Fatma Tarlaci, PhD is a machine learning scientist and trainer Open Teams. Her work focuses on Natural Language Processing (NLP), AI, and open source software. Her remarks on AI To Uplift Humanity expanded on the next generation of natural language processing.
NLP in AI
Most people’s familiarity around NLP is what they know from Google Assistant, Siri and Alexa. But the next generation of NLP will be focusing on common sense and emotion. “We still are not very good at understanding common sense, our emotion detection in textual or even speech data.” Dr. Tarlaci explained. “Human language is very complex. And it's not only the textual data that we are looking at, which most of the applications have been doing up until now. We need to incorporate voice, we need to incorporate vision, and only then we will have a better understanding.”
Computer vision will also assist NLP in the near future. “A human will interpret body language and tone in someone's voice and make assumptions and some judgments about what to say next.” Tarlaci said. “When we incorporate these different dimensions and domains [in AI] then we will get better at understanding human language and communicating in generating it correctly.”
Is Open Source Needed To Make AI Transparent?
One concern in AI development is the lack of transparency. So, is an open source approach the solution? “Not necessarily or not fully, but I think it makes it slightly better because then we have multiple stakeholders and a lot of people are have great intentions,” Tarlaci said. “If an open-source project is becoming very useful and being used by millions of people in their AI, [or] different AI applications, then there is a chance that someone is going to notice that something is not right or it's not going well, at least it's not behind closed walls…we can do things to understand and explain the model if it is open-sourced. But it goes beyond open source.
Having an open source approach to the training is important as well. “If you don't have access to the actual training that is still a problem.” The industry is making good strides however. “The recent focus on explainability and responsible AI, I think we will have some good steps towards [these] very powerful tools that we are using in almost every application in the world today.”
Tarlaci advocated for many domain experts to be at the table in development of AI in the future. Teachers, lawyers, and those who aren’t familiar with the nuance must get involved. “The people who are experts in the field should create those platforms and domains,” Tarlaci said. People moving AI forward should be the ones to create platforms to solicit input and make the industry more transparent.
Fatma Tarlaci, PhD Resources
Connect with Dr. Tarlaci on LinkedIn here.
Learn more about Open Teams here.