All the sessions from Transform 2021 are available on-demand now. Watch now.
Edge AI is becoming increasingly important with the internet of things (IoT) boom, but the security challenges make it more difficult for companies to adopt edge computing. As the size of the IoT ecosystem has grown, so have the risks.
Edge AI has an advantage over cloud-based AI, McAfee chief data scientist Celeste Fralick told VentureBeat executive editor Fahmida Rashid during a session of VentureBeat’s Transform 2021 summit. Backhauling large amounts of data collected from various sensors up to the cloud, processing it, and then bringing it back down is both costly and time-consuming.
In the case of a farmer out in a field where there are no wireless networks, getting online to access cloud applications and process the data from the sensor is not going to work well. But edge AI runs locally, so the same farmer can access answers right from the field. Edge AI also avoids some of the privacy concerns associated with cloud — which is particularly important in fields like health care.
“The pros of the cloud, of course, are the cost, the reliability, the ability to merge a lot of information very quickly,” Fralick said. “But you also have the cons of cloud, where you have the issue of your data privacy and security not as perfect as you’d like it to be with some companies.”
There is a specific downside to edge AI, Fralick said, noting that IoT devices — specifically edge devices — tend to be brittle. If the devices have a security flaw, the manufacturer is the one that has to repair it.
“The risk at the edge is it’s much more brittle and much more complex than it would be in the cloud. So your security risk is much higher,” Fralick said.
Because of the nature of edge AI and the IoT, it makes sense to keep your routers secure. “If you’re a consumer, you always have to ensure that you’re updating your software,” Fralick said. “Certainly, if you’re an enterprise, business, you have to update your software constantly. That’s one of the most important things that we can do. And as a home consumer, ensuring that you have security on your router as well is very important.”
Data management and AI
People need to realize that there are different types of AI. “[You have] simple statistics and machine learning, you have deep learning, and you certainly have neural language processing,” Fralick said. At large conferences, Fralick often approached other booths to ask what kind of AI was being used in their product or how models were being trained and rarely received satisfactory answers. If someone at the booth tried to claim 100% accuracy, Fralick knew they did not really understand AI.
“Ensure you have monitors throughout your development process to get to as close to 100% accuracy as you can. But understanding why you have false positives and false negatives and understanding the root cause of why you have those are definitely the most important,” Fralick said.
VentureBeat’s mission is to be a digital town square for technical decision-makers to gain knowledge about transformative technology and transact.
Our site delivers essential information on data technologies and strategies to guide you as you lead your organizations. We invite you to become a member of our community, to access:
- up-to-date information on the subjects of interest to you
- our newsletters
- gated thought-leader content and discounted access to our prized events, such as Transform 2021: Learn More
- networking features, and more
Author: Zachariah Chou