What are the latest developments in AI Robotics? - with Mike Wooldridge
ฝัง
- เผยแพร่เมื่อ 21 ก.ย. 2024
- Why are robots so much more complicated to develop than AI software? Find out with 2023 Christmas Lecturer Mike Wooldridge and the Oxford Robotics Lab.
Find out more about the 2023 Christmas Lectures here: www.rigb.org/c...
Michael Wooldridge is a Professor of Computer Science at the University of Oxford. He has been an AI researcher for more than 30 years, and has published more than 400 scientific articles on the subject. He is a Fellow of the Association for Computing Machinery (ACM), a Fellow of the Association for the Advancement of AI (AAAI), and a Fellow of the European Association for AI (EurAI).
Join this channel to get access to perks:
/ @theroyalinstitution
Subscribe for regular science videos: bit.ly/RiSubscRibe
The Ri is on Twitter: / ri_science
and Facebook: / royalinstitution
and TikTok: / ri_science
Listen to the Ri podcast: podcasters.spo...
Donate to the RI and help us bring you more lectures: www.rigb.org/s...
Our editorial policy: www.rigb.org/e...
Subscribe for the latest science videos: bit.ly/RiNewsle...
Product links on this page may be affiliate links which means it won't cost you any extra but we may earn a small commission if you decide to purchase through the link. - วิทยาศาสตร์และเทคโนโลยี
To me, the most impressive example of AI in robotics is the development in Ukraine of self-learning AI embedded into prosthetic limbs, so that the limb learns to interpret the motor nerve impulses of the patient and gradually improves with use, and even gives feedback through the same sensory nerves which previously provided it, allowing the patient to self-teach the robotic prosthesis to move with exactly the right force and speed as they require.
The reason we see thousands of robots used by industry and very few for domestic applications is simple - money. Industry has a vested interest in automation as it reduces cost and improves accuracy and repeatability, but not many households are willing or able to stump up the sort of money it takes to develop general-purpose domestic help robots.
The Royal Institute...
Bringing you the latest information from the previous decade.
Kek!
To the comms folks and/or producers of this video: good job! It's no small thing to make academics and their research pursuits approachable and appealing.
This had about the same insight as the snack aisle at a grocery store.
Summary: In this video, Michael Wuldridge, a professor of computer science at the University of Oxford, discusses artificial intelligence (AI) and its application in robotics. He explains that while AI is often associated with robots, progress in robotic AI has been slower compared to text-based AI. The Oxford Robotics Institute is working on state-of-the-art AI robotics, including designing robots for various applications such as industry, inspection, medical, and safe human interaction. The speaker highlights the challenges of robotic AI, including dealing with noise, sensors, and interactions with the environment. They express optimism about the progress being made but acknowledge that there is still much work to be done.
Key themes:
1. The challenges of robotic AI compared to textual AI, including noise, sensors, and interactions with the environment.
2. The advancements and applications of AI robotics at the Oxford Robotics Institute.
3. The need for further scientific advancements in robotic AI.
This lab is a fan of Really Useful Boxes. I approve.
How good/fast is virtual embodiment (accurate physical simulation of imperfect sensors and actuators) these days? Surely it then comes down to massive deep training of the kind used so successfully with the LLMs.
00:13 what questions were you asking chatgpt, Mike?
Very basic interview does not even scratch the surface
Now I understand better the huge differences between pre-programmed robots and teaching robots to think for themselves. Not sure I can live with that without some important legislation being put in place to 😮protect human beings😮
1:10 NO. "AI" isn't any "harder" in the physical world. It's only hard because what you have IS NOT "AI".
No mention of real world simulations for training robots at the speed of a computer? There are now examples of actual robots trained entirely in virtual environments that cope quite well on their first encounter with the real world.
It should be noted that Artificial Intelligence has nothing to do with human intelligence. It in no way mimics the way that humans learn or think. AI does some things, well enough for them to be used in certain cases via statistical modeling (ie: language translation), but humans don’t operate this way. This is why AI language translations, for example, will never 100% match human translations. This is compounded by the fact that currently, AI runs on classical computers, while the human brain leverages quantum mechanics to process information.
