Recently, a post on Quora, about a story of data scientist got fired, has spread out so fast. We are now in the revolution era of AI but we are also in the hype of AI.
Look forward to seeing you guys debate/discuss on this post.
I resonate with many of his points.
ML/DL/DS are definitely not the magical tools that solve every problems, even in academia.
Speaking from my own experience, I am in grad school with a very famous Professor, expecting to apply ML into wireless communication problem, only to find out that the application is very limited.
I 100% agree with him, and am tired to see how AI is presented to people compared to what the reality of it is. Like about the whole world is collectively and almost willingly missing the point. ML just does statistics, but it does it very well, so we can get meaningful information from loads of data. Hell we can even predict the future with good enough data! But, with what Deep Learning is now capable on doing, especially with computer vision and NLP, everybody starts thinking that we’ll create C3PO in 10 years and that we can already do some magic tricks with AI.
The main problem is that practitioners are accomplices in this matter, because selling dreams attracts investment which in the end allows said practitioners to practice. By always showing the few examples that work well within an experiment to the public (hi GANs and text gen !), we make everybody think the technology is fully mature and applicable. So everyone starts freaking out that AI will replace humans everywhere and will solve everything on their own, while totally ignoring the real dangers it poses with what it actually can do right now (profiling, fake texts/videos, arbitrary decisions bases on biased AI, etc.). I fear that a new AI winter is coming when more companies figure out that it isn’t magical at all, and therefore stop investing altogether until someone finds a new way of doing AI.