Figure Eight has written a report regarding the data science community and observes that the field is evolving. There are a lot more machine learning projects and data that is required to power them than before. Therefore, jobs in these fields are multiplying. This reports states that this community focuses more on ethical issues even if data privacy has always been a concern. Artificial Intelligence has increased a lot of questions as it is used in many situations, such as medical diagnoses and courtroom sentencing.
In this report, Figure Eight asked questions to data scientists and ethical professionals such as doctors, clergy and police officers, their opinions regarding AI.
Data Scientists don’t like their job, they love it.
How often do you get contacted for new job opportunities?
The first results show that 85% of data scientists are contacted at least once a month and 50% at least once a week. Of course, it wasn’t always like this. Today, more companies realize the enormous potential of data. There are so much of it everywhere that businesses know the real value of having a data scientist in their team.
The thing about data scientists is they’re greedy […] no matter how much data they have, chances are, they need more.
What holds data scientists back?
55% cited quality and quantity of training data as being their biggest challenge. In 2017, Figure Eight asked the same question and the same answer had come up. Data pros affirm that anything your organization can do to help data and machine learning teams should be the highest priority.
High quantities of high quality data are what build accurate models and inform smart decisions.
But what exactly data scientists do with all that data anyway?
1 to 25% admit to use data for AI projects , 17% use it 75% or more of the time and 10% none of the time. There is an important increase in investment for these kind of inititatives and it will be interesting to see how much it will be next year.
Tell us about the tools behind the talent…
In 2015, Figure Eight asked about the tools data scientists used and Excel had still an important role. Today, open source tools are mostly used.
What about ethical issues?
The ethical issues regarding building and deploying AI have been magnified. Many examples of algoritmic bias in sub-disciplines, such as facial recognition, audio assistants and employment application review. However, data scientists are very optimistics about AI en general – 75% of them afirm that AI will be good for the world. In another hand, only 39% of ethics professional who participated to this survey think the same, and 45% think that it will be bad.
What about real-world AI implementations?
What we can say is that implementing AI across all sectors of society isn’t something the community has a big appetite for. When the people behind the technology are, preaching slower, more sober approach, we’d all do well to sit back and listen.
The graph below shows in which scenarios it would be appropriate for AI to make decisions without human interactions.
What are the issues of audio interfaces?
To release or not to release, that is the question.
Audio interfaces are becoming more prevalent. Comscore states that 50% of all searches will be voice searches by 2020. In fact, there are already aroung a billion voice searches each month. However, we see a struggle with regular and everyday speech. For instance, when people aren’t native speakers, have regional accents or dialects.
Figure Eight asked data scientists what is the exact issue with audio interfaces, and what shoul be done about releasing a home assistant that couldn’t understand accents and dialects. The results showed that data scientists are cautious about these implementations and are in favor of transparency.
48% of data scientists affirm that it should be released with warning label that notes some people may not be able to use the product. 12% of them say that there should be regulations in place preventing the release of a product in this state.
If a self-driving car were proven statistically safer than your average human driver, would you rather drive yourself or use an autonomous vehicle?
For most of our survey, both groups tracked pretty similarly. They largely felt that AI was a force for good, that products should have labels informing who they work for and who they don’t, and were more comfortable with AI-driven product recommendations than AI-driven mortgage approvals or judicial sentending.
However, regarding the matter of autonomous vehicles, the reactions were completely different. In fact, 75% of data scientists state that they woul ride in car, and 75% of ethics professionals would drive themselves.