Are you certain it works differently? How do you know humans don't use statistical modelling? Also that is a little bit like saying its just physics or just math. What evidence do you have that quantum mechanics has a relevant role in how information is processed by the human brain? As far as I can tell it is all at the electrical level, see electrical probe stimulation, transcranial magnetic stimulation and inhibitors and inducers working etc.
"Never"
What you're missing, imo, is that AI will be combined with QC. Once QC is practically on a wide scale, it's game over for A LOT of things that 20thC minds are used to. And they're not "evil" things, despite what your granny or Gramps will tell you.
@@PhoenixProdLLCYou are assuming AI will even work with Quantum computers. At best thats a hypothesis that it can even work.
The UK is living in the architecture, infrastructure and psychology of the past while the London Financial Sector empowers foreign innovation and Laogai manufacturing?
Erm I don't believe this I mean look at Boston Dynamics In what they show us and what they don't show us via the military side.
Much of what Boston Dynamics shows is not necessarily AI robotics, but rather robots following pre-programmed movements. Extremely sophisticated and robust, but completely different from AI robots than can, for example, learn and adapt in real-time to situations (including its own damage) that it has never been explicitly programmed for.
Why would you even NEED to see it if YOU'RE not in the military?😂
@@PhoenixProdLLC Think about the answer to the question you just asked.
90% of that BD robots do is scripted, not AI-driven. Change a couple of environmental parameters (make the door handle a different shape, make the floor slightly slanted or irregular, etc.) and they fail spectacularly.
I wonder how soon there will be androids like star trek's data, with super strength! Thank you for sharing.
Give or take around 10 years. If the progress doubles each year
😂About 100 years...sw/hw/fw don't all evolve at the same pace
@@redditchrc960That's what has been said over and over since the early 1970's. 50 years and still dangling that carrot.
@@williambrandondavis6897 Were getting closer
As soon as the laws of physics stop applying.
Demos shown do not look as if we made much progress at all. Tesla robot currently appears to be at the apex in AI robotics.
And the operative word there is "APPEARS".
These are real robots, being worked on by engineers to achieve useful things. Not puppets being guided through little animated set pieces by a team of students.
@@JohnBailey39🤣
Tesla is a joke. Even Boston Dynamics robots (which are light years ahead of anything Tesla does) use a lot of scripting.
In the future, there will be robots.
What depth of insight!
What? So, there's still no plausible substitute for my laziness?
There was a time when humans spent LESS TIME WORKING, even in a day, then the modern world does. WAY less. FACT.
This is one of the best TH-cam channels...but people is not playing attention. Human kind is taking wrong roads
It feels like you made this video 20 years ago. For example, Tesla has over 5 million robotic cars on the road right now which use pure vision (from 8 or 9 video cameras, depending on model age) and neural nets to be able to see the 3D environment and drive around the road systems. Likewise, Tesla is also developing a humanoid robot using the same sensors and neural nets. After only 2 years of development, the robot is already many years advanced from anything you show here and is well on its way to solving Artificial General Intelligence. Please catch up with the times.
Im the first to like it, I liked the topic
I predict that in about 30 years or less we will be able to build a life sized Johnny 5 who can do literally any and all jobs that a Human can do, except create good art.
Will they be able to paint technically perfect paintings? Yes. Will they be as emotionally resonating? No.
Even if they are beautiful a work without an entity behind it to express something is empty
@@stephenr80 which is why art made by artificial intelligence won't compare to art made by Humans or other biological life forms.
Also I believe that an artificial intelligence can't have hopes, dreams, ambitious, or goals, except for what is given to it by a Human.
Why th would anyone create a robot to have it pretend to be an artist? 😂 Silly. Ask that money and time is to do things like mining and building and business and accounting and a lot of tedious admin work. Not to do our dancing and singing and expression and dreaming FOR us! 😂
@@PhoenixProdLLC We already have apps that we can download onto our phones, where you can tell it a few key words and have it make a picture for us.
