Presentation Slides (coming?)
01:00:00 Lloyd: https://robothon.org/
01:03:19 Lloyd: https://www.bainbridgeisland.com/business/barn-bainbridge-artisan-resource-network/
01:06:03 Terry James: The BARN – Bainbridge Artisan Resource Network.
01:07:16 Terry James:
Snocomakerspace. 11604 Airport Rd, Ste D200 Everett, WA 98204
01:10:14 James Newton: Topic idea: Teleoperation. Anyone good at that? It’s an iterest and I have ideas, and a tiny bit of experience, but not much. Would be good to find a speaker or hear from others on that topic.
01:14:05 Bob: Washington State Charter Schools Association Conference https://wacharters.org/conference/
01:15:43 Bob: I am trying to put together some kind of display/presentation for them. If you would be interested in joining on this please let me know. robertphiggins@comcast.net
01:22:47 Steven Kaehler: https://www.maketory.com/
01:28:32 Terry James: Galactic Robotics Society
01:33:07 Terry James: I liked the robot in the new Lost in Space Series
01:33:35 Terry James: I, Robot
01:34:27 Chris Powell: The Cybernetic Imagination in Science Fiction by Patricia Warrick.
01:34:39 Terry James: Lost in Space the movie robot, https://www.google.com/url?sa=i&url=http%3A%2F%2Fthegreatcanadianmodelbuilderswebpage.blogspot.com%2F2015%2F02%2Frobot-from-lost-in-space-1998.html&psig=AOvVaw2kLq80OOiUFGNxsswwrRpk&ust=1705862627160000&source=images&cd=vfe&opi=89978449&ved=0CBMQjRxqFwoTCKDpstHP7IMDFQAAAAAdAAAAABAN
01:34:56 Colin L: Reacted to “Screenshot2024_01_20_104554.jpg” with 👍
01:37:06 Terry James: Robot / AI ship. Disney’s Flight of the Navigator
01:37:28 Abhishek Gupta: Yann is at NYU. He actually won a Turing award 🙂
01:37:59 James Newton: My thoughts and links to videos and discussions about Robots (Automation) and Human Worth.
https://docs.google.com/presentation/d/14hfPI3YTJp3wy9JAdQNyI779PR6kRHPOjW2NjjiCNuM/edit?usp=sharing
01:39:09 Terry James: Robert Miles AI safety.
01:39:09 Chas: I’m going to go with a very Hollywood movie.. I liked Real Steel — but I think it’s because I have a passion for building and he was digging parts out of a salvage yard.
01:40:00 Terry James: AI use for medical diagnosis is very promising
01:41:28 Bernardo Rocamora: https://www.youtube.com/watch?v=144uOfr4SYA
01:41:42 Bernardo Rocamora: Munk Debate on Artificial Intelligence | Bengio & Tegmark vs. Mitchell & LeCun
01:41:54 James Newton: https://marshallbrain.com/manna1
01:43:25 Bob: Here is the debate with Max Tegmark and Yan LwCun on the Ai existential crisis
01:43:34 Bob: https://www.youtube.com/watch?v=144uOfr4SYA
01:45:12 Colin L: Like going 1000 miles to go to an Arby’s.
01:49:50 Terry James: My sound was very static like and then I lost Steve’s last comments.
01:50:30 Terry James: During the presentation we may want to encourage everyone to mute their microphones to reduce feedback.
01:54:56 Terry James: Best use of AI.
01:56:27 James Newton: Reacted to “Screenshot2024_01_…” with ❤️
01:56:38 James Newton: Reacted to “Screenshot2024_01_…” with 😂
01:56:57 James Newton: We all geniuses here
01:57:13 Terry James: Electrical and Optics engineering backgrounds
01:57:42 Terry James: Many inventors.
02:00:04 Terry James: Operant conditioning
02:01:00 James Newton: Make as many paper clips as you can.
