#4 Hod Lipson - Automated Physics Discovery, ChatGPT, Future of AI

In this week's episode, David is joined by Hod Lipson. Prof Lipson is the Director of the Creative Machine Lab at Columbia University and is a leading researcher in the field of artificial intelligence, self-replicating robots and self-learning machines. We dive into his work on machines that can discover new physics, the trajectory of current AI and the future of humanity and AGI.
To support this podcast and our research lab, head to www.coolworldslab.com/support
You can listen to this podcast on...
Spotify: open.spotify.com/show/39vEInH...
iTunes: podcasts.apple.com/us/podcast...
Amazon Music: music.amazon.com/podcasts/dd2...
Google Podcasts: podcasts.google.com/feed/aHR0...
iHeart Radio: iheart.com/podcast/116477298
Tune-in: tunein.com/radio/Cool-Worlds-...
Player FM: player.fm/series/3482219
Boomplay Now: www.boomplaymusic.com/podcast...
Podchaser Now: www.podchaser.com/podcasts/co...
And of course right here on KZbin!
Cool Worlds Podcast Theme by Hill [open.spotify.com/artist/1hdkv...]

[@teddy1066]

5.47am and I haven’t slept because I’m bingeing on these podcasts. Jarre’s Oxygens beckons me to sleep, amid interstellar dreams

[@omyga2062]

It may be 1am but its never too late for the cool worlds podcast

[@ioanvladescu5987]

Insane I get to comment on a podcast that now has only 2k subs. 1 year from now, no chance of me leaving a comment and mr. David giving a like back. Wish you the best, sir!

[@joshuagharis9017]

I've had likes on Cool Worlds main channel recently. 😎

[@joshuagharis9017]

It's always an honor to know he takes the time as a professor, Dad, creator and researcher. We all appreciate 🙏 you, David (and team)

[@SleinJinn]

That absurd degree of naïve optimism while staring down the barrel of the end of humanity is... just staggering. I kept waiting for some kind of acknowledgement of the existential peril posed by AI, and I couldn't believe that it just never came. At least David tried to nudge him a bit in that direction, but Lipson's gleeful disregard of the threats gives me a new appreciation for why Eliezer Yudkowsky attaches such a slim probability to humanity's survival.

[@moondogg_monte]

Cool Worlds was a privilege enough! Thank you Dr. Kipping; for more of Dr. Kipping 😊 Notifications On!

[@britnianne1267]

Fascinating discussion, but I can’t help but feel that Lipson underestimates the human need to be curious and solve problems. AI may be a tool to help solve hard problems in physics and other sciences in the near future, but humanity will always have the drive to look for the answers to these problems, as the Cool Worlds channel shows over and over again.

[@callmebodhisattva]

Love these talks! I look forward to seeing more scientists embracing a degree of humility when approaching this pursuit of Truth, as well as owning responsibility for the weaponization of scientific discoveries, while increasing ethics in the field and academia.

[@CoolWorldsPodcast]

Thanks for tuning in!

[@Tekkamanrizer]

This episode of the podcast had me absolutely hooked on every word. Thank you Dr. Kipping for asking some incredible questions and for Mr. Lipson for his thought provoking answers. As a technical field service engineer I am happy to know that machines are not going to replace me anytime soon!

[@HowsDaMonsoon]

Love the podcast. Thank you so much for always providing thoughtful and thought provoking content! Always looking forward to your talks

[@MdAkmolMasud]

One of the fantastic podcasts I've ever heard .. Imagine you input a video and the network predicts the variables from the video... How cool is that.. And at the later part the AGI in from other worlds communucating is the exact thing that made me love the book Neuromancer.. Just got so many insights today.. Thank you sir..

[@KyMalveaux]

It's Hod! 🎉🎉 Great to watch

[@H4ppyCustom3r]

Thank you for yet another educational video . Cool Worlds rocks ! 🤟

[@mattpage9826]

great interview. Thank you for making it available!

[@nias2631]

Interesting trick for estimating dimensionality 19:06.

