The explosive successes of AI in the last decade or so are typically chalked up to lots of data and lots of computing power. But benchmarks also play a crucial role in driving progress—tests that researchers can pit their AI against to see how advanced it is. For example, ImageNet, a public data set of 14 million images, sets a target for image recognition. MNIST did the same for handwriting recognition and GLUE (General Language Understanding Evaluation) for natural-language processing, leading to breakthrough language models like GPT-3.

A fixed target soon gets overtaken. ImageNet is being updated and GLUE has been replaced by SuperGLUE, a set of harder linguistic tasks. Still, sooner or later researchers will report that their AI has reached superhuman levels, outperforming people in this or that challenge. And that’s a problem if we want benchmarks to keep driving progress.

So Facebook is releasing a new kind of test that pits AIs against humans who do their best to trip them up. Called Dynabench, the test will be as hard as people choose to make it.

Benchmarks can be very misleading, says Douwe Kiela at Facebook AI Research, who led the team behind the tool. Focusing too much on benchmarks can mean losing sight of wider goals. The test can become the task.

“You end up with a system that is better at the test than humans are but not better at the overall task,” he says. “It’s very deceiving, because it makes it look like we’re much further than we actually are.”

Kiela thinks that’s a particular problem with NLP right now. A language model like GPT-3 appears intelligent because it is so good at mimicking language. But it is hard to say how much these systems actually understand.

Think about trying to measure human intelligence, he says. You can give people IQ tests, but that doesn’t tell you if they really grasp a subject. To do that you need to talk to them, ask questions.

Dynabench does something similar, using people to interrogate AIs. Released online today, it invites people to go to the website and quiz the models behind it. For example, you could give a language model a Wikipedia page and then ask it questions, scoring its answers.

In some ways, the idea is similar to the way people are playing with GPT-3 already, testing its limits, or the way chatbots are evaluated for the Loebner Prize, a contest where bots try to pass as human. But with Dynabench, failures that surface during testing will automatically be fed back into future models, making them better all the time.

For now Dynabench will focus on language models because they are one of the easiest kinds of AI for humans to interact with. “Everybody speaks a language,” says Kiela. “You don’t need any real knowledge of how to break these models.”

But the approach should work for other types of neural network too, such as speech or image recognition systems. You’d just need a way for people to upload their own images—or have them draw things—to test it, says Kiela: “The long-term vision for this is to open it up so that anyone can spin up their own model and start collecting their own data.”

“We want to convince the AI community that there’s a better way to measure progress,” he adds. “Hopefully, it will result in faster progress and a better understanding of why machine-learning models still fail.” 

Source link

قالب وردپرس

If we had to make a guess at the single piece of electronic bench equipment owned by the highest proportion of Hackaday readers, it would not be a budget oscilloscope from Rigol, nor would it be a popular portable soldering iron like the TS100. Instead we’re guessing that it’s a multimeter, and not even the most accomplished one.

The DT830 is a genericised Chinese-manufactured 3.5 digit digital multimeter that can be had for an astonishingly low price. Less than a decent hamburger gets you an instantly recognisable plastic case with a chunky rotary range selector switch, and maybe a socket for some kind of transistor or component tester. Make sure that there is a 9 volt battery installed, plug in the pair of test leads, and you’re in business for almost any day-to-day electrical or electronic measurement. They’ve been available in one form or another for decades and have been the subject of innumerable give-aways and loss-leader offers, so it’s a reasonsble guess that you’ll have one somewhere. I have three as far as I know, they make great on-the-go instruments and have proved themselves surprisingly reliable for what they are.

Persuading You Is Going To Be A Tough Sell

If you talk about the DT830 in polite company, you might be greeted with snorts of derision. It’s not difficult to find reviews that tear one down and compare it to a more expensive meter, and not surprisingly find the pricey meter to be of higher quality.

And it’s certainly true that for a couple of dollars, you get a switch that won’t last forever and high voltage isolation that maybe isn’t quite up to spec. But I’m going to advance a different take on the DT830 that may surprise some of you: to me it’s a modern classic, an instrument that provides performance for its price that is nothing short of phenomenal. Because that pocket-money meter not only measures voltage, current, and resistance, it does so accurately and repeatably, and to compare that with what might have gone before is to show just much better a device it is.

