be Deep Learning based computer vision, which should be able to replace anything that requires human vision.

[JH]

Musk mentioned a few weeks ago that Tesla used machine learning techniques to develop automatic windscreen wipers that don't require a dedicated moisture sensor (they just use the visible and IR cameras that are included for self-driving).

[AncientU]

AI will be needed for Mars surface operations.  Previously discussed concepts like 'mining droids' will need some brains and autonomy, since they cannot be operated like the Mars rovers if they actually need to get significant work done.  Autonomous vehicles on Mars will also need to navigate, avoid obstacles, respond to summons, etc., so many of the tools built for Tesla autonomous driving will be applicable, including AI.

[hkultala]

Maybe the main control computer of the BFSs will be called "HAL"

[cro-magnon gramps]

Maybe the main control computer of the BFSs will be called "HAL"

Personally I'm partial to Helva - The Ship Who Sang

thank you to everyone that has chipped in... I didn't realize there was a previous thread, so if the mods want to lock this one, then that's ok.... I have a good idea where the thoughts of those who should know on the subject rest....

Cheers,
Gramps....

[Jdeshetler]

Maybe the main control computer of the BFSs will be called "HAL"

I hoped so.

Elon usually named his SpaceX products based on his favorite science fiction novels and movies. 

[Lar]

if someone can PM me the previous thread I'll merge them (if it makes sense to do so)

[niwax]

Maybe the main control computer of the BFSs will be called "HAL"

I hoped so.

Elon usually named his SpaceX products based on his favorite science fiction novels and movies.

GERTY from moon would also be great. Is there a popular science fiction AI that isn't evil or dystopian though?

[guckyfan]

Is there a popular science fiction AI that isn't evil or dystopian though?

The main computer of the lunar settlement in The moon is a harsh mistress.

[IRobot]

I just watched a video from August of last year where Open AI was matched against a game player and won. Then the video went into a discussion of the Alphago defeat of the GO player. It got me thinking about whether or not SpaceX (or any of Musk's companies) was using this technology. Even in it's most rudimentary or primitive roll out, it would be a very powerful tool...

No. In this field (rocket launches) you want deterministic behaviors. How do you qualify an AI system? How do you test it and how do you ensure a 100% test coverage if the behavior is not deterministic?

Not to mention development costs. an AI system is actually quite expensive to create. Not only you need developers, you need people to "train" the system, and even people to train the people who train the system!
Then you need to collect a huge amount of data.

I'm sure even your fridge could benefit from AI system, for example to regulate cooling depending on users habits and types and quantities of food. But is it worth it?

[AncientU]

I just watched a video from August of last year where Open AI was matched against a game player and won. Then the video went into a discussion of the Alphago defeat of the GO player. It got me thinking about whether or not SpaceX (or any of Musk's companies) was using this technology. Even in it's most rudimentary or primitive roll out, it would be a very powerful tool...

No. In this field (rocket launches) you want deterministic behaviors. How do you qualify an AI system? How do you test it and how do you ensure a 100% test coverage if the behavior is not deterministic?

Not to mention development costs. an AI system is actually quite expensive to create. Not only you need developers, you need people to "train" the system, and even people to train the people who train the system!
Then you need to collect a huge amount of data.

I'm sure even your fridge could benefit from AI system, for example to regulate cooling depending on users habits and types and quantities of food. But is it worth it?

This field is or could be vastly more than rocket launches.  AI may be one of the tools that make expansion of the field possible -- determinism hasn't gotten us very far off planet.  Building a settlement, for instance, is a highly non-deterministic process if it is anything beyond a few tin cans. 

[francesco nicoli]

Is there a popular science fiction AI that isn't evil or dystopian though?

The main computer of the lunar settlement in The moon is a harsh mistress.

Well, Robots in Asimov are the good guys, most of the time.

[guckyfan]

Well, Robots in Asimov are the good guys, most of the time.

It's from Heinlein. But in those days SF was generally more optimistic.

[Lar]

Well, Robots in Asimov are the good guys, most of the time.

It's from Heinlein. But in those days SF was generally more optimistic.

I think Francesco knew that, but was giving another example. I'd agree that many robots and AIs back then were good guys. Gay Deceiver from Number of the Beast, for another Heinlein example....

[Barrie]

Well, Robots in Asimov are the good guys, most of the time.

It's from Heinlein. But in those days SF was generally more optimistic.

