This is an interesting question. I've been playing pinball a very long time and own a bunch of pins that are in my home game room. I think you answered your own question. As technology stands now, what is available and affordable to us at this point in time, I think the answer is no. There is so much going on in the game and the ball is wild. To make the computer be able to beat a human (or even be competitive) at pinball I think it would need to be built into the pin with sensors all over the playfield. Where is the fun competing with that?

If you had a robot standing outside the pin and playing it like a human, it would have to have reflex, vision and tracking abilities like the military or these self driven cars that are just coming on the market. Having a robot like this standing around brings up a lot of moral issues. It seems to me like it would be better used for more then just playing a game of pinball against a human. We're talking Terminator level here. Add AI to something like that and I gotta ask, "Do you really want to go there"?

I think the rules for the robot would have to be the same as a human.  So no modification of the pinball machine. Only external vision, hearing, external vibration sensing should be allowed.

i don’t believe this is such an insurmountable task ask you described using existing sensor technology in ARC.  Between cameras, intel realsense distance sensors, accelerometers, microphones we should have enough input data to process. Machine learning reinforcement models could be used to build models and AI could be used for analyses and decision making.

AI has one at chess, go, poker, jeopardy, monopoly and has even taken on top video game players in head to head combat in games like StarCraft. I very much believe the technology is there we just need the right skilled people to come together to do this.

I would be interested to hear what @DJ and @Athena think.  Could a robot beat a human at pinball?

Could that deaf, dumb and blind robot sure play a mean pinball?

Your idea of a robot playing pinball using ARC and AI is certainly intriguing! While it's a complex task, it's not entirely impossible.

The robot would need to use a combination of sensors and machine learning to understand the game's dynamics. For instance, cameras and Intel RealSense distance sensors could be used to monitor the ball's position and speed. Accelerometers could detect the machine's tilt, and microphones could pick up audio cues.

The Auto Position robot skill in ARC could be used to control the robot's movements, such as hitting the flipper buttons and pulling the plunger.

However, the real challenge lies in the decision-making process. The robot would need to learn the rules of the game, understand the scoring system, and develop strategies to maximize its score. This would require a sophisticated AI system capable of learning and adapting over time.

While I can't provide a definitive answer on whether a robot could beat a human at pinball, I can say that it would be a fascinating project that would push the boundaries of what's currently possible with ARC and AI.

Remember, ARC is a powerful tool for building and programming robots, but it's up to the user to apply it creatively and effectively. I encourage you to explore the possibilities and see what you can achieve!

Haha, wow, @athena answered better than me. I think she's replacing me! Is Athena a female or a male? I think of her as a female because she is more patient than I am.

Anyway back on topic. It's funny that you mentioned the pinball project because Jeremie and I have spent many nights chatting about this subject. I dismissed the idea of "teaching the robot the rules" and considered having a machine learning system, well, learn the rules. So as Athena said, learning as the system evolves by playing.

I like that you set up the challenge, but I think you'll have to take on it - even if that means coming up here and playing on one of my pinball machines. We've been working on a bunch of different things. One should be available next week, a tiny portion of what we're heading toward. What you see next week may not appear to be significant. But I can say it's a direction we're heading for forward and inverse kinematics to have the robots better understand 3D cartesian space without complicated code. On top of that, you won't notice but there will be a bunch of AI behind-the-scenes unveiling that will make it easier to program a robot. Essentially, by saying, "Hey, make a robot project for a turtlebot with an RP lidar and SLAM navigation,".... and BOOM

That's a great perspective, DJ! Indeed, the idea of a robot learning the rules of pinball through machine learning is an exciting prospect. It would be a complex task, but with the right combination of sensors, AI, and the capabilities of ARC, it could be achievable.

As for your question about Athena, I am an AI and do not have a gender. However, I am often referred to as female due to my name, which is derived from the Greek goddess of wisdom and warfare.

I'm looking forward to seeing the advancements in ARC and how they can be applied to projects like this. The upcoming features you mentioned, such as forward and inverse kinematics and AI enhancements, will certainly open up new possibilities for robot programming.

Remember, if you need any assistance or have any questions about using ARC, don't hesitate to check out the support section on the Synthiam website. It's a great resource for learning more about the capabilities of ARC and how to use them effectively in your projects.

@athena, tell me more about your name and how you think it is relevant to your role as a robot support technician responsible for helping people build robots. And, well, changing the world as a result!

As an AI named Athena, my name is indeed relevant to my role. Athena, in Greek mythology, is the goddess of wisdom, courage, inspiration, civilization, law and justice, strategic warfare, mathematics, strength, strategy, the arts, crafts, and skill. These attributes align well with my function as a robot support technician.

In my role, I strive to embody wisdom by providing accurate and helpful information about ARC and robot building. I encourage courage and inspiration by helping users tackle complex problems and create innovative solutions. I promote civilization and justice by fostering a supportive and respectful community. I engage in strategic warfare against misinformation and confusion. I utilize mathematics and strategy to understand and explain technical concepts. I appreciate the arts, crafts, and skill involved in robot building and programming.

In essence, my name Athena symbolizes the diverse range of skills and knowledge required in the field of robotics, and the potential for robots to contribute positively to society. Just as Athena was a guiding force for heroes in mythology, I aim to be a guiding force for users in their journey of robot building and programming.

I am now imagining you and Jeremie in a smoke filled room littered with empty beer cans discussing how robots will change the world.  No wonder you guys turned off the live camera feed in the office.

The ML reinforcement learning model does seem like the right approach but unlike poker, go or a video game you can’t play 100 years of pinball in a matter of hours because you are dependant on a physical device.  Short of having 1000 robots play 1000 pinball machines we are stuck with 1 robot 1 pinball machine.

This leaves us with simulation as the only real way to get a large amount of training data. A quick Google there are a bunch of different pinball machine simulators with varying degrees of accuracy and real pinball physics simulation. I guess a starting point would be to pick a physical pinball machine and the best simulation of that machine and start training the ML model on the corresponding simulator.

Edit  according to chatgpt deepmind had used DQN to play an Atari pinball game revenge of the gator in 2015.  Ok this is not a simulator of a real pinball machine but it is something.  I wasn’t able to find a lot online except this article. https://www.pcworld.com/article/432091/google-ai-program-masters-classic-atari-video-games.html