[HN Gopher] DeepMind AI learns simple physics like a baby ___________________________________________________________________ DeepMind AI learns simple physics like a baby Author : mdp2021 Score : 58 points Date : 2022-07-11 20:14 UTC (2 hours ago) (HTM) web link (www.nature.com) (TXT) w3m dump (www.nature.com) | mdp2021 wrote: | Another divulgative article: | | -- DeepMind AI learns physics by watching videos that don't make | sense - An algorithm created by AI firm DeepMind can distinguish | between videos in which objects obey the laws of physics and ones | where they don't - | https://www.newscientist.com/article/2327766-deepmind-ai-lea... | | > _Luis Piloto at DeepMind and his colleagues have created an AI | called Physics Learning through Auto-encoding and Tracking | Objects (PLATO) that is designed to understand that the physical | world is composed of objects that follow basic physical laws. // | The researchers trained PLATO to identify objects and their | interactions by using simulated videos of objects moving as we | would expect [...] They also gave PLATO data showing exactly | which pixels in every frame belonged to each object. To test | PLATO's ability to understand five physical concepts such as | persistence..., solidity and unchangingness..., the researchers | used another series of simulated videos. Some showed objects | obeying the laws of physics, while others depicted nonsensical | actions_ [with the latter, correctly the AI returned wrong | predictions, showing an acquired intuition of physics] | | From the submitted one: | | > _[Jeff Clune, Uni British Columbia, Vancouver: ]<<[Comparing AI | with how human infants learn is] an important research direction. | That said, the paper does hand-design much of the prior knowledge | that gives these AI models their advantage>>. // Clune and other | researchers are working on approaches in which the program | develops its own algorithms for understanding the physical world_ | Jeff_Brown wrote: | Data and analysis alone, without experimentation, don't seem like | enough to achieve real intelligence. From its title this article | sounded like it would be about progress in learning by doing. | Alas, it's not. | mdp2021 wrote: | > _experimentation_ | | Right. "You have to tell yourself stories", as the late Prof. | Patrick Winston said (you are intelligent because you can | predict the unexperienced). Because you need concept | development and critical thinking - an active process. | thomasjudge wrote: | I feel like there is an inferential leap implied, to greatly | simplify, from "A does X and B does X" to "A and B must operate | relevantly similarly." For example, walking and flying are both | modes of transportation, but you can't really learn anything | interesting about one from studying the other | sebzim4500 wrote: | >For example, walking and flying are both modes of | transportation, but you can't really learn anything interesting | about one from studying the other | | You can figure out newtonian mechanics entirely on the ground | and that clearly helps you understand flight. By analogy, | getting more understanding about what limits exist in ANNs | could plausibly help understand how the brain works (and vice | versa). | ChikkaChiChi wrote: | Once you understand that walking gets you from where you are to | where you want to be, you start to define the characteristics | of motion. Then, seeing other forms of conveyance that are | faster underpin the concept of efficiency. | | Walking and flying may not have a lot in common to you and I, | but to a thing learning to crawl, there is a lot to be | understood. | jonbaer wrote: | Looking at the photo I don't think the AI is going to realize | eating the piece it is about to pick up and it will choke. I | would actually like to see more reinforcement learning agents | like that, the action space on infant movement is quite small so | it's even really about "action space discovery" to some point. | Things it discovers are way more interesting, like if food were | not on the floor/level and it has to stand to get it, it will | eventually get there after N attempts (over time), and then if | you introduce another agent if learning to block the other agent | will award more food, etc, then it discovers 50/50 and | equilibriums (better to eat now than wait). PLATO seems like a | step in that direction. | danielmorozoff wrote: | Similar work has been pursued for a number of years now in a | Darpa program called Machine Common Sense: | https://www.darpa.mil/news-events/2018-10-11 | | I recall Tenenbaum's lab had a similar paper a few years back. | mdp2021 wrote: | Also https://www.machinecommonsense.com/ , | | in which animations are shown which reveal the close similarity | to the DeepMind project. | mikolajw wrote: | Clickbait title. | | I wish ML researchers stopped using anthropomorphizing language. | This has decades of solid tradition, but that's no excuse. Any | comparison of a machine to a human misleads the public. Machines | aren't like babies, artificial neural networks aren't like actual | neural networks or brains. Machines shouldn't be given human | names (PLATO is a borderline case). | | I know this is like talking to a wall -- money requires hype -- | but still, please stop doing that. | mdp2021 wrote: | A further article and a commentary just appeared on The | Conversation from a Professor of Psychology and Infant Studies: | | https://theconversation.com/researchers-trained-an-ai-model-... | | > _Typically, AI models start with a blank slate and are trained | on data with many different examples, from which the model | constructs knowledge. But research on infants suggests this is | not what babies do. Instead of building knowledge from scratch, | infants start with some principled expectations about objects | [...] The exciting finding by Piloto and colleagues is that a | deep-learning AI system modelled on what babies do, outperforms a | system that begins with a blank slate and tries to learn based on | experience alone_ ___________________________________________________________________ (page generated 2022-07-11 23:00 UTC)