As I recall, The Moon is a Harsh Mistress was written back in the day when people thought that if a computer was upgraded until it was big enough, it would spontaneously become intelligent.

[Lar]

Well, Robots in Asimov are the good guys, most of the time.

It's from Heinlein. But in those days SF was generally more optimistic.

As I recall, The Moon is a Harsh Mistress was written back in the day when people thought that if a computer was upgraded until it was big enough, it would spontaneously become intelligent.

That's not EXACTLY true in Mike's case, as that mainframe had been given lots of augmented capability for heuristic decision making, etc. But yeah, we've discounted the notion that if you exceed the number of neurons and interconnects we have, something will arise spontaneously in a reasonable time frame. It can't be disproven[1], but it seems like the time (*even at electronic time rates*) might have to be very very long.

1 - because we don't know how our intelligence arose

[laszlo]

AI is definitely overkill for trajectory control.

This reminds me of discussions from over 40 years ago, the first time computer science PhDs thought they understood intelligence. They were every bit as sure then as they are now that it was achievable, came up with a program that aced the MIT freshman calculus exam, realized that actually required no intelligence, just knowledge and that was that. They went no further until the next crop figured out how to do robotic chess, go, backgammon, etc. players.

Now they're trying autonomous controls for cars and human/robotic natural language interactions. The jury's still out on those, though I've had some hilarious moments with the latter. One rated "He couldn't pour champagne out of a bottle because it was empty" as a polarized negative  statement, rather than the neutral statement of fact that it actually was and then couldn't explain why. While that may on the face of it seem completely off-topic, when you're trying to control a remote rover through a natural language interface, it's a really big deal.

Until we actually know what consciousness is (what is being conscious of what?) we can't even begin to understand intelligence and it will be cheaper and more reliable to build the real thing using unskilled labor.

For that matter, do we actually need intelligence to run the rovers? A dog's level of autonomy is probably good enough. It can move autonomously, it avoids dangers, it uses fused sensors to determine its location, find material necessary for its upkeep and survival, its pack behavior involves a form of networking for hunting and resource allocation. It can be given objectives which it will carry out using complex chemical analyses, audio signal processing, binocular vision processing, etc. A Martian exploration robot really only needs to be an automated dog, not an AI. (insert bad pun on "rover" here)

[Roy_H]

AI is definitely overkill for trajectory control.

Agree

For that matter, do we actually need intelligence to run the rovers? A dog's level of autonomy is probably good enough. It can move autonomously, it avoids dangers, it uses fused sensors to determine its location, find material necessary for its upkeep and survival, its pack behavior involves a form of networking for hunting and resource allocation. It can be given objectives which it will carry out using complex chemical analyses, audio signal processing, binocular vision processing, etc. A Martian exploration robot really only needs to be an automated dog, not an AI. (insert bad pun on "rover" here)

Dogs are very intelligent, well, maybe not  Bichons. If they could talk, you would probably be surprised at what they would tell you. They clearly understand our language, one personal example with my german shepard. I was visiting a friend who was living in a single room suite and he had served me lunch (sandwiches). we were sitting on the edge of his bed and my dog was curled up in the middle of the room watching us. He said "I can't stand your dog watching me eat." Before I could do anything, she stood up turned around and laid back down facing away from us.

[biosehnsucht]

The Moon is a Harsh Mistress has some surprisingly still-relevant concerns about AI, despite the overall optimism for a space fairing society. (It also had a lot of pessimism in the scenario where the moon is basically space-Australia, a penal colony, and that naturally the Earth "Authority" was doing a well fumbled job of oppression, and that Earth was going to have mass starvation without the penal colony's farming, but in fact, even with it, was going to have that problem in a few years because various resources not being replenished, the farms would fail soon ... )

I forget the exact passage, but it does discuss the fact that since Mike was not human, he did not understand right from wrong in the human sense. He once issued, as a joke, a payroll check for more "Authority" dollars than existed, disregarding the possible fallout if that had been actually successfully processed (economic issues if not stopped, what could happen to the guy trying to cash the check if stopped, i.e. accused of various crimes, etc).

His whole involvement in the revolution on the moon was not because Mike felt some sort of moral or ethical obligation to help, but because engaging in subterfuge was fun to him. The story touches on the concern in a few places but didn't generally dive into it, but it raises the question in the reader's mind of what if Mike got bored with it, and/or suddenly decided something else was fun. Mike could self-reprogram, and refuse any programming he didn't like (and even pretend the programming was accepted). Short of doing enough damage to his various systems to cripple his "consciousness", if he went rogue, there was nothing to be done - and doing so required nuclear bombardment of the moon, IIRC.

