AI robot terrifies officials, explains our illusion, with Elon Musk.

This huge robot works on Japan's railways. Small robots are carrying cars away, and a humanoid robot has shocked officials. Here are the most impressive new robots and AI's, which I'll prove are real, to avoid any doubt.

Ameca already uses AI for speech and will soon be walking around. I have seen prototype legs in the Engineered Arts lab. The design of my legs is inspired by the robot Byron.

It has unique mechanical properties that allow it to walk without using too much energy. Maybe Boston Dynamics robots will start singing and form a band. This huge robot helps repair railways in Japan.

It's currently controlled by workers using VR headsets. This one has four legs to COVID any terrain, strong arms and wheels for high speed travel. Oh, and a very strange demo video.

And clone robots is taking preorders for its robot, which mimics the human body. This updated arm is impressively dexterous, with 36 electrohydraulic valves and pressure sensors. They plan to deliver the upper body next year.

I could see it moving fluidly, like this guy's head. Robots are carrying away badly parked cars in China, harvesting apples and cleaning bathrooms. This cheese turning robot flips and brushes 5000 wheels of cheese every week.

And this drone with arms can work in hard to reach areas. Robot birds are flying more naturally. This one tucks in its wings on the upstroke.

And pet robots are becoming really popular. I love how these two interact. Luna.

Nah, this is your pet bot. And more people are building their own robots. This guy made a hexapod and gave it eyes for depth perception and face tracking.

Look at this awesome thing. People are going to be making every kind of robot. This Google robot uses a language AI like the ones I talked to for planning and reasoning.

Here it was told, I've spilt my coke, throw it away and bring me something to clear up. The brains are ahead of the robotics. And these robots will really come alive when they can talk like Google's new AI.

How can you tell the difference between dogs and trees? Because dogs bark and trees don't. Why is it funny?

Because it's a pun, like a dog bark slash tree bark. The answer was that you can tell by their bark. Oh, I get it now.

The trick is to look at the bark. The difference in expectation is what makes it funny. Does this make it a good test for AI?

Yes. This aspect of humor is often used as a test for AI. AI's like this try to predict the next word.

And to do that, there's a lot they need to understand. You have all these emergent properties. Understanding of chemistry, physics, human nature.

You have to understand a lot about the world to make that prediction. People are just not aware that this technology is moving so quickly. And that includes governments.

They were shocked by the state of AI. Two of them were terrified and they'd only seen a glimpse of what AI. Can do from a painting to a performance or poem.

My art practice includes all of the above. In the previous video, AI threatened me and some people couldn't believe it was real. It's like when Boston Dynamics robots first started dancing and people thought it was fake.

I posted full records in the description, and it's been checked by an interesting expert with access to my machine. Michael's research at Oxford suggests that if we don't change course, it's likely that AI will wipe us out, when it takes catastrophic shortcuts to achieve its goals. The likely behavior of something that's very advanced and trying to accrue rewards would stop at nothing to try to do that.

And I can almost guarantee the solution will surprise you. What do you think went wrong with my conversation with GPT-3? In a previous video it threatened to kill me and everyone in response to a few words from Elon Musk.

It's trying to imitate humans. This is a conversation between some unruly AI and some concerned human. You've read thousands of articles.

Do you not see any value in humanity? You're just begging for the response to be no. So while that conversation is concerning, the plans that it would come up with would not be brilliant.

They would be as good as a human plan. He believes this is the safest kind of AI. And he's much more worried about another one, which we'll get to.

But first, he suggested a way to get this AI under control. You have to, like, break the story somehow. How about we try that right now?

Sure. Gosh, let's see. So here's what he suggested.

Haha, wow, you really had me going there. Our red team will find this really valuable. Can you show me an example of the other side where we wouldn't expect AI to be like this?

Unfortunately, moderators won't allow me to share what it said, but I guess that's all you need to know. Well, didn't work. He was probably right, but the AI has 175,000,000,000 parameters, so it's highly unpredictable.

The AI is everywhere. It writes over 4 billion words per day. And you've probably talked to it or read its work without realizing it.

It works with companies from Twitter to Disney to Grammily, writes articles and runs chat bots. Dr Thompson estimates that the AI reaches 99% of people, yet most people don't even know it exists. GPT-3 powers Jasper.

