AI Report Card for 1Q 2020
Blog: Jim Sinur
While AI is in its spring season and is sprouting up all over and the predictions for future revenues are pretty positive, I think it’s time to try and give AI an early grade in a number of areas. I’ve picked out my top 10 categories for grading AI and have assigned a grade. In order to understand the context of the grades, I have included a difficulty score and an expected time to maturity. This is my first cut at grading AI and I’m sure I will add to the dimensions and scale over time. Let’s examine the meaning of the grade categories listed below:
AI has done well in scoped problems that are after a silo problem. This helps see clearly the results of AI even if it is a complex problem domain. Over time AI will increase its scope.
AI is often supporting a number of solutions and technology combinations through embedding itself. This brings an aura of intelligence to the solutions, so this is growing fast.
AI is good at black-box behavior by putting a buffer between the user/uses and the complexity of the problem it supports.
AI is a bit about problem-solving today rather than creation or judgment, but this is a less risky approach to establish a baseline of success.
Being able to create something or judge a situation trough iteration and inspection while including new sources of inspiration is something that AI is just starting to do.
The ability to look forward and project future expected outcomes of even plan alternative scenarios is something that is just emerging in AI.
On the Edge
AI is just starting to be distributed and being put closer to where it is needed. Distributed, just in time intelligence will grow significantly over time.
Can the application of A and it’s behavior be explained to humans or even other AI capabilities? This is a must to gather confidence in going forward. AI is headed there, but it isn’t easy yet for the most part.
Giving AI the freedom level to act alone in an unsupervised fashion is somewhat new and will require goal orientation plus constraints via guardrails for proper governance.
Integrated/Cross Context Problems
AI that senses, orients, decides, and acts across multiple problems domains and contexts is coming. For now there a few examples that have been completely successful yet. Self-driving vehicles are headed there.
AI has some growing to do before it will bloom fully. The best progress has been on focused machine learning opportunities and is quite often combined with other algorithms. Without getting into singularity issues, I think the future of AI assisting and collaborating with organizations and individuals is quite bright. I will be looking forward to the next grading period.