"When it comes to automation of existing tasks and workflows, you need not adopt an ‘all or nothing’ attitude’ "
Growing interest towards automation technologies (AT) in 2018 vs 2017: a) Deep Learning (DL) and Neutral Networks: 77%; b) Artificial intelligence (AI): 72%; c) Machine Learning (ML): 30%; d) TensorFlow: 17%; e) NLP (Neuro Linguistic Programming): 11%. New tools attracting interest are PyTorch (ML) and Ray --a tool of reinforcement learning (RL) to approach multi-step decision problems in change of supervised learning--. According to a Harvey Nash/KPMG CIO Survey, the level of investment in AI and AT depends on: 1) a company’s self-positioning --digital leaders invest more--; 2) location --e.g.: China invests more--. Reasons why AI technologies will grow: 1) DL follows the improvement in data sets, hardware and software tools, plus a culture of sharing cases and thoughts --e.g.: arXiv, easy-to-use, open source ML libraries--. 2) Companies need more data to benefit. China is “the Saudi Arabia of data”. New tools to get more data: a) services for generating labeled data; b) tools like GANs and simulation platforms to get realistic synthetic data to train ML models; c) open source tools to improve data liquidity: cryptography, blockchains and secure communication. 3) Demand for hardware (compute, memory, host bandwidth and I/O bandwidth) is getting higher. The leaders of DL hardware startups are China and the US. Data, models and compute are to be considered from the point of view of simplicity to explain fairness, privacy, security and reliability. AI is already used for automation of workflows --e.g., as AI assistants: now (FAQ), within 2 years (contextual), in 3-5 years (personalized), in 10 years (autonomous organization of assistants)--.
[TAGs: automation technologies, deep learning, machine learning, artificial intelligence, Neutral Networks, TensorFlow, NLP, PyTorch, Ray, reinforcement learning, arXiv; GANs, simulation platforms, synthetic data, cryptography, blockchains, compute, workflows]