Chinese scientists claim breakthrough in quantum computing race

[Signifyin(g)Monkey]

...so basically I think what we're seeing is the initial stages of the construction of competing supply chains for the next stage of the computational arms race.

 

Quote

Google said last year it has built a computer that could perform a computation in 200 seconds that would take the fastest supercomputers about 10,000 years, reaching quantum supremacy. The Chinese researchers claim their new prototype is able to process 10 billion times faster than Google’s prototype, according to the Xinhua report.

 

Quantum computing is going to be truly nuts.  And if there has to be a Cold War 2.0, let us at least get insanely powerful computers out of it.

[SilentWorld]

How long until I can buy it on Newegg? 

[Jason]

4 minutes ago, SilentWorld said:

How long until I can buy it on Newegg? 

 

Tomorrow, but UPS won't deliver it until sometime next year.

[TwinIon]

Google's computer from last year used 53 qubits and this Chinese one uses 76. So a good improvement, but I think the exponential nature of qubits makes that improvement sound like a bigger deal than it is.

 

Of course, we're still pretty far away from making these things do anything particularly useful.

[Jason]

1 minute ago, TwinIon said:

Google's computer from last year used 53 qubits and this Chinese one uses 76. So a good improvement, but I think the exponential nature of qubits makes that improvement sound like a bigger deal than it is.

 

Smaller?

[CitizenVectron]

Hoping each time that the next computation would be the computation home.

[TwinIon]

2 minutes ago, Jason said:

Smaller?

I just mean that I think the difference between 53 and 76 is a better indication of the leap being made than "10 billion times faster."

[mclumber1]

9 minutes ago, Jason said:

 

Tomorrow, but UPS won't deliver it until sometime next year.

[Commissar SFLUFAN]

That's great and all, but how does it run Crysis?

[Chadatog]

9 minutes ago, Commissar SFLUFAN said:

That's great and all, but how does it run Crysis?

Like shit

[johnny]

How many computations until everybody can get healthcare 

[vaxick]

How many seconds will this shave off of a Windows update? 

[Spawn_of_Apathy]

On 12/4/2020 at 5:50 PM, Commissar SFLUFAN said:

That's great and all, but how does it run Crysis?

you can almost set off a chain explosion at upwards of 24fps. 

[Remarkableriots]

We have to stop @legend from downloading himself into that computer and taking over the world!

[legend]

7 hours ago, Remarkableriots said:

We have to stop @legend from downloading himself into that computer and taking over the world!

 

If it makes you feel better, quantum computing isn't really useful for AI. At least nothing has been proposed that indicates as much.

 

I'll stick to downloading myself into conventional super computers for now.

[CitizenVectron]

2 hours ago, legend said:

 

If it makes you feel better, quantum computing isn't really useful for AI. At least nothing has been proposed that indicates as much.

 

I'll stick to downloading myself into conventional super computers for now.

 

Could quantum chips be useful components of future computer/AI designs, though? As in, when something that the quantum computer excels at is needed (as determined by the normal computer), it offloads that calculation to the quantum chip, and then receives the result? From what I understand, quantum computers are really great a a niche area of brute-force computing, but are not flexible. From my (very limited) extrapolation of that, could they not be used as components that specialize in a greater, more flexible overall design? 

[legend]

17 minutes ago, CitizenVectron said:

 

Could quantum chips be useful components of future computer/AI designs, though? As in, when something that the quantum computer excels at is needed (as determined by the normal computer), it offloads that calculation to the quantum chip, and then receives the result? From what I understand, quantum computers are really great a a niche area of brute-force computing, but are not flexible. From my (very limited) extrapolation of that, could they not be used as components that specialize in a greater, more flexible overall design? 

 

There are not really any clear applications for even sub parts right now, or at least I've never seen one proposed that made me think "oh that could be useful." Part of the issue is QC doesn't even make intractable problems tractable (by the CS definition), it just makes them non-trivially faster, but still intractable. That's why the way you make symmetric encryption robust to the existence of QC is just by doubling the key size and calling it a day. (My understanding is the way asymmetric encryption works this is not as straightforward, but there are still other approaches out there that would be. To be clear though, I'm not an expert in encryption)

 

Bearing that in mind, now note that basically every problem in AI is fundamentally intractable and the only way we make progress is by figuring out useful approximations. 

 

Maybe at some point in the future someone will find an interesting connection, but if I had to pick investigations into new kinds of computing architectures for AI, I'd explore other non QC architectures first.

[legend]

@CitizenVectron I was just looking at this and thought you might find it interesting: Tensorflow with quantum computing:

 

Google AI Blog: Announcing TensorFlow Quantum: An Open Source Library for Quantum Machine Learning

AI.GOOGLEBLOG.COM

---

Posted by Alan Ho, Product Lead and Masoud Mohseni, Technical Lead, Google Research “Nature isn’t classical, damnit, so if you want to m...

