I met about 40-50 academic roboticists this week after a major conference in Macau. Approximately 90% of them were applying some form of ML to motion planning or evaluation tasks, maybe 1/3 of which were cobotics related. 7% were applying it to CV. 1% were creating semantic maps between natural language and popular consumer robotics domains (navigation). 2% were doing off the wall stuff like 'soft robotics' with paper and strings. As someone working in the space without an academic background and coming from software, I was absolutely floored how few practical industrial or commercial problems were being tackled or even considered. At the academic level, ML has the funding. Cross with cobot for extra funds. Bonus funding points if you use public funds to develop 'dual use' technology: search and 'rescue' was a meme. Rather morally bankrupt.