AI Defeats Anti-AI Test

so I'm glad we can stick with CAPTA. Perhaps the most interesting part of that name is the Turing test. Named after a computer scientist and codebreaker Alan Turing, the Turing test has evolved into an interesting concept. The Turing test of today is a scenario in which a human interrogator uses some sort of interface, often a computer terminal, to communicate with someone else. That someone else might be a human being, or it might be a computer program.

If enough human interrogators are fooled into thinking the computer program is actually a person, that program is said to have passed the Turing test. The program has appeared to be indistinguishable from a human being to a significant percentage of interrogators. In the case of captures, the Turing test is a sort of shorthand for a gateway that is meant to allow humans to pass through while stopping artificially

intelligent constructs at the border. The general idea behind the captured test is that it should be something relatively simple for a human to complete, but very challenging for a program to do. This is important because you don't want to capture to be challenging for humans to complete, since that would discourage people from filling out the form in the first place. It doesn't do you any good if

you prevent anyone from going through the gate. There are some things that robots and software can do much better than humans, such as drive a car under normal conditions or play a game of chess. But there are some things that we humans can do pretty easily that AI has a really tough time handling. At least that's how things stand now. AI is getting better all the time, and we humans have seemed to stalled out all the

little bit. Evolutionarily speaking, a simple capture example that you've likely seen would be a series of letters and numbers that are deformed or semi obscured by other shapes. As humans, we can recognize which shapes represent letters or numbers and which are just distractions, But computers have a lot harder time separating the signal from the noise. Enter the AI research firm Vicarious, a company that has received funding from

such tech luminaries as Mark Zuckerberg and Jeff Bezos. Vicarious created an artificial neural network that the firm claims can solve capture puzzles. That's pretty impressive considering some of the trickier puzzles can stomp human beings. Google's Recapture has an eighty seven percent solution rate among humans. The neural network can suss out the shape of letters and numbers while ignoring all that distracting noise. A neural network mimics the

way our brains work. Instead of neurons the nerve cells and our brains, the network uses network computer systems. The artificial neurons attack problems in layers, with each neuron and effectively tackling a different part of the problem, and then combining all of those solutions together to come up with an overall solution to the caption. These networks aren't just automatically good at completing tasks. You have to actually teach them how to do it, just as you would a

human being. Back in two thousand twelve, folks across the Internet reacted in glee at the news that a network of sixteen thousand computers had learned how to recognize cats by examining ten million images culled from YouTube videos. What could be more representative of the modern online experience, But that example illustrates how complicated this process is. It took thousands of computers working together to learn how to identify a cat based off a sample size of ten million images.

So while it is possible to teach computers to do things like recognize cats or captures, it's not easy. Humans are still better at this than machines are at learning. Unfortunately, captured tests aren't really about learning, or else we would still have a leg up on the robotic competition. As AI has become better at recognizing captures, designers have created more difficult puzzles that obviously impacts both humans and non humans,

according to researchers with Vicarious. Since Vicarious researchers sounds weird, even with new and improved captures, their software can solve the puzzles sixty six point six percent of the time. With Google's recapture system fifty seven point four percent of the time for Yahoo's captures and fifty seven point one percent at a time with Paypals captures. According to security experts, this means we can expect to see hackers make use

of those same techniques within a few months. That will mean the capture will cease to be an effective gatekeeper to keep bots out. We may see more creative approaches. For example, some sites use a short written passage followed by a simple question about the reading material to test for human nests. Eventually, AI and neural networks may be good enough to navigate those systems that we humans are used to without any trouble at all. At that point, we may need to re examine how we protect sites

and services from abuse. Or maybe by then we'll have merged with the machines and entered the singularity. To learn more about artificial intelligence and how machines and human intelligence may one day merge into one. Subscribe to the tech Stuff podcast. We also talk about other texts like catapults, sort of like a grab bag of awesome tech topics. The show publishes on Wednesdays and Fridays, and it's a deep dive into all things tell I'll see you again soon.