Thirty years ago, a digital multimeter was an expensive item, and most multimeters were still analogue. A cheap multimeter was therefore invariably a small pocket analogue device, and the very cheap ones could be astoundingly awful. Accuracy and repeatability in reading wasn’t their strong point, and while I am a great fan of analogue multimeters when it comes to spotting dips and trends in tweaking analogue circuitry, even I can’t find reason to praise the inexpensive ones. By comparison the DT830 delivers reliable and accurate readings with a high-impedance input, something I would have given a lot for in 1985.

That Performance Is No Fluke

So given that it costs considerably less than a pint of beer in a British pub, how does such a cheap instrument do it? The answer is, by standing on the shoulders of giants. My colleague Anool Mahidharia supplied the answer here back in 2017 when he took a look at the Intersil 71XX series of integrated circuits; the archetypal DT830 contains an ICM 7106 3.5 digit digital panel meter chip, whose roots lie in a much more exclusive stratum of the industry.

(Despite there being a load of newer and more accomplished multimeter chips on the market I was surprised to find that none of them had found their way into the meters I’d opened.)

The ICM 7106 was based on work Intersil did in 1977 to produce the part in Fluke’s first portable DMM, the model 8020A.

So you’re not getting anywhere near the physical design or component quality of that expensive meter, but you are benefiting from the tech that made its ancestor a very good instrument for the 1970s. The dual-slope integrating ADC and precision reference are the same as the ones in many far more expensive meters, which is what makes the reading from your few-dollar DT830 one you can trust. Not bad for something you might dismiss as a piece of junk!

If there is something to be gleaned from this story, it is a very real demonstration of the power of semiconductor manufacturing. Assuming it has passed acceptable factory QA, every 7106 is as good as any other 7016, from the first one made by Intersil in the 1970s through to the unknown-origin chip hiding under an epoxy blob in my cheap meter. The manufacturer can skimp on every other component in the meter, but assuming that there’s no money in counterfeiting a 43-year-old chip that long ago left its premium product phase behind and has been manufactured by many sources over the years, they can’t skimp on the chip that powers it. To be an ICM7106, it must have the same features as the original from the 1970s, thus my bargain-basement meter still shares something that matters with one of far higher quality.

The DT830 multimeter, then. It may be a heap of junk, but it’s an astonishingly good heap of junk. I for one, salute it.

Source link

قالب وردپرس

Call of Duty Mobile is a mammoth battle royale game with millions of players worldwide. Like every other mobile game, most Call of Duty: Mobile players enjoy the game using the touch controls. However, what most players don’t know is that they can also play Call of Duty Mobile with a controller on both Android and iOS devices.

Activision announced the controller support for Call of Duty Mobile back in November 2019. However, most players are still unaware of this feature. So, if you want to play Call of Duty Mobile with a controller but don’t know how to connect it with the game, then you’re in luck. In this article, we’re going to guide you through the entire process to connect a controller with Call of Duty Mobile.

Supported Call of Duty Mobile Controllers

As of now, the official Dualshock 4 PS4 controller, except the first generation, and the official Xbox One controllers can be used to play Call of Duty Mobile. However, the good thing is that over time, Call of Duty: Mobile will bring support for new controllers as well.

If, somehow, you manage to connect a non-supported controller with Call of Duty Mobile, you might still run into problems while playing the game. So, we highly recommend that you only use one of the supported controllers.

How To Connect A Controller With Call Of Duty Mobile?

That’s it; now you’ll be able to play Call of Duty Mobile with a controller in both Battle Royale and Multiplayer.

Note: If you leave your phone idle for a long time, then the controller might get disconnected from your phone, so you’ll have to connect it again.

COD Mobile Controller Support: Everything You Need To Know

When you play Call of Duty Mobile with a controller, you will only get matched against players using the controllers. Moreover, if you’re playing in a squad where one of your squad members is playing with a controller, even then, you’ll have to go against players using the controller support.

It’s important to note that you can’t use COD Mobile controllers to explore the in-game menu. So, if you want to change the weapons loadout and customize your character before entering the match, then you’ll have to do that using game’s native controls.

Also, make sure you enable the controller before entering the match. If you forget to do that before entering the pre-game lobby, the controller won’t work in the entire game.

The post How To Play Call Of Duty Mobile With A Controller? appeared first on Fossbytes.

Source link

قالب وردپرس