[ad_1]
In recent years, many of the world's leading technology companies – from Google to Facebook and Microsoft – have become interested in artificial intelligence and how to integrate it into almost all of their products. For example, Google even renamed its Google search division as Google AI before this year's developer conference, during which AI was in the spotlight. Mark Zuckerberg also explained how Facebook was using artificial intelligence to try to suppress hate speech on its platform during its F8 conference in May.
The AI market is also booming as companies continue to invest in cognitive software capabilities. According to the International Data Corporation, global spending on AI systems is expected to reach $ 77.6 billion by 2022, more than triple the $ 24 billion projected for 2018.
But the industry still has a long way to go and much of its progress could depend on the ability of academics and industry players to find ways to equip computer algorithms with learning capabilities similar to those of humans. Systems based on artificial intelligence, that is, algorithms used by Facebook to detect inappropriate content or virtual assistants from Google or Amazon that power the smart speakers of your home, still can not deduce the context in the same way as humans. Such a breakthrough could be essential for Facebook, which is stepping up its efforts to detect online harassment and identify terrorism-related content on its platforms.
"There are some very obvious cases, and the AI can be used to filter these or at least to allow moderators to decide," said Yann LeCun, chief scientist of AI for Facebook AI Research in a recent interview with Business Insider. "But there are a lot of cases where something is hate speech, but there's no easy way to detect it unless you have a broader context …" this, the current AI technology is just not there yet. "
A key element to advancing the field of artificial intelligence, especially with regard to in-depth learning, will be to make sure that there is a hardware capable of handling it. . This is one of the main topics discussed by LeCun Monday at the International Conference on Semiconductor Circuits, where he discusses a new research paper describing the main trends that will have to be taken into account by the sellers of chips and researchers over the next five to ten years. "No matter what they build, it will influence AI's progress over the next decade," he said.
Before the conference, LeCun spoke to Business Insider about the direction of the field of artificial intelligence, what it could mean for the devices we use in everyday life, the state of artificial intelligence today. and the biggest challenges ahead. Below are the key points to remember from our conversation.
The energy consumption of the machines needs to be significantly improved so that the AI can improve.
Imagine a vacuum cleaner smart enough to map your living room so that it does not clean the same place twice, but also able to detect obstacles before hitting them. Or an intelligent lawn mower that can intelligently avoid flower beds and branches by mowing your lawn. For gadgets like these to work and become prevalent – in addition to technologies in which companies like Facebook and Google Alphabet, such as augmented reality and autonomous cars – LeCun says that a less energy-hungry hardware is needed . Such an advance is not only necessary for such technologies to flourish, but also to improve the way companies like Facebook identify the content of photos and videos in real time. Understanding what's happening in a video, transcribing that activity into text, then translating that text into another language so that people around the world can understand it in real time requires a "huge" amount of computing power, says LeCun.
We will continue to see the progress of AI in smartphones in the short term before improvements appear elsewhere.
Over the next three years, LeCun thinks that most smartphones will integrate artificial intelligence directly into hardware through a dedicated processor, making features such as real-time speech translation more prevalent on phones. This is probably not a surprise for those who have been keenly interested in the smartphone industry in recent years, as companies such as Apple, Google and Huawei are increasingly integrating AI into their mobile devices, which, according to LeCun, will allow "all kinds of new applications".
Giving machines "common sense" will be a major goal of AI research over the next decade.
While humans often discover the world through general observations, computers are usually trained to perform specific tasks. If you want to design an algorithm that can detect cats in photos, for example, you need to help understand what a cat looks like by exposing it to a mine of data, which could contain thousands of photos labeled as including cats. . But, according to LeCun, the Holy Grail of the next decade to advance AI is to perfect a technique known as self-supervised learning. In other words, allow machines to better understand how the world works with the help of data rather than simply learning to solve a particular problem – such as identifying cats.
"If we actually train [algorithms] To do this, the ability of machines to capture context and make more complex decisions will grow significantly, "said LeCun, who added that this technique currently only works reliably for text, but not for videos and breakthroughs could be what companies like Facebook need to improve the moderation of content on their platforms.While nothing knows when this solution will come, LeCun said: is not something that will happen tomorrow. "
[ad_2]
Source link