@@PhoenixProdLLC I expect the art/marketing/advertising departments of many corporations would be happy to have creatives who work super fast, 24/7, for the price of a few kWh of electricity!
🤖 😮❕
Lol, If I want to know about the latest advancements in AI and robotics I would listen to people making those advancements, not the Royal Institute. What a joke.
unneeded, distracting, annoying background music
makes this video unwatchable.
boston dynamics
Физиков и философов придётся поставить в угол и выпороть хорошенько, а почему? Естественный интеллект обогнал развитие Человека в обществе людей и уже создаёт искусственный интеллект, а развитие общества людей отстало от естественного интеллекта, что может вести к окончанию цикла без начала нового цикла. Только в обществе людей человек может стать Личностью и обладая полномочиями от общества контролировать механизмы структур в том числе государственных с их кровотоком финансов. Что же произойдёт при незрелом обществе людей так и не научившихся держать под контролем естественный интеллект, перекладывающих основы Человека ответственность на механизмы искусственного интеллекта? В чём же причина незрелости? Философия как основа Человека и общества людей отсутствует напрочь, осталась демагогия " натянуть сову на глобус". В интересное время находимся в материальных процессах, естественный отбор с причинно следственными связями и цикличностью в иерархии, но, пришедшие в сей процесс " созданные по образу и подобию" так и не создавшие зрелого общества способного перейти из одного большого цикла в следующий отдав всё на волю естественного интеллекта или всё таки дозреем до разумности взять под контроль стяжательство и прочую пребуху материальных процессов.
I just don't get it. Robots and software have been around of decades. In the years that I've been around, very little real progress has been made. This may sound a bit offensive but the engineers working on these are just stupid. There is maybe 1 exception, the Atlas robot from Boston Dynamics.
What a waste of time. This woman knows nothing
Robotics is harder than text based AI.
. . .I'm not saying he's wrong, I AM saying it would be nice if he would justify the claim.
There is a reason robotics is harder than language, and 'embodiment' is not the right angle to view it from.
What would you say is the right angle to view it from?
It's definitely one of the two main angles. The other is that there's a ton of easily digestible material to train text-based AIs on, but not nearly as many physically-accurate simulated environments to train robots in, which means they need to acquire data through their own sensors, and that is not only more complex (and more expensive) but also takes much longer.
If you believe replacing humans with machines to do work then those humans need to be paid accordingly. Regardless of how much the machine costs you can't replace the wage to the human so that's where the solution must come from. Not from the AI. Peace 😎 ✌️
Machine looms replaced humans. Farming machinery replaced farm labourers. Mine workers lost all their jobs in the UK. Cheaper overseas labour replaced much heavy industry in the West. Car factories replaced many people with robotic machinery. All call centre people will be replaced with AI probably within our lifetimes.
@Safetytrousers then there's no reason left for you to be around. Is there? Just asking. Careful when replacing human effort for cost certainty. Just saying. Read the club of Rome and the 300 then you will understand. Peace 😎 ✌️
Wwhy is there only something about the “AI” scam here?
Wow putting blocks on top of each other, please tell me this isn’t funded by the tax payer. They clearly haven’t progressed in 40 years, pathetic.
Show us what you have programmed?…… pathetic
@@Orcrez probably something that’s taken your job, so apologies
At some point in your life you were just stacking blocks too. Probably shitting your diaper at the same time!
@@Orcrezpeople are ALLOWED to criticize and question things and in public without being shut down by some randobot on the Internet. Gtfu
The thing about research is that it can be tricky to see what is hard about a problem from the outside. I Googled Professor Maiolino's work and it looks really interesting and important. When you see a robot just moving objects around it is actually using a sophisticated soft-body gripping control system with fine motor sensors. Engineers have been able to manually program rigid claws to do very specific motor operations for years, but this research is about robotics that can be adaptable, precise, and delicate. It's an enormously more challenging engineering problem than the robots on assembly lines. Solving this has ramifications for medical robots, and much more.
Please actually look into a topic before you say things like this.
00:13 what questions were you asking chatgpt, Mike?