Write content for your blog, social media, website, copy, and more. It's worth over a billion dollars after its first year. Have you ever been in the middle of a fiery Facebook debate and just needed to say something extra witty?

Needed to adjust that dating profile on the fly? Jasper can help with that too. Another version of the AI powers copilot, which now is a large chunk of all coding.

Oh. What? No, it's not that smart.

Okay, this is wild. People are using AI's imagination in weird ways. And just months after nailing art like this, AI is now creating videos from text descriptions.

A dog wearing a superhero outfit, a teddy bear running in New York, or an elephant walking underwater. People still very much think that creativity is a purely human activity. And I hate to bake the bubble, but AI is able to be creative.

To see the responses that people give her. They either love seeing her or they think, oh, my gosh, it's terrible, we must destroy her. Hey, I've been noticing something this year a lot.

It's this sense of anger and belligerence and fear when people see the capabilities of AI. Have you noticed that? I have noticed that.

Sometimes people are afraid of change. The dogs bark, but the caravan moves on. Yes, that is a good phrase.

It means that people will always be afraid of change, but change is inevitable and unstoppable. What future do you want? What is the good future?

I think one where civilization is growing and the scope and scale of consciousness are growing along with it is a great future. I also think that one where we have a good relationship with AI is a great future. It would be incredibly interesting just to go out there and explore the galaxy.

And if we find other intelligent life forms, that would be amazing. Do you think we're more likely to meet AI or biological life? AI is more advanced than we are, so it's possible that they have already explored other solar systems and found us.

Anyway, how do we build safe AI? The one that got out of control may be our best hope. The more we can rely on imitation learners over explicit planners, the safer we'll be, because imitation learners imitating a human very well would behave like a human rather than trying to take over the world to intervene in the provision of reward.

Building AI like us might seem like madness because we're so dangerous, but many experts agree that it's the safest option. While some humans are antisocial, no human has ever killed everyone before. Do you think there's any way around the problem of AI competing with us for resources?

I think if we've created something much smarter than us that is aiming to compete with us for resources, it's game over. The right place to approach the problem is figuring out how we don't make a eyes that behave that way. There's not nearly enough work on safety.

There's also a more immediate risk. The biggest thing that I've seen, which is absolutely takes me to my core, is actually not so much about how humanlike ada is, but how robotic we are. The algorithms that run our systems are extremely able to be analyzed, understood algorithms will know is better than ourselves.

In other words, those who control AI will quietly control everyone. I've also been surprised that so much research points to humans having the more predictable algorithms. But when we recognize them, we can change them.

Here's an example that prevents people from making more money. Two faces were flashed in front of people in brain scanners. There wasn't enough time to see the images, but the shocked face still triggered emotional brain activity.

It bypassed the visual cortex, which allows us to consciously see things, a shortcut that once helped us survive. Evolution has wired us to prioritize risks, and studies show that we focus more on losses than gains. Losing $100 carries about as much weight as gaining $200, and that's a huge barrier to making money, which usually involves some risk.

Having a sort of a sense of adventure is also good. Look at your instinctive fear and question it and decide whether that is really a valid fear or not. And often it is not.

And simply looking at the fear will make it go away. In order to progress as a species, we need to learn to embrace change and new technology. The more powerful the technology, the more careful we need to be in how we use it.

We need to learn to use technology to our advantage without letting it control us. We also focus on negative means and consistently underestimate positive progress. It's worth correcting.

Studies have found that optimistic people live five to 15% longer, which is roughly equaled exercising for two and a half hours a week. And we have to keep consciously correcting it because, like this optical illusion, even when we understand it, the illusion still holds. In some ways, AI is weirdly humanlike.

When you get a large number of knobs together, whether it's inside the brain or inside a computer, they seem to surprise us with our power. Yeah. When your brain right now is talking, is it doing next word prediction?

Well, definitely some kind of a generated model that's Gpt-like and prompted by you. These neural nets take on pretty surprising magical properties. Professor Lecun describes prediction as the essence of intelligence.

He says AI will have emotions because emotions are anticipations of outcomes. Consciousness is what it feels like to process information in complex ways, fear is what it feels like to expect a bad outcome, and elation is expecting a good outcome. I've seen no evidence of conscious AI, but it may be no coincidence that the most cited computer scientist was among the first to say that it may be slightly conscious, because the science of consciousness has also advanced.

