At Bennett Data Science, we’re seeing a lot more interest in A.I. lately. Especially since about June of this year. This week I’ll discuss why I think this is happening and how A.I. will be an even bigger differentiator to companies that utilize it effectively (read: companies that actually get value from their analytics spend). I’ll back up my thoughts with some data from around the globe.
First off, here are some of the changes from 2019 to 2020 (reference):
- 6.8% more companies are successfully creating data-driven organizations
- 4.7% more companies are using data to drive innovation
- 3.1% more companies are considering data as a business asset to be managed
These are great for our industry and for the companies successfully using data science to drive revenue.
In terms of the types of analytics being used, an O’Reilly study found that machine learning is still used by the majority of early analytics adopters, while deep learning is most popular among organizations that are still evaluating A.I. This makes sense as machine learning is much easier to understand and (often) to implement than neural networks.
The gains, however, can be much greater where companies can successfully use deep learning. Deep learning models have been shown to deliver huge performance leaps in areas such as voice-to-text, image retrieval, and natural language processing. I wouldn’t be surprised to see machine learning garner a smaller percentage of the overall analytics pie in the coming years as deep learning becomes more ubiquitous.
There are some tricky issues when considering A.I. adoption. Surprisingly, the lack of machine learning (ML) and A.I. skills aren’t the biggest impediment to A.I. adoption. Almost 22% of respondents in the O’Reilly study identified a lack of institutional support as the most significant impediment to growth. Furthermore, few organizations are using formal governance controls to support their A.I. efforts.
Governance is a tricky issue, especially because of ever-increasing cybersecurity, ethics, and bias; all issues we’ve covered briefly here. What do you do when your voice translator was trained on one particular ethnicity? Does it generalize well? Probably not, and building diverse models is not only ethically right, but such bias can also expose companies to finger pointing and poor press, even when there was no intended malevolence.
In terms of build vs. buy (or in this case, rent) Deloitte’s 2020 edition of its annual “State of A.I. in the Enterprise” report, released in July, indicates that many enterprises are investing heavily in A.I., and many are buying cloud-based A.I. products instead of building their own. A significant 93% of adopters use cloud-based A.I. services, with far more enterprises buying rather than building A.I. capabilities.
These cloud-based A.I. products are big drivers of adoption. Before, we required many months to deploy useful analytics so that stakeholders could reap the benefits. With today’s cloud-based options available from Google, Amazon, and Microsoft, mid-level data scientists can do a lot of the heavy lifting in a couple of weeks. The value of such easy deployment has fundamentally changed how we approach projects. We can now talk about delivering insights-to-action in a fraction of the time it used to take.
Finally, since the pandemic started, we’ve seen big shifts in unemployment. In terms of analytics, more companies than ever are looking towards automation to replace workers and make up for lost bandwidth. Hence, A.I. and increased tech are where companies are looking for a solution.
More than two-thirds of US companies increased their investment in A.I. technology during the pandemic (reference). 63% of companies already using A.I. said it helped them stay resilient during the big changes they faced during the pandemic. At the same time, a majority of the executives polled (82%) expressed concern over other countries surpassing the United States in terms of A.I. tech development.
Lastly, the most noteworthy trend I see is this one: From 2018 to 2020, A.I. adoption has grown from 48% in 2018, to 72% in 2019, to 81% in 2020. And the pandemic is only pushing that curve forward.
In the coming months, I expect A.I. to deliver more on its promise to deliver value.. Companies are moving forward in record numbers, but with more caution. This is great for our field!
I expect bias and governance to grow in importance in 2021. As non-sexy as these topics might be, they are some of the most important ones to address. We must be as ethical and fair in how we treat each other, as to how we automate machines to treat each other.
Have a nice week!
The State of A.I. in 2020 Likely Sees More Adoption
It’s been a strange, uncomfortable year, for A.I. and most other industries. COVID-19 swept in, leaving hundreds of thousands dead, the global economy in tatters, and millions of people out of work. The full impact of the coronavirus, still raging around the world, has yet to be felt. That said, it’s difficult to precisely gauge the state of A.I. in 2020. While historical data indicates A.I. adoption should increase this year, and anecdotal data indicates many enterprises are turning to A.I. and simple automation to augment a reduced workforce during the pandemic, it’s impossible to predict how the coronavirus will affect A.I. spending and adoption in the long term. But preliminary information shows that A.I. adoption is on the rise, despite, and, in some cases, because of the pandemic.
MIT Team’s Cough Detector Identifies 97% of COVID-19 Cases Even in Asymptomatic People
Part of the challenge in controlling the coronavirus pandemic is in identifying and isolating infected people quickly – not particularly easy when COVID-19 symptoms aren’t always noticeable, especially early on. Now scientists have developed a new artificial intelligence model that can detect the virus from a simple forced cough.
Top Open Source Recommender Systems in Python for Your Machine Learning Project
Recommender systems have found enterprise application by assisting all the top players in the online marketplace, including Amazon, Netflix, Google, and many others. These systems are the decision support systems that make the personalisation process better as well as smoother. It predicts and estimates the content of user preferences by extracting from various data sources such as previous database, data history, among others. Here’s an overview by Analytics India Magazine of the top eight open-source recommender systems in Python, in no particular order, that you could try for your next project.