As new artificial intelligence (AI) platforms and applications become available, businesses must leverage their powerful capabilities and keep pace with their competitors. From marketing to sales to customer experience, business users can take advantage of highly personalized messaging when communicating with both existing and potential customers. However, it is not without concerns. The rise of misinformation campaigns, inaccurate data, bias, and cybersecurity risks are forcing the business community to take a hard second look at AI.
Hosted by the Center for Wireless and Information Technology (CEWIT) at Stony Brook University, the virtual panel discussion “AI for Business Optimization” explored the accelerating AI landscape with a diverse group of experts. Christine Gilbert, Assistant Professor, Department of Marine and Atmospheric Sciences (SoMAS), Department of Communication and Journalism, Alan Alda Center for Communication Sciences, moderated the discussion.
Panelists were asked how they are implementing AI in their businesses. Jeehye Yun, his CEO and founder of the Baltimore-based text analytics company RedShred, said his company has been working on features like generating copy for press releases, white his papers, and simple marketing applications. explained how they are using the AI of While this has been very successful, Yun stressed that AI output still requires human editing and checking for correctness.
“When AI fails, it’s because of the kind of technical content we’re asking it to do,” Yun said. “In the case of aircraft maintenance, for example, the information must be accurate. There is no room for error. These procedures must be followed according to the letter.”
Pepe Valiente, head of strategy for IBM Software Support, issued a similar warning.
“AI can solve problems, but its failures can grow exponentially,” he said. “Everything starts with data, so it’s important to have good data to work with and have the opportunity not only to test the model, but also the time to train it. , you need to see if there are any anomalies in the models you own, and you also need a governance layer that helps you control what is going on.”
Cole Ingraham, Principal Data Science Instructor at the NYC Data Science Academy, said of its vast capabilities, it’s important to know exactly what users want out of AI.
“Some specific occupations are better prepared for AI training than others,” he said. “We have to get over the assumption that AI can do ‘everything’ because it can’t. We have to know how to approach it. Some people use it to try to increase their revenue. but technologists like software engineers and data scientists might try to understand how these things work fundamentally, how they can be improved and how they can be monitored. I can’t.”
Messenger technology reporter Benjamin Powers tackled the challenges of data bias and misinformation from a journalism perspective.
“A big reason this is happening is that people can now use it as easily as they want,” he says. “So it’s not just the one he’s or the two he’s doing this. There’s a model that’s spreading very quickly and data that’s not as carefully constructed as the companies that have been working on this for years. You can train in. There are a lot of different challenges.”
Powers added that companies such as Apple and Samsung have banned employees from using generative AI tools because they could inadvertently compromise sensitive data.
“These models can take information, learn from it, and spew that information out to someone else who has absolutely nothing to do with the company itself,” he said. “So we need to be very aware of what we enter and who has access to these models and the data used. And this will lead to other issues related to privacy.”
All panelists agreed that despite the amazing advances AI can potentially enable, it is important for users to understand and respect its limitations.
“AI doesn’t really understand what it’s reading,” says Yun. “Like humans, we don’t understand what it’s producing. It’s a ‘next word’ predictor.” They’re built on layers of models trying to understand what happens next, but there’s no superintelligence behind them right now. So does it really know what it’s doing? The answer is no. Because they are trained on material written by humans. “
“It’s not only a technical problem, it’s also a human problem,” Valiente added. “How can we make good things and regulate them? How do we reconcile everything that’s going on around us? Ultimately, this is what happens one way or another, but we get to the core of what we can do with all this power. We have to find a way to humanize this to go back.”
“When you think about AI in terms of risks and benefits, AI is like an entertainment venue that mirrors ourselves,” says Powers. “It works smart for the specific tasks it can perform, but it has no sentience. Because we are human, it has advantages and many problems.We are not infallible.”
— Robert Enproto