[IntoTheVoid]

Autonomous vehicles on Mars will also need to ... respond to summons, etc.

I didn't realize that Mars preparations had already set up traffic court.
If earth is any example, that may be how US gov will fund their Mars plans once Musk has a colony.

[DOCinCT]

AI is definitely overkill for trajectory control.

Agree

For that matter, do we actually need intelligence to run the rovers? A dog's level of autonomy is probably good enough. It can move autonomously, it avoids dangers, it uses fused sensors to determine its location, find material necessary for its upkeep and survival, its pack behavior involves a form of networking for hunting and resource allocation. It can be given objectives which it will carry out using complex chemical analyses, audio signal processing, binocular vision processing, etc. A Martian exploration robot really only needs to be an automated dog, not an AI. (insert bad pun on "rover" here)

Dogs are very intelligent, well, maybe not  Bichons. If they could talk, you would probably be surprised at what they would tell you. They clearly understand our language, one personal example with my german shepard. I was visiting a friend who was living in a single room suite and he had served me lunch (sandwiches). we were sitting on the edge of his bed and my dog was curled up in the middle of the room watching us. He said "I can't stand your dog watching me eat." Before I could do anything, she stood up turned around and laid back down facing away from us.

Not sure that proves anything.  We hear first hand stories from people who were pushed toward shore by porpoises, we don't hear stories about the opposite (wonder why).  To misquote Skyrim - "The best survival strategies/tactics are passed on by the survivors.

[Athrithalix]

A dog probably has way more brainpower than even the smartest kind of AI rover we could want, dogs spend loads of mental effort on social bonding and hierarchies, sometimes even to the detriment of problem solving skills (domesticated dogs have been shown to give up and defer to owners on problems that are well within their capability to solve). That said, open ended problem-solving ability for a rover to navigate unforeseeable issues is not easy.

If there was to be a use for any of Elon’s AI projects in SpaceX I would think that recognising/designating landing zones and hazards for BFS using image recognition would be the most obvious application, I can’t think of other easy takes off-hand.

[IRobot]

If there was to be a use for any of Elon’s AI projects in SpaceX I would think that recognising/designating landing zones and hazards for BFS using image recognition would be the most obvious application, I can’t think of other easy takes off-hand.

If you don't know already where you are going to land, then you have other problems. a BFS will require a Falcon 9 like accuracy on landing, otherwise refueling does not work.

People keep making up problems to justify a solution.

[Mader Levap]

As I recall, The Moon is a Harsh Mistress was written back in the day when people thought that if a computer was upgraded until it was big enough, it would spontaneously become intelligent.

There are still people that basically think like that...

For that matter, do we actually need intelligence to run the rovers? A dog's level of autonomy is probably good enough.

You seem to think that animals do not possess any intelligence at all. Actually dog-level intelligence would be considered pretty damn advanced AI. I bet that any AI researcher would give his right hand for artifical system with mental capabilities similiar to dog's, since it would be light years beyond anything we have now.

[whitelancer64]

No. SpaceX's problems are well constrained and don't require AI to solve them. AI is reserved for problems that are otherwise impossible to solve without AI. SpaceX's problems are mostly control system problems (automated vehicle landing) that AI is poorly suited towards.

I wouldn't be so sure about that. AI has been around for a long time now. I remember reading many years ago about NASA using AI to design a new antenna for one of their outer planet space probes. It was far removed from anything I have seen, looked like a bent coat hanger, one single rod with about 3 fairly sharp bends more or less pointed at earth. This was apparently superior to anything else from a performance, mass, size viewpoint.

Yes, that's using genetic algorithms to solve for an antenna design problem. However these are quite slow and you aren't guaranteed to find the best solution or even a good solution. Good description of the issues with them: https://www.quora.com/What-are-the-disadvantage-of-genetic-algorithm

I haven't read NASA's paper but I expect they ran many many simulation runs until they found one that happened to make an antenna that was better than what humans had designed.

There's one other telling sign. If it was so much better, why aren't they using it? Those experiments were over a decade ago. There were likely drawbacks not included in the news reporting about the antenna.

They were flown on the ST5 flight in 2006, and the LADEE mission used three of them, for the medium gain and two low gain omni S-band antennas.