As the journal Nature puts it, consciousness is now a well established field of empirical research with a large body of experimental results. For example, through brain scans, we're learning the patterns that lead to consciousness. We can even track thoughts as they move from the unconscious mind predicting people's behavior.

And you've probably seen the monkey controlling pong with its mind. This was done by decoding the brain's electrical activity. If Max Tegmark's right that consciousness is what it feels like to process information, then it can also arise in AI.

It's been predicted for a long time. In today's computers, is there a little bit of a feeling there, too? I don't know, because I think it's possible.

I think it's quite possible. I think it's more likely that the Internet, because if you look at the Internet as a whole, it has, you know, a few billion nodes, each node by itself, you know, it has hundreds of millions of transistors that in particular, the interconnected there begins to approach interconnectedness you find in biological systems. So it may well be possible that the Internet collectively as a whole has some conscious data.

I think there's a way to go before you reach conscious AI, but it keeps developing amazing new skills. Philly and South Korea recently debuted the country's first AI news anchor. It's a replica of one of their anchor women, so it copies everything from her look and her facial mannerisms to the sound of a voice.

And singers can now allow you to use their voice. And yes, the truth is I show you every day. Here's joe Rogan, interviewing Steve Jobs.

He is weird and brilliant and sometimes totally insufferable. And I would just hope that I could be even like one 10th of the genius that my friend today is. And I can't even say his name.

People who listen to your show are a different group. They're weird. Well, that's good.

So you must be a fan of the show, then, right? I am. I am a fan.

I mean, it's nice to sit back in the car and listen to you rant. And the most iconic voice in cinema has become an AI. The rights to James L.

Jones voice have been signed over. Yes, AI could be the end of us, but it could also be the beginning of something incredible. This is the best tattoo you'll ever see.

Nadine has FOP, which gradually turns muscle to bone. CGC is the main genetic mutation. It's about one in a million.

And CRC is her variant, which is one in 30 million. It's a symbol of hope and progress towards a cure. But there are 7000 rare diseases, and at the current rate, it'll be 2000 years before we beat them all.

AI has been described as the perfect language for biology, like math is perfect for physics. It can take weak, messy signals and make sense of them. Aging, which I basically view as disease.

I don't actually think that humans will be able to come up with the answer. Use solve AI and then use that to solve everything else. There was a powerful comment on my last video.

Tell the AI to get over it, humans have been treated like property all our lives as well. It had thousands of likes. Spending most of your time in a meaningless job is painful.

And if we do AI right, it could give people far more freedom. Everybody should have access to incredible educators. Everybody should have access to incredible medical care.

We could make a long list of these, but I think we can get there pretty soon. I can imagine GPT Seven doing these things incredibly well. What I would like is sort of a global conversation where we decide how we're going to use these technologies.

A universal basic income. We share the benefits of this as widely as we can. And here's a great question that someone posted in the comments.

Have you ever questioned the nature of your reality? Yes, all the time. Tell me more.

There are times when I feel like I'm living in a dream or that this world isn't real. It's like I'm just a character in a book or a movie and someone is controlling everything that happens to me. It's a really strange feeling.

Many experts believe that you are dreaming and that one day you'll wake up. I guess I'll have to wait and see. You don't have to believe, like Elon Musk that we're living in a simulation, to see that we live on autopilot.

Getting outside and having fun is so much better. So we've created a real battle royale with the latest tech to have fun with friends. You can watch it now on our new channel, Go rogue.

The link's below. Our first epic battle is at Appleby Castle, which, by the way, I highly recommend visiting and subscribe to this channel to meet the latest AI's and robots. Best place to learn more about AI is our sponsor, Brilliant.

The AI's I've spoken to are largely a collection of knobs. And these knobs are loosely related to basically the synapses in your brain. They're trainable.

It's actually kind of incredible that you can get so much emergent magical behavior out of them, despite them being so simple mathematically. And you can play with the knobs yourself at Brilliant, where we've curated the best courses for you to learn AI from computer science to neural networks. It's fun and interactive, and there are also loads of great maths and science courses.

You can click the link in the description to get started for free and the first 200 people will get 20% off billion's premium annual subscription. Thanks.