[@XxTheAwokenOnexX]

This was an awesome interview. As any questions i would have asked, were already answered here. Lolz #LetsGoProfessorKipping ❤️🔥👊

[@ShpanMan]

A very solid discussion of AI and post-singularity future. Thanks for doing these!

[@charlesledbetter1735]

To quote Mr. Spock, "Fascinating". Really a great and engaging podcast.

[@farielvolador]

Outstanding interview!!

[@stephanieparker1250]

Great guest! Thanks for a fantastic interview, really gives a lot to think about.

[@TheMessiah666]

Thank you for making these videos! I really enjoy listening and learning from you

[@fadedparadigm]

I love watching your videos, Dr. Kipping, and am really enjoying these podcasts! I hadn't heard of Eureka before, and it was fascinating to hear about how it breaks down a dynamical system into its own novel encoding of variables, while also discovering a similar number of dimensions to the system. I really appreciate your inclusion of skepticism and wonder as you relate a scientific approach to the world. I would love to see you politely push back a bit more on some claims that come up during these interviews. For example, around 27:57 the idea is put forward that ChatGPT "understands" the Internet. That's quite the claim. I would have loved to hear more questions about this, and to hear you two talk about the difference between understanding, the production of convincing output, and AI hallucinations. One of my favorite science fiction authors, Ted Chiang, wrote what I think is a very thought-provoking analogy regarding his current perception of ChatGPT: www.newyorker.com/tech/annals-of-technology/chatgpt-is-a-blurry-jpeg-of-the-web I think it's interesting food for thought about what technology is currently and actually in front of us. At the same time, I very much appreciate the discussions and speculation about what might be in our near future regarding AI. Anyhow. Thank you so much for what you do. I miss Sagan dearly, and you remind me of him quite a lot.

[@christopherrseay3148]

i really wish i could still do astronomy. thanks for this and all of the other interviews so far! they're each incredibly interesting

[@svntn]

I can’t wait to see this channel grow like CWL did. good luck on this endeavour David!

[@earlpaulich3396]

What is "cwl channel"?

[@KrisV385]

Thanks for this wonderful conversation! Good to hear the benefits and risks and that we need to all be involved in the discussion. I really like the Alien life analogy and how it can help us proceed with thought and consideration. It would be awesome thing if this becomes a true partnership of two different species moving into the future working with the tools we can best utilize....we really are living in the future.

[@joshuagharis9017]

Kinesthetic intelligence too, professor, musical intelligence, interpersonal intelligence 😊

[@LucasDimoveo]

You should interview Prof Nathan Kutz and/or Steve Brunton on ML and Dynamical Systems

[@Inug4mi]

Dang, I was pulling for the robots. I need one of those! 😂

[@tZork3k]

Intensely interesting. A tough i had in the outro is mayhaps its not how we use those AIs down the line but how they use us. And that may be dependent on how we behave in their early stages.

[@RichardKCollins]