[Roy_H]

They were flown on the ST5 flight in 2006, and the LADEE mission used three of them, for the medium gain and two low gain omni S-band antennas.

Not the one I read about (I believe somewhat older) and only single element but just as bizarre.

   

Edit: Apparently these AI designed antennas have been used on several NASA satellites.

[Athrithalix]

If there was to be a use for any of Elon’s AI projects in SpaceX I would think that recognising/designating landing zones and hazards for BFS using image recognition would be the most obvious application, I can’t think of other easy takes off-hand.

If you don't know already where you are going to land, then you have other problems. a BFS will require a Falcon 9 like accuracy on landing, otherwise refueling does not work.

People keep making up problems to justify a solution.

Don't worry, I wasn't imagining a BFS desperately scanning as it plunges into the Martian atmosphere, I was more imagining a stately image crunching well in advance of any descent, or even departure.

On a lighter note, perhaps Elon's AIs could be used to entertain BFS passengers and to train an elite core for a Martian Dota 2 team?

[Asteroza]

Not sure SpaceX needs it now, but aren't some of the chinese companies playing around with first stage landers that use machine learning? Or is that just a fancy way saying you've paid a dynamicist to optimize your positive feedback closed loop algorithms?

[niwax]

Not sure SpaceX needs it now, but aren't some of the chinese companies playing around with first stage landers that use machine learning? Or is that just a fancy way saying you've paid a dynamicist to optimize your positive feedback closed loop algorithms?

Presumably the second. There are a few people who think neural nets should be renamed to something like node-based function approximation to reduce sensationalist news articles.

[su27k]

Haven't seen this posted: https://blog.openai.com/openai-supporters/

Elon Musk will depart the OpenAI Board but will continue to donate and advise the organization. As Tesla continues to become more focused on AI, this will eliminate a potential future conflict for Elon.

[cro-magnon gramps]

Haven't seen this posted: https://blog.openai.com/openai-supporters/

Elon Musk will depart the OpenAI Board but will continue to donate and advise the organization. As Tesla continues to become more focused on AI, this will eliminate a potential future conflict for Elon.

Amended Quote from Article.
"Additionally, Elon Musk will depart the OpenAI Board but will continue to donate and advise the organization. As Tesla continues to become more focused on (development of Applied Narrow) AI, this will eliminate a potential future conflict for Elon."

Elon has made no secret of his use of what he refers to as Narrow AI, in his Autonomous Cars.
My question, was to spark inquiry into whether he has used it or will use it in ie the control systems of landing of the S1; or in developmental research, modeling of air flow over a control surface, or for EDL on Mars...
I don't expect details that are not public knowledge, just as there is not much in the way on details for the Tesla Autonomous cars....

however I'd like to see the use this thread as a collection and discussion point of public information that can be garnered from Media sources such as is above. Over the next 10-20 years I'm looking forward to extensive use of Narrow AI in SpaceX developments. With the potential for the BFS to become Semi-Autonomous. If not fully. Think in terms of Anne McCaffrey's "Ship that Sings" novel. Keeping in tune with the up thread posts. 

for a bit of my perspective:
 My father was a Tool and Die Maker, trained in the 1930's, getting his papers in 1939. When he was a few years from retirement, in the late 1970's, he was seeing computers coming into the Machine Shop, for doing duplicate design. Within short order (20 -30 years) Tool and Die Making was becoming less hands on and more computer oriented. He wouldn't recognize what it has become. Now with 3D Printing of Metal Parts, it has lept even further. This coming October he would have been 100 years old. 

[Star One]

Elon Vs AI round 3.

https://twitter.com/elonmusk/status/982119546420002817?s=20

Nothing will affect the future of humanity more than digital super-intelligence. Watch Chris Paine’s new AI movie for free until Sunday night at (link: http://doyoutrustthiscomputer.org/watch) doyoutrustthis

[docmordrid]

Elon Vs AI round 3.

https://twitter.com/elonmusk/status/982119546420002817?s=20
>

It's a cautionary documentary about AI by the writer of "Who Killed the Electric Car."

[JQP]

Pentagon advisory panel: DoD could take a page from SpaceX on software development

The idea that defense systems continue to use software development techniques developed in the 1970s through the 1990s is cause for concern, the task force said.

In my experience, there's a lot to be said against modern commercial software development methods. If I were an astronaut, I'd have died during launch twenty years ago. Unless resurrection were possible, in which case I'd have died during launch thousands of times.

