Federal Tech Talk

This is not your father’s artificial intelligence

This week on Federal Tech Talk, Arash Rahnama, head of Applied AI Research at Modzy, joins host John Gilroy to talk about artificial intelligence, data security...

John Gilroy

November 9, 2020 10:19 am

1 min read

Best listening experience is on Chrome, Firefox or Safari. Subscribe to Fed Tech Talk’s audio interviews on Apple Podcasts or PodcastOne.

This week on Federal Tech Talk, host John Gilroy spoke with Arash Rahnama, head of Applied AI Research at Modzy. They discussed artificial intelligence and federal IT.

AI has been around for a while. With the advent of cheap cloud storage and computers, the concept of artificial intelligence has matured to a point where we can refine the discussion. Rahnama brings several ideas to the table: Poisoning data, explainability, and adversarial defense.

Today’s attackers will try to slightly alter the data set you use for training your artificial intelligence application. For example, a malicious actor could alter a stop sign to cause an accident with autonomous vehicles. This seemingly innocuous attack can have major implications in the intelligence community

A lowly data scientist may have to explain her decision to her boss, in which case Rahnama suggests you remember what happened in high school math — you must show your work. Be able to explain how you drew that conclusion.

A federal IT professional should expect that outsiders are attacking the data set used for federal artificial intelligence. The interview also into aspects of new terminology like data parsing and data filtering.