DARPA's Latest Chip Is Designed To Be Bad At Arithmetic (technologyreview.com)

Reader holy_calamity writes: Pentagon research agency DARPA has funded the creation of a chip incapable of correct arithmetic, in the hope of making computers better at understanding the real world. A chip that can't guarantee that every calculation is perfect can still get good results on many problems but needs fewer circuits and burns less energy, says Joseph Bates, cofounder and CEO of Singular Computing. The S1 chip can process noisy data like video very efficiently because it doesn't need the extra circuits or operations needed to ensure every mathematical operation is performed perfectly. This summer DARPA will put five prototype computers, each equipped with 16 of the inexact S1 chips, online for researchers to experiment with.

Comments

Intel already tried something similar

[marius.muja]

https://en.wikipedia.org/wiki/...

Re:Intel already tried something similar

[Pfhorrest]

Daisy, daisy
Give me your answer do
I'm half crazy
Can't divide three by two
All my answers
I can't see 'em
They are stuck
In my Pentium
I could be sweet
My answers fleet
With a workable FPU

-David Pogue, probably

Re:Intel already tried something similar

[Locke2005]

At Intel, quality is job 0.9999999998!

Simple Solution

[cogeek]

Just program the chips to use Common Core math

Re:Simple Solution

[werepants]

I know whining about common core is a popular pastime among people who have an irrational fear of change, but have you ever taken more than a few seconds to actually look at how common core teaches things? It has much more in common with real mental arithmetic than the standard method we all learned in grade school, and is very intuitive if you actually take a moment to understand it.

It's funny, because most of the complaints I've seen cherry-pick examples to intentionally make common core look more complex, but gloss over the convoluted aspects of the standard method (75+22 makes standard method look obvious, but 99 + 99 has many more steps because of the carrying). I for one (and probably most people) never actually carry numbers mentally - in the previous example I would add 100 + 100 and then subtract two, or some other shortcut that fits with human cognition rather than optimizing to be easy to write on a whiteboard.

Re:Simple Solution

[cogeek]

Thanks for jumping to conclusions, but I happen to have a BS in Civil Engineering, so I've done my share of math. I also have 4 children currently ranging from HS to College level and all have experienced Common Core to varying degrees. I did my homework when Common Core was first being discussed and all the way through it being shoved down my childrens' and their teachers' throats, have you? It was approved by one individual to the outcry of every single other person on the review board. It's complete and utter rubbish. Anytime you'd like to sit down and solve a complex math problem using your Common Core vs my usage of Common Sense, please let me know. I'll even give you a 10 minute head start.

Re:Simple Solution

[MightyYar]

The problem is that there is a fundamental disconnect between the people who devised these new methods of instruction and the materials and curricula which use the new methods. There is nothing wrong with teachers arming themselves with several different ways to teach a particular math concept. Some gets won't "get it" using one method, and so you have these other two which you can draw upon.

The problem is that they force the kids to learn all three methods. They then practice all three methods in the same time space that used to be allotted for one method, more or less. In addition to confusing kids who would have "gotten it" the first time, it significantly reduces total practice time for whatever method they do eventually settle on. I think that the teacher should introduce a single method of their choosing, and then use their discretion to apply the other methods. Then each child should practice their chosen/assigned method until it is mastered.

Either that or give kids 3x the math instruction so that they can learn all of the new methods, which I think is an insane choice.

I think you'll find teachers, on the whole, a bit dissatisfied with the new approaches. So it's not just us engineer types who had no trouble learning the older style math.

Re:Simple Solution

[yodleboy]

"I happen to have a BS in Civil Engineering"

You do understand that your math skills are probably far greater than the majority of parents sending kids to public school right? Your kids are fortunate enough to have a parent that really gets math and can help them understand. Let's face it, not every kid is going to need to do complex math when they grow up. Common Core seems to be very good at teaching kids practical math that they can do mentally. Something the majority of adults I run into seem to have a lot of trouble with. I have two kids in elementary school, and both of them regularly impress with their math skills. I'm OK at math and my wife is actually good at it, so sometimes we cringe when we see their homework assignments, but the facts are, they are learning math, they can apply math even at this age, and they are not intimidated to try more difficult math later on. I always wonder how many kids with actual aptitude have been turned off at an early age because of math taught in traditional ways.