Import thing to note here though is this is largely geared toward layering ML on top of quantum systems and algorithms, not for using quantum mechanics to solve fundamental AI problems.

[Air_Delivery]

IMO "AI" will never exist without some kind of biological component. It is not possible for the technology as it is no matter how powerful to become intelligent in a way a biological entity is.

 

Machine learning or machine pattern recognition is much more accurate terminology. No doubt it can do things far beyond the capability of any human but its scope is extremely narrow and prone to catastrophic failure if anything unexpected is thrown its way. It will never have the ability to adapt to an unexpected and changing environment like a human. For example "AI" could never have brought back the astronauts from Apollo 13.

[legend]

9 hours ago, Air_Delivery said:

IMO "AI" will never exist without some kind of biological component. It is not possible for the technology as it is no matter how powerful to become intelligent in a way a biological entity is.

 

Machine learning or machine pattern recognition is much more accurate terminology. No doubt it can do things far beyond the capability of any human but its scope is extremely narrow and prone to catastrophic failure if anything unexpected is thrown its way. It will never have the ability to adapt to an unexpected and changing environment like a human. For example "AI" could never have brought back the astronauts from Apollo 13.

 

Computers are Turing complete. Theoretically there is nothing stopping them from having human-like intelligence. To explain that a bit, this means there's nothing a biological architecture can compute (equivalent to behavior) that a conventional computer system could not. The practical concern is whether computer hardware is effective for it from a performance stand point. I suspect it's not -- as it is we are already designing better hardware that is specific for making AI operations faster (or require less energy, or better memory handling, etc.). But there's also a good chance the biological hardware of people isn't ideal either; it's simply what evolution had to work with.

[CitizenVectron]

59 minutes ago, legend said:

 

Computers are Turing complete. Theoretically there is nothing stopping them from having human-like intelligence. To explain that a bit, this means there's nothing a biological architecture can compute (equivalent to behavior) that a conventional computer system could not. The practical concern is whether computer hardware is effective for it from a performance stand point. I suspect it's not -- as it is we are already designing better hardware that is specific for making AI operations faster (or require less energy, or better memory handling, etc.). But there's also a good chance the biological hardware of people isn't ideal either; it's simply what evolution had to work with.

 

Right, as I understand it, if you could theoretically just build a human brain (in the same shape, layout, etc) with electronic-equivalent components, it could function the same way. So if that is the case, then obviously it would make sense that a machine could be sentient in the same way with different designs that are more efficient. Evolution is a fantastic best-route finder, but it's not perfect.

[legend]

5 minutes ago, CitizenVectron said:

 

Right, as I understand it, if you could theoretically just build a human brain (in the same shape, layout, etc) with electronic-equivalent components, it could function the same way. So if that is the case, then obviously it would make sense that a machine could be sentient in the same way with different designs that are more efficient. Evolution is a fantastic best-route finder, but it's not perfect.

 

You can do something even dumber. You can, in principle, just simulate a human brain on an existing computer (as in not in replace with electronic components) if you had a computer with enough memory (you don't  ). But it would run impossibly slow. We could probably do better even with simulation since a lot of the physics of the brain is mostly unimportant and doesn't need to be directly modeled, but it would still be far too slow even with that.

 

We also don't know enough about the workings of the brain to properly simulate it even if we wanted to try.

 

 

The big open question is whether we can abstract out enough of what the brain is doing such that something similar to a conventional computer could still do it, or if we really do need radically different hardware architectures to run it efficiently.

 

 

I will also add that few of us actually want to replicate a human brain. We already know how to make humans, no thanks, don't want that  What we want is the intellectual capability (ideally better) without all the other cognitive cruft that comes with humans.

[b_m_b_m_b_m]

51 minutes ago, legend said:

We already know how to make humans, no thanks, don't want that  

This is the opposite of Chuck Yeager

[legend]

1 hour ago, b_m_b_m_b_m said:

This is the opposite of Chuck Yeager

 

I know who Chuck Yeager is, but clearly not enough to understand this relationship  

[b_m_b_m_b_m]

3 minutes ago, legend said:

 

I know who Chuck Yeager is, but clearly not enough to understand this relationship  

this guy fucks

 

 

Joke sounded better in my mind

[Signifyin(g)Monkey]

America strikes back!

 

Quote

A viable quantum internet — a network in which information stored in qubits is shared over long distances through entanglement — would transform the fields of data storage, precision sensing and computing, ushering in a new era of communication.

 

This month, scientists at Fermilab, a U.S. Department of Energy Office of Science national laboratory, and their partners took a significant step in the direction of realizing a quantum internet.