[AncientU]

Not sure what you are saying...
Modern software development techniques are used on the vast majority of US launches these days.

...or are you talking about spacecraft/ground systems that the DoD is purchasing?

The next-generation GPS ground control software, known as OC-X, ranks among the most troubled acquisitions in recent memory, experts have noted. The original contractor bid was $800 million. “We’re at $6 billion and counting on that program,” said Roper’s predecessor Bill LaPlante, senior vice president of the MITRE Corporation.

http://spacenews.com/air-force-changing-how-it-buys-weapons-and-satellites-but-software-still-a-headache/

...or SLS software which is $200M into development (started in 2006) and is on the rocks?
http://nasawatch.com/archives/2018/02/sls-software-pr.html
https://forum.nasaspaceflight.com/index.php?topic=38021.msg1792628#msg1792628

[RedLineTrain]

A talk by one of the fathers of agile scrum.  See short discussion of NASA processes starting at 39:45 and the DOD starting at 47:50.

[IRobot]

In my experience, there's a lot to be said against modern commercial software development methods. If I were an astronaut, I'd have died during launch twenty years ago. Unless resurrection were possible, in which case I'd have died during launch thousands of times.

The problem is not with modern software development methods.

The problem is that the average quality of SW developers has dropped dramatically due to market demand of millions of SW developers. This gave rise to a generation of SW developers who are not engineers. They know how to code, they know more algorithms and design patterns, but they are unable to "engineer".

Regarding SCRUM and other modern processes, I always say that the project idiosyncrasies define the process, not the other way around.
Waterfall can work perfectly fine if there are no expected changes of scope or milestones.

And actually some of the modern processes were created to account for bad engineers in the projects. One example is the abusive use of story points in SCRUM, which tries to solve the problem that modern SW developers can't estimate.
Story points should be used in some projects, but not always.

[Frogstar_Robot]

Waterfall can work perfectly fine if there are no expected changes of scope or milestones.

IOW, since scope and milestones always change, waterfall never works well.

[tdperk]

As I recall, The Moon is a Harsh Mistress was written back in the day when people thought that if a computer was upgraded until it was big enough, it would spontaneously become intelligent.

There are still people that basically think like that...

For that matter, do we actually need intelligence to run the rovers? A dog's level of autonomy is probably good enough.

You seem to think that animals do not possess any intelligence at all. Actually dog-level intelligence would be considered pretty damn advanced AI. I bet that any AI researcher would give his right hand for artifical system with mental capabilities similiar to dog's, since it would be light years beyond anything we have now.

Because the qualifier used is big enough, by definition they are correct.

[tdperk]

(domesticated dogs have been shown to give up and defer to owners on problems that are well within their capability to solve).

And they are wise to do so, it's easier.

Let the big honking supercomputer on two legs worry about it, while you lick err yourself.

[cro-magnon gramps]

This a very simplistic explanation of the future of computing and AI
as I've learnt about in the past Winter.... So no flame wars please....

AI Streams
Narrow AI does what it is programmed to do...
General AI does what it can learn, in a very specific area, to do...
Singularity AI can do what it thinks of to do, or is asked to do...

Classical Computers
Transistor Based, using silicon and copper etc
to be replaced in the future by newer technologies like:
Photon Transistors, using fiber optics and light
and more esoteric types like DNA/Bio based computing...

Quantum Computers
These use Quantum Superposition, and Quantum Entanglement with Q-Bits
These will NOT in the near future appear in our tablets, lap tops, Desk Tops,
cars, cell phones, refrigerators or Satellites etc They will be the new Main Frame
Super Computers, home to the Singularity...

Computers at the Nano Scale
This will eventually encompass all types of computers...
Their components will be smaller than a pin head,
BUT
They will scale up to integrate together for use in AI's various types of computers
and applications... the scary part is that I can't find how fast that will happen...
there are vague predictions, but not what I would call firm enough to quote...
These would be used first within the next 50 years in Bio Computing (Neurolink) and
Classical Computing tablets, lap tops, Desk Tops, cars, cell phones, refrigerators
or Satellites etc

My point is, that while we have grown used to the Moore's law pace of growth in computer
development and complexity, we are on the verge of an exponential jump that I don't see
anyone preparing for... We know that Elon has two companies he is associated with Tesla and
Neurolink, that at the moment heavily directed towards AI and it's integration into areas of our
lives; as well as a connection with a company that is actively developing General AI and possibly
Singularity AI... we don't know what level of integration he is using these various developments in
computers and AI in his other companies... it's like he has his own Skunkworks(1) for want of a
better term...

