And I think that's behind a lot of the things that people are worried about. These simulations that I've run say that things can get Much better in terms of GDP.
training, what do they actually do? I mean, are they out there physically shooting targets together? Are these computer simulations ? Do we know what they look like? Yeah, majority, most of it is computer simulation, you know, war gaming. This is, you know, most of it is done inside
So that's what makes a supercomputer. Our simulations look so much like the real universe that there's gotta be something in the universe that is like what we simulate.
Without an open mouth, you're doomed to failure. These simulations are games without the scoring.
So these are much more expensive simulations . These simulations cost 100,000 to a million CPU hours per galaxy. These simulations cost 10 or 20 million CPU hours, and you get tens of thousands of galaxies.
These simulations cost 100,000 to a million CPU hours per galaxy. These simulations cost 10 or 20 million CPU hours, and you get tens of thousands of galaxies. But at much lower resolution.
display a free plane on this SLM and create a focal point here. We're doing exactly the same thing here, and then we use computer simulations to calculate what kind of image to create in the sound field and create the same image. So what we're doing is light and sound. I think half of you programmers write programs, so it's like the
relatives as I helped them kick off what's now 50 challenger centers around the globe that are kind of space science education facilities where they have shuttle simulations that students go through and groups like the Nature Conservancy just happens to be a charity that I particularly fond of that I also partnered with in my mission that I'll tech to talk about but for many years even though I was backing all this and
with the way that agents are working and tool use, it started to become really useful for, still early days of it, but genuinely useful in Either robotics or simulations .
And I'm going to show you two more simulations , right? Two sets of simulations . In one, AI is just a continuation of the historical patterns of automation that we've seen.
Okay, here's the growth rates over time. So the simulations I'm showing you are kind of holding constant the measurement, and that would be one reason why things could be a little faster.
I have two event horizons. And in the simulations you can see a bobble and they merge together and they make one bigger black hole. And then it radiates in the gravitational waves.
- That, actually, we're able to do now. You can run simulations of the formation of planetary system. So if you run the simulation, really, where you wanna start is a cloud of gas, these giant interstellar clouds of gas that may have,
everything that goes near it gets sucked in. can do advanced hyperrealistic simulations . And if you think about things like I know virtual reality at the moment or video games and you
everything that goes near it gets sucked in. in hyperrealistic simulations . Theory number two, or possibility number two, is that advanced civilizations do exist,
everything that goes near it gets sucked in. probabilities not on simulations .
You've been running simulations on a war with Iran. Yep, every strategy for 20 years and it's playing out right now. So, I can tell you that we are losing control of
Yeah. What is the from everything you know 30 30 plus years studying this stuff? Iran running simulations on Iran, advising the White House, being a master and probably arguably the most informed person in the United States right now about air attacks like the one the US performing on Iran. What is the headline that you're trying to send to the world
We have a few different planes, a couple of different environments. And these simulations can be set by the trainer to put the trainee or the de-icer who needs recertification in for any scenario that they choose with all of these different variables.
on the intricacy of your particular respiratory system. Using supercomputer simulations based on scans of your lungs, we can predict with high precision where particles will flow and then design devices that can deposit drugs exactly where your body needs them.
and the computational effort of some of those so that you can keep the whole show on the road to implement the actionable decisions you referred to, because we really need to get these simulations to make predictions ahead of time, not after the event, so they can inform those clinical decisions. Yeah. Yeah. In fact, Matt, you kind of, just to jump in there, a lot of medicine's done with post hoc rationalizations
with all these simulations in mindwandering.
Why do we run simulations ?
You guys are just simulations .
We can run simulations or small neural networks on big computers.
had been invented then, whereby we could-- the story I've told you-- we could calculate it in a big computer. Is that where simulations and simulating universes and galaxies come into play. That's where they came into play, yes.
So that's what makes a supercomputer. So there are many simulations .
And even did some simulations of that big weight that was hanging to see if it would swing in weird ways and rack into things, and made a few modifications
We conducted mission simulations in which we faked the clocks out on the bird, in flight, in the middle of nowhere between Uranus and Neptune.
We had 40 mission simulations .
They had to use simulations for a lot of stuff.
They run simulations in computer. This area has grown significantly, and we call them computational astronomers.
They can do simulations , games-- whatever it might be.
and we were all shocked. But our simulations look a lot like these images that we're now getting from Hubble, so we're beginning to figure it out. And one of the ways that we're hoping to improve our understanding is using deep learning technology.
We don't get to see how galaxies evolve because they evolve on a time scale of hundreds of thousands of years to hundreds of millions or even billions of years. Cosmological dark matter simulations show large-scale structure and the properties of the lumps of dark matter that we call halos, and then we can try to put galaxies into those.
I get tens of millions of CPU hours from Google and-- from NASA Ames and other places. So this is some simulations that we don't like so much anymore because it turns out that the galaxies really are elongated and they don't have as many stars as these simulations did at the same redshift.
They just slowly grow in size, but fade out. In our simulations we have some slow track, we have some fast track. These are organized the same way as the picture.
But actually when they start galaxies look like pickles. And our simulations are showing exactly that same phenomenon. There's the dark matter.
They built these simulations , down to the way muscles jiggle.
can get-- continuous simulations and test these builds along the way.
In other words, I'm using waves transmitted through space itself to create images, shapes, and tools that are beneficial to humans. When you do that, the results of computer simulations and things like projectors that are output become detached from human perception, so it feels like the data itself takes shape and starts moving along with various things. I'm constantly thinking about how such things can relate to existing 3D
and of new simulations .
They're not simulations . They are the real phenomena.
And it's in simulations where you learn what you really want to do.
running these simulations . What this allows scientists to do is carry out science and ask questions more quickly and get feedback faster than they
to run molecular dynamics simulations .
And one of the simulations that I've done is take a decade-- the first decade of the 2000s, which
we can run simulations of models.
You wanna have interactive simulations .
is a role that simulations will play. So, yeah?