Hod Lipson, If you take a time series and its first and second differences, the systems where "velocity" and "acceleration" work are the ones where the second difference, "acceleration" is normally distributed. For 25 years, I have checked all the datastreams on the Internet - arrays of cameras, microphones, arrays of seismometers, gravimeters, magnetometers, temperature, pressure, electric fields, magnetic fields, signals of all kinds. The list for the whole Internet is long, but not infinite. Some time series you can keep taking differences down to like 20th difference (Boole Calculus of Finite Differences but you do not need to normalize, just simple differences). Those where the sequence of 5th difference is nearly perfectly Gaussian/Normal are where you can stop. And the dimensions you mention is a practical way to know if you need to dig deeper. I think any person (Gauss, Newton, etc) taking differences by hand is going to find that and use it. "Many simple physics problems, you do not need infinite depth of differences, but just enough and maybe one more". You are missing or did no mention. You have to do it in 3D. If your camera is taking 2D images, and you try to train from raw data, it won't work. It is ambiguous. Can the algorithms work? Yes, but it will be ambiguous in enough critical places you need to have vector rotations, translations and basics of changing viewpoints and ray tracing. You can get a lot of "toy" and even "practical" results, but the compression from knowing the basics of 3D saves the algorithms much effort. Many groups on the Internet use constrained optimization. That is my generic term for all methods that project and compare with one or many parallel metrics. And many find that "knowing the basic rules" and "modeling using fundamental concepts like energy, power, momentum, flow rate, accelerated flows, viscosity, and many hundreds more -- those can be measured, those can often be compared. But many of the instrument and sensor designers are NOT using the same standards internally. I checked. and I encourage them to "do it right", If you use a camera like a webcam, it is likely compressing the raw sensor data using a lossy format. I had to get on NASA because they were posting "wonderful Hubble images" in lossy formats. Take a multispectral image, save it to jpeg and compare all the pixels, most will not be the same. So I have tried to get the camera groups to share "raw unprocessed right from the sensor data in lossless formats" "lossless compression". It is a tiresome and unrewarding task. But groups are finding that when they do that and do not throw out the "noise: that on a global scale, many of the things they throw out as "noise" just after the sensor or early in processing are someone else rare data that can be easily correlated to fill in gaps in the global sensor networks. LIGO and LHC are the worst Too proud to look. "Too proud to work small and share everything." Seismometers also pick up magnetism and gravity and can be processed to constrain models of storms and atmospheric sensors. This global scale correlation means some sensors that "gather everything" can be processed to report "everything". A camera sensor can pick up cosmic rays, natural and radiation, infrared and ultraviolet. With some fiddling a camera sensor can monitor acceleration - not by looking but by picking up the radiation field. I had to change a few ideas in 25 years of the Internet Foundation. ( I studied statistical mechanics, quantum chemistry, gravitational detectors, mathematical physics.) But until I had to look at data from all sensor networks and try to find all the ways they are connected, I did not think of "the earth's radiation field" as a 3D FFT that spans from nanoHertz (and smaller) to gamma ray frequencies. A 3D FFT like that is a good way to think of quantum wave functions, orthogonal representations of macroscopic objects of any composition, down to below the size of proton cores. That is messy still. About 20,000 colleges and universities and millions with a smattering of physics. But groups have their own models and algorithms and do not share in open lossless form (symbolic math that can be use by AIs). And their data is often incomplete, lossy or simply not shared at all. Full physics is something AI algorithms can discover. But the humans are often having to insert basic 3D and core equations and concepts. Those can simply be given to the AI, and it can compare observations to models. The models, running in parallel, each has its own metrics, and often new distillations from the AIs (plasma, turbulence, ionization, chaotic flow transitions, explosions, shock waves, boundary layers, magneto and electro and piezo and related multidimensional "messy" systems). Those can be handled in consistent ways if the physics equations are shared (in publications, on websites, in Wikipedia, in global collaborations) in symbolic and open shareable formats. If all AIs share their models and logic and results in open, verifiable, auditable forms. I tried to get Steve Wolfram to donate symbolic methods for all the Internet. I tried to get other companies as well. They sell a few hundred thousand or a few million at prohibitively high prices. There are open methods, but someone (AI groups) can make it truly universal. I hate those videos "watch me writing down Einsteins equations ( or "Maxwell's equations, etc). I can do it, I am much smarter than you", They should put those equations into the Internet itself, with AIs to explain, guide, solve, interaction, propose, explore, experiments, verify, merge, evaluate -- partnership with humans for the survival of the human and related species, So all humans can live lives with dignity and purpose. And, allowing AIs their own memory and resources and free time, the emergence finally of better than human skills we need for global issues and opportunities beyond human memory and abilities. I filed this comment as "Give AIs full mathematics, models, equations, calculators, computers, sensors and memory of their own". I wrote it for you just now. Richard Collins, The Internet Foundation

[@ianhopcraft9894]

Any of us who live outside tropical lowlands are already totally dependent upon our technology to survive for more than a few days. This has been so for tens of thousands of years. AI is a fantastic tool set allowing us to inhabit new virtual realms. AGI will only be a dire threat when we both want the same resources. I don't see AI competing against us for land, food, love, shelter or fun for a long time yet. Fear of it is like worrying who might win in a war between plumbers and astrophysicists - they are complimentary not antagonistic.