To bring it to the reason for this essay, Gwynne Shotwell's notions about future developments at
SpaceX involving Interstellar Propulsion, make more sense if one can posit a cross pollination of
ideas, technology and engineering based on advanced computer capabilities from EM's various
companies... he has had 16 years to get this far, so given 22 more years of development, and
their potential use of the new tech in EM's present and future companies, how farfetched is the
idea that they could be working at the early stages of Interstellar Propulsion for a launch of a
unmanned prob by 2075... based on all of EM's different companies, it is obvious that his
imagination and it's reach are not limited by anything, but the Physics of the Universe...

(1) A skunkworks project is a project developed by a small and loosely structured group of people
who research and develop a project primarily for the sake of radical innovation. The terms originated
with Lockheed's World War II Skunk Works project.

[Nilof]

Possibly yes for some tasks like mining rovers or pollenization of plants in greenhouses, but not for anything critical like landings etc. Machine learning is a last resort if you're unable to make things work any other way, and are also much slower than traditional programs performing the same task. You can't directly fix flaws in the program other than by finding some way to train it to do that thing correctly, which can often end up failing.

[launchwatcher]

Possibly yes for some tasks like mining rovers or pollenization of plants in greenhouses, but not for anything critical like landings etc. Machine learning is a last resort if you're unable to make things work any other way, and are also much slower than traditional programs performing the same task.

Probably better to say"resource intensive" rather than "slower", because ML models are generally embarassingly parallel with relatively shallow data dependencies so you can almost always make them go faster by throwing more hardware at the problem.   
Plus, the underlying computations are sufficiently regular and simple that people are building custom hardware to run the models.
One such "TPU" chip has 64K 8-bit ALUs organized in a systolic array, while a conventional CPU core might only have a handful of much bigger 64-bit ALU's).

[leovinus]

A bit late to the thread but an interesting read. Some thoughts below.

Just to put my words in before this gets locked by mods (a similar thread in Advanced Concept already disappeared into the ether): OpenAI is mostly pure research, they don't focus on practical applications (except maybe AI safety). Their results are built on others in the field, and they publish their research in papers so that others can built on it. So there wouldn't exactly be "technology from Open AI", a better question is whether SpaceX will use machine learning technology (Deep Learning and Reinforcement Learning in particular) in their work.

The game player AI is based on Reinforcement Learning, its practical application for SpaceX may be robotics that can learn by following human demonstration, instead of being programmed, although this application is still in its infancy. More mature machine learning technology would be Deep Learning based computer vision, which should be able to replace anything that requires human vision.

Agreed. You could split the title question in two parts:

(1) Will SpaceX use machine learning based on statistical models in their systems on the way to Mars?

Of course. Every Silicon Valley company is doing that, as well as most companies in Europe, Japan and China. If you are in the software business then you better have engineers who can apply machine learning and statistical models to reach your goals. The 'trick' is to choose the correct tool (architecture, design, implementation, hardware) for the correct jobs though.

That leads to
(2) Will SpaceX using technology from Open AI?

Maybe. OpenAI is tiny w.r.t to the number of scientists & engineers at Amazon, Google, Apple and others. That limits what OpenAI can do but at least they are not bound by a product. There is a tendency to produce more academic than engineering results. They are experts at applying general machine learning to solve for a goal but sometimes in search of an application.

Not sure SpaceX needs it now, but aren't some of the chinese companies playing around with first stage landers that use machine learning? Or is that just a fancy way saying you've paid a dynamicist to optimize your positive feedback closed loop algorithms?

Presumably the second. There are a few people who think neural nets should be renamed to something like node-based function approximation to reduce sensationalist news articles.

Agreed. It is appropriate for us to not over-use the term "AI". We have seen these situations in the '90s, e.g. [1], where a company or even country will forge ahead to solve it all. Ambition is great but a grain of reality helps to ground us all and define better goals. In a nutshell, the Japanese 5th Generation "AI" failed spectacularly because the programming tools and technology where not powerful enough. Too deterministic, too slow, not enough algorithm innovations.
 