02:01:12 James Newton: What could possibly go wrong? ,o)
02:01:39 Lloyd: Reacted to “What could possibly …” with 😂
02:05:26 Bernardo Rocamora: I was at IROS this fall (in Detroit). In one of the plenary talks, Disney introduced us to an amazing robot that was trained with RL
https://spectrum.ieee.org/disney-robot
02:20:42 Terry James: How do you identify a hole?
02:21:37 Terry James: yes
02:23:49 Bernardo Rocamora: Is the prior data in these cases obtained with teleoperation?
02:25:42 Terry James: That was also done by Daniel in Karate Kid
02:27:06 Steven Kaehler: Catch any flies?
02:27:28 James Newton: Reacted to “Catch any flies?” with 😂
02:27:59 JooJu: Reacted to “Catch any flies?” with 😂
02:31:09 James Newton: Of course the reality is, if you can express the idea in all that complex math, then it can actually be implemented in a computer. You can’t just give the common sense advice in English and expect the machine to do it.
02:33:23 Terry James: Human rewards
02:34:16 James Newton: So align the reward algorithm with the human, then let the algo reward the training system.
02:36:25 James Newton: Local minimums
02:42:29 James Newton: Moms are an underappreciated resource… they’ve been training dumb robots for millenia. ,o)
02:42:46 Thurman Gillespy: Reacted to “Moms are an underapp…” with 👍
02:42:52 JooJu: Reacted to “Moms are an underapp…” with 👏
02:42:54 Terry James: Did you see a difference between mens and women approaches to learning and intuition?
02:44:28 Terry James: For example it seems that women work well with young children to guide them but not override their actions which men sometimes do.
02:44:49 James Newton: Reacted to “For example it see…” with 👍
02:45:07 Steven Kaehler: Sometimes good solutions to problems come from unexpected approaches that aren’t logical or obvious to most people.
02:45:55 Terry James: Would that mean that you may want to build randomizing into the rewards to expand the scope of solutions.
02:47:18 Terry James: You mean like unloading the dishwasher.
02:48:39 Terry James: What about a robotic puck?
02:48:45 James Newton: But notice that a randomization from an existing prior data example is a /small/ change from a known solution, not a totally wild thing.
02:48:45 Terry James: Or a magnetic return.
02:50:43 James Newton: Another idea is 1. Locking the simulation to the real world (not easy) via some sort of instrumentation / vision / whatever and then 2. For every real world good training point 3. generate 1000 simulated training points via small random changes from the real world point.
02:52:01 Chris Powell: Simulated annealing?
02:52:13 James Newton: Reacted to “Simulated annealing…” with 👍
02:54:44 Bernardo Rocamora: How do you guarantee that the process is not divergent? In the sense that if the forward task and the backward task are not fully known, errors can accumulate in both directions
02:55:46 Thurman Gillespy: How do vacuum robots do so well at mapping the room? They don’t seem to need as near as much ‘training’.
02:56:26 Terry James: Doesn’t that depend on how cluttered the room is?
02:57:35 Thurman Gillespy: The problem space is much more constrained.
02:58:46 James Newton: Even better than watching paint dry.
02:59:31 Terry James: Yes, chaos theory.
03:12:28 JooJu: 👏
03:12:35 Richard Greenway: Reacted to “👏” with 👏
03:12:35 James Newton: Really an excellent talk and so much more detail to gain from that.
03:12:48 James Newton: 👏
03:13:06 Terry James: Have a great month! Gotta head to work.
03:14:23 James Newton: Also gotta run… wife has a “honey do” list and it turns out I’m her primary robot. :sigh:
03:14:34 Greg Maxwell: Thanks, great topic. Need more physical robot work.
03:14:43 James Newton: Reacted to “Thanks, great topic…” with 👍
03:14:49 Bernardo Rocamora: I need to log off. Thank you to SRS & Dr. Gupta, see you all on the next meeting! Let me know how I can contribute (bernardo.rocamora@gmail.com).