AI and Privacy Implications

The session welcomes Igor Yablokov, founder and CEO of Pryon, known for developing the technology that forms the basis of Amazon's Alexa. The conversation will focus on AI, machine learning, neural networks, and their crossroads with data protection and privacy.

Igor explains how his work at IBM on a multi-modal research team led to the idea of AI assistants. He founded a venture-backed company in 2007, ahead of its time, which eventually became Amazon's first AI acquisition, codenamed Pryon, forming the core of Alexa, Fire Phone, and Fire TV. He shares how challenging it was to be ahead of the market and how people didn't understand the concept initially.

Igor clarifies that 'AI' is largely a marketing term, while 'machine learning' is the actual practice where systems learn from observations and feedback. He uses an analogy of a new employee gaining experience to explain machine learning. Neural networks are described as a technology that allows for unsupervised clustering of data, like images, without pre-defining features, contrasting it with older 'feature engineering' methods. He also highlights how CAPTCHAs are often used to train these AI models.

Igor discusses how big tech companies often fail to adequately inform consumers about data collection, citing the recent discovery of link previews leaking private information. He argues that companies prioritize adding features for convenience while covertly collecting more data. He criticizes the lack of transparency in app updates and settings, suggesting that critical information is often buried or made difficult to understand.

The discussion pivots to the regulatory challenges surrounding automated decision-making and the inherent biases in AI algorithms. Igor highlights the difficulty of regulating these technologies, as companies can often find ways to infer information even if direct data collection is restricted. He points out the tension between the desire for explainability in regulated industries and the rapid, often unexplainable, advancements seen in areas like medical research during crises.

Igor emphasizes the importance of educating citizens to be critical thinkers in an information-rich world, akin to trained intelligence analysts. He suggests that platforms like WireWheel are crucial for inventorying and understanding data flow within systems. He uses the example of Georgia Pacific finding a 'poison document' in their AI system to illustrate the value of knowing what data is being fed into AI. He foresees a future where technology can be used to protect privacy by offering clear 'ingredient lists' for digital products.

Igor introduces Pryon, explaining that it’s named after the codename of the engine that became Alexa. He contrasts consumer AI's home use with enterprise needs, highlighting the differences in scale, accuracy requirements for critical workflows, heterogeneous IT environments, and stringent security postures. He emphasizes Pryon's focus on building a purpose-built, air-gapped enterprise AI platform that allows businesses to control and regulate their AI to ensure data security and compliance.