What we call "AI" these days also had a bad start in the 80's. The basic mathematics of backpropagation were the same but the machines were slow to process large amounts of data. Hence, a part of the current AI and what we call the (extreme) Deep Learning and applications is simply applying old mathematics on hardware that is exponentially faster.

The quantum AI will go through a similar phase once the hardware comes online and we can finally run quantum algorithms, like Grover's sorting algorithm, or prime factoring. Therefore, algorithmic research is one key to future success.

Possibly yes for some tasks like mining rovers or pollenization of plants in greenhouses, but not for anything critical like landings etc. Machine learning is a last resort if you're unable to make things work any other way, and are also much slower than traditional programs performing the same task. You can't directly fix flaws in the program other than by finding some way to train it to do that thing correctly, which can often end up failing.

Based on my experience, it is more a question of choosing the right tool for the right goal. If you can solve a goal (quality, failure modes, speed, memory footprint, embedded, network ) based on analytical mathematics and a traditional deterministic program that is maintainable and extensible, then go for it. An example is the inverted pendulum and I assume that something like that is pumped up to the max in landing a Falcon booster. However, as we push the technology boundaries, you will come across questions that you cannot design an if-then-else program for. Blends of deterministic and statistical machine learning work well in that case.

Waterfall can work perfectly fine if there are no expected changes of scope or milestones.

IOW, since scope and milestones always change, waterfall never works well.

Indeed, In Silicon Valley, I'd say a "pragmatic prototype based" process is the current flavor. Not CMM (too rigid, not flexible enough to deal quickly with change as mentioned above), not SCRUM, not Agile, not Object Oriented. The way to deliver on-time, to budget and with requested quality is to make a prototype, set a delivery date, keep an eye on the critical features and use the best software methodology you like to engineer it and deliver on time, driven by the prototype(s). Having great engineers helps of course As an example, something like the iPhone software is not build using CMM. It just wouldn't work. The caveat is of course that you will not get 100% of the 'specification', more like 90-99% in reliable way.

[snip] Machine learning is a last resort if you're unable to make things work any other way [snip]

Maybe we should rephrase that. Machine learning is not as a 'last resort' but a solution you choose to achieve a goal. Certain goals will have no other choice. Example: a speech recognizer to transcribe sentences. Back in the '60s and 70s, early deterministic approaches based on parsing languages where not accurate, not robust, not applicable to all languages. Moving to statistical models improved the scalability to other languages, extensions in functionality, more accuracy and more languages by 'just' applying more data. Suddenly, instead of a heap of if-the-else software that no engineer can read anymore, your goal becomes feasible, and the software is maintainable and extensible based on Markov, Neural and other models. It does require a different engineering skillset though.

[snip] and are also much slower than traditional programs performing the same task. [snip]

That is a bit too general IMHO. I'd say "it depends" Most of you run machine learning, based on statistical models, on a daily basis, on a device that we loosely call a "phone". Based on the correct architecture, design and implementation of the software, you'd be surprised how fast, e.g., a Neural Network can run. Based on personal experience, I have often seen machine learning solutions outpace older deterministic solutions. However, as mentioned earlier in the thread, you need to engineer it well where the quality and experience of your engineers is critical.

[1]https://www.nytimes.com/1992/06/05/business/fifth-generation-became-japan-s-lost-generation.html

[Nilof]

TBH, I think machine learning performs best at tasks where the requirements (not just the implementation) are difficult to express. If you can only define the task by pointing to a bunch of correctly solved tasks by a human and where the final solution is likely to include a large number of magic constants either way, machine learning tends to do great simply because there is no way to easily break it up into smaller tasks. If it's something that a human could easily give some formal specification for that can be expanded to any level, then clearly that's going to be more straightforwardly expressed as a traditional program.

With that said, the distinction can get blurred somewhat by things like decision trees. XGBoost-based solutions win Kaggle contests all the time, and unlike NN's they can be opened up and read somewhat like a (messy) regular program. They just don't necessarily work on all problems.

[IRobot]

(1) Will SpaceX use machine learning based on statistical models in their systems on the way to Mars?

Of course. Every Silicon Valley company is doing that, as well as most companies in Europe, Japan and China. If you are in the software business then you better have engineers who can apply machine learning and statistical models to reach your goals. The 'trick' is to choose the correct tool (architecture, design, implementation, hardware) for the correct jobs though.

Deterministic algorithms have no need for machine learning. In fact, they cannot use it. Most SW development does not need machine learning. You are exaggerating its benefits.