Bet you this is the key to real AI

[gurps_npc]

As in, in order to get a real AI, it will need to have this fuzzy logic.

Which by the way will end up making our new Robotic overlords require human slaves to do math for them.

Which we will do incorrectly, causing their entire robotic empire to fall in a matter of hours.

Re:Bet you this is the key to real AI

[shawn2772]

As in, in order to get a real AI, it will need to have this fuzzy logic.

Which by the way will end up making our new Robotic overlords require human slaves to do math for them.

Which we will do incorrectly, causing their entire robotic empire to fall in a matter of hours.

I've thought this for a long time (well, not the humorous bits), that it's entirely possible that general intelligence fundamentally requires fuzziness and imprecision, and that by the time we succeed at creating really smart artificial intelligences, we'll find that they're just as error prone and fallible as people are. Which isn't to say we can't make them smarter than we are or that they won't be incredibly useful.

I also wonder if we will find we need to make systems that are deliberately crippled for many of the tasks we want them to do. I mean, imagine installing Marvin the Paranoid Android's brain in a car and requiring it to spend all of its time driving people around. It would likely soon drive off a bridge just to end the boredom. Yeah, Marvin is a fictional character, but it's believable to me that true artificial intelligences will get bored, distracted, make mistakes, etc., just like people, making them perhaps not much better than people at many of the moderately mindless tasks we'd like them to take on. So we'll have to limit them to make them good at what we want them to do... which may make them not so good at what we want them to do.

I think we may also have to deal with the equivalent of mental illness in AIs, and be unable to fully diagnose and/or fix the problems because the system is complex enough to be opaque to us, the same way we don't (and may never) fully understand our own brains and their malfunctions.

Intelligences general and flexible enough to do anything may do everything somewhat badly, and systems sufficiently specialized to do a task extremely well may be unable to cope with the unexpected. Or not. The next few decades are going to be very interesting.

Re:Don't see how this should help

[pla]

Yes, I guess you can make arithmetics slightly faster when you allow errors, but is that where todays CPUs spend a lot of time?

Not so much in error checking, but in the choice of the algorithm itself.

As an example, Quake 3 famously used a crazy-fast inverse square root routine. It didn't give an exact answer, but rather, one "close enough" to suit its intended purpose (calculating surface normals for reflections) in software, in a quarter of the time FPUs of the era could get an answer using dedicated hardware. The FPU would always give a much more accurate answer, but not every use needs a much more accurate answer.

the question on everyone's mind:

[Gravis Zero]

Does it run Linus... cause you know, close enough.

Isn't Visual Basic non-deterministic enough?

[dsmatthews9379]

Why spend all that money on research when Microsoft already had the perfect product for their needs.

Re:An irrational fear of change...

[cogeek]

You should have provided her a piece of paper and a pencil so she could draw out 1,212 little circles and then cross off 687 of them, then count up the remaining circles. Would be a system she's likely to be much more familiar with, unfortunately.

Remember:

[Locke2005]

At Intel, quality is job number 0.99998643!

Re:An irrational fear of change...

[Anonymous Coward]

For example, I tried to give this young woman at Panda Express 12 dollars and 12 cents, because the bill was 6 dollars and 87 cents, so that I could get a $5 bill and one quarter back from the transaction so I wouldn't have to carry around so many separate bills or extra coins, and she looked apoplectic.

That's funny, because the fast food restaurants I've visited tend to have this thing called a "cash register" where the employee keys in the *exact* amount that customer wants to pay with. She wouldn't have had to do any math whatsoever if you had handed her the amount you claim you did.

What probably happened was that you handed her a ten, she posted the transaction, and like a true jackass you said, "Oh wait I got change," and started counting out the remaining $2.12 in nickels, dimes, pennies, and atm receipts. She saw the line forming behind you, rolled her eyes, and your brain registered that as "bitches can't do math."

But yeah, I feel you: some people have a hard time with change.

Re:An irrational fear of change...

[radarskiy]

"so I wouldn't have to carry around so many separate bills or extra coins"

Sounds like you couldn't cope with change, either.