A Sneak Peek at the State of AI 2016
Early trends from our survey about Machine Learning, Natural Language Processing and Computer Vision
Yesterday we started collecting responses to our State of AI survey. Although the survey’s only just begun, we couldn’t resist peeking at the results, which already suggest some interesting patterns.
If you’ve already taken part in the survey, thank you! We hope you enjoy this little look at how others have answered. The data won’t be truly useful until we get more responses from researchers, developers and product owners working on Artificial Intelligence technologies. If that’s you, we hope you’ll take the survey, and share it with your colleagues.
AI is being used to open new lines of business
We’re looking forward to seeing how this breaks down per industry.
Still very early in the AI adoption cycle
Of the respondents who are work on commercial systems, 24% say their systems are already profitable. We already have a lot of questions about this. For instance, are the new lines of business already profitable? We’ll give everyone the raw data once the survey is complete, so you can dig into it however you like.
Nobody thinks academic papers are irrelevant
There’s a very strong transfer of knowledge between research and production. This has made advanced research degrees very valuable, and encouraged companies to invest in internal research.
Python is the (wrapper) language of AI development
These early results are likely to be somewhat Python-biased. We’re the makers of spaCy, a Python NLP library, so Python developers are over-represented in our network.
Data quality is a bigger problem than data quantity (but not by much)
We also asked about efficiency, memory usage, accuracy and versioning problems. Are the problems the same for everyone, or does it depend on the language you’re using, what industry you’re in, and where your company is in the development cycle?
Nobody likes relying on SaaS
It will be interesting to see which types of AI development are the most SaaS friendly, and which types are least happy to rely on it.