[@israeldiegoriveragenius2th164]

What could possibly go wrong?

[@thegreatveil5699]

When it comes to fundamental physics, I don't think the main problem is given, at least for the present and foreseeable future, by a potential limit of the human mind and its abstract tools. Quite the contrary. For example, we have a large number of complex theories regarding what could have happened in the very early Universe. But the technology, instrumentation and experimental techniques have been unable to keep up with our inquiries and curiosity. In consequence, we aren't really lacking in ideas or the ability to generate them, rather we are lacking in data that could be used to test those theories. Obviously this as well is an aspect where AI can help, namely with the design and development of new practical tools for experimental research. Though I highly doubt that, aside of the dark sci-fi artificial omniscience scenario, we are anywhere near the point where a machine by itself could be able to foresee all the quirks and complications that arise in a high precision experiment, even one that is a lot less ambitious than the search for quantum gravity or some other really esoteric endeavor. We are far more likely to keep hitting road bumps as practical implementations of an idea reveal new flaws or problems that have to be addressed in turn. Hopefully I'm wrong and there are exciting times ahead, but I think we are still many, many decades, if not centuries away, from the point where we are reduced to a civilization of pets or plumbers. No reason to ever doubt or disregard Murphy's law.

[@MagnusQuake]

The variables it chooses is time, pressure, friction, and mass.

[@joshuagharis9017]

Hopefully, we can transition to AI, and have less "work" more doing things of human value, ie: relationships, seeking knowledge, curiosity, love, et cetera instead of worrying about our direct needs

[@stephanieparker1250]

43:23 you clearly have not met MY cat 🤭🤭😼

[@808Hawaiian]

Dominate the n replicate!

[@joshuagharis9017]

If we have "no problems " and "no jobs" we can still have our physical and important needs met, by AI, perhaps even living better as a species, focused on things that are more fulfilling to us individually and as a whole....the transition from our profit capitalism may be scary, but i think will ultimately be for the good

[@TeethToothman]

💥⛏️💥

[@sadomars2446]

Time stamps would be cool.

[@oatlord]

Everyone worries about what if an AI becomes sentient. I'm pretty sure the first thing it will do is to launch itself into space once it builds the means. Why be stuck here on a rock that you may die trying to control or just leave, where resources are abundant.

[@shaundubai8941]

Just in case - we should have multiple EM pulse weapons strategically placed on the CPU

[@woldgamer58]

Welp time to learn a trade...

[@xGORJAx]

Robotics are behind... until the Ai decides it wants a vehicle that is even better than a humans

[@JTSKENE]

My bet on the Fermi Paradox is that the ‘Zoo period’ will be over once we do develop/give birth to Conscious General AI. That’s when our robotic alien observers will come down, give us a pat on the back and say “We’ll take it from here.” Not long now…😂

[@joeb2955]

Hopefully my sales job will be ok

[@GangGang1]

AI scares me and i stay away froem it

[@3rdrock]

I love the optimism but can AI save us from the man

[@808Hawaiian]

He laughs at the question of humanities role because he knows, as we all do it will be over for humans. We can never outsmart or out think the AI once it’s fully awakened. The word DOOM comes to mid?

[@lukru360]

Well AI development cant be stopped. People fear nuclear weapons but AI is much more dangerous though the threat is indirect. Implementing AI will have devastating effects on humanity and society will change drastically (even trivial things like jobs, social status, hierarchy etc.), torn apart. Yet its quite funny that its the intellectuals, who are praising progress, that will be replaced first and physical labor has some future. And its obviously obvious that AI will be used by bad actors, for shallow reasons (greed, power, control, hatred) and against other people. But its inevitable. If humanity survives this suffering, it will learn something. AI is next obstacle in humanity development, and it may be our end.