Concerns over the artificial intelligence’s ability to spread disinformation may be overblown, according to one expert.
“I think it’s a little overhyped,” Richard Socher, chief scientist at software firm Salesforce, told CNBC in an interview during his trip to Singapore.
That’s because humans are already adept at creating fake news without the help of algorithms, Socher said. Furthermore, fake news usually has some sort of “agenda” behind it — something that AI inherently lacks, he added.
AI can already be used to create fake, superimposed images onto another person in videos. There have been rising concerns that such videos, also known as “deepfakes,” could be used to spread misinformation.
Socher, an expert in natural language processing in the field of machine-learning, said current attempts at AI-generated text are often “not as coherent” when compared to humans who would write fake news.
“I honestly think that people are much better at creating fake text news … than AI will be for a long time,” Socher said.
“There are already people who look at being able to classify pictures or text and see if they were created by … an algorithm or not,” he said. “As with most things in security, there will be measures and counter-measures and an ongoing sort of race, but of all the concerns I have, this is not as high on my priority list.”
Instead, Socher spoke about the potential bias in AI as well as its possible impact on jobs — topics he said were currently at the forefront of discussions about AI ethics.
“Now that AI really works, we need to think about the impact that it has on people,” he said.
“In some ways, AI holds a mirror in front of us,” he said, adding that the algorithm essentially processes data generated from human actions before distilling them and automating processes.
As a result, it is important to create AI algorithms that make “unbiased decisions that don’t discriminate,” he said.
On the subject of AI’s potential impact on jobs, Socher said the view that algorithms would make all jobs redundant overnight was “unrealistic.”
“What I think we will see is like, there’s certain kinds of jobs that have more potential for automation, and there’re others you know that will have less,” he said. “As you work through that spectrum, you basically have to think about all the people that are impacted by it. And then I think it’s a combination of both companies, as well as governments, to try to help those people.”
Asked about where he though AI was headed, Socher said he was “very optimistic” in the long-term but less so in the short-term, acknowledging the challenges and concerns stated above.
To illustrate this, Socher said that in the early days of the internet, “zero experts” predicted that there would be social media, or even a Twitter marketing manager position.
“Nobody could predict that from the early internet, and AI will have an even bigger impact on productivity, and so on, than the internet,” he said.
“A hundred years from now, I have no worries that people will find new kinds of jobs, in like travel and … fun extreme sports and doing YouTube videos,” Socher said.
