The undelivered promise of AI
The perception of novel solutions in engineering is convoluted, especially in artificial intelligence (AI). E.g. machine learning is promising to build the next generation of safe, lightweight and performant designs based on historical data.
Yet, rightfully the acceptance of these ambitious claims is declining. The AI models’ brittleness, embedded bias, and sluggishness in working with ‘uncertainties’ and not forgetting their innate weakness in math are becoming show stoppers. Rafinex stochastic topology optimizer Möbius is radically different because it doesn’t fall into those traps. On the following link, you’ll find a blog article by IEEE Spectrum about seven ways AIs fail: [read here]