AI
Watch Tesla’s neural net labeling complexity for stop signs+
Insightful and interesting video from Andrej Karpathy on the challenges of tuning and labeling different situations for self-driving. The stop sign variables show how difficult a problem this is. Getting to 80% is probably not too difficult. Getting to 95% is very difficult. Getting the last 5% and then 1% is not only a massive challenge but you can see how labeling could introduce false positives and negatives and actually hurt system performance. For many applications, 95% is good enough but not here.
Andrew Breen is a partner at The Buy Build Fund and the principal of Assert Digital Ventures where he acquires & invests in small, cash flow positive digital businesses with growth potential. He uses his years of digital product experience to expand the market. Currently focused on health & wellness, Andrew has grown ADV’s acquisitions significantly to date. In addition, Andrew advises leading companies from startups, investors to Fortune 1000 companies on digital products and transformation. Known for his deep knowledge of the Lean framework, Andrew has significantly restructured the digital products, processes and culture of a range of companies. He is an adjunct professor at both NYU's Stern School of Business and Courant (CS) Institute teaching on a range of tech product management and innovation topics. He is contributing author on two books on tech product and cultural topics.
