With half of the year in the rear-view mirror, it’s a good time to re-examine the annual predictions we made back in January to see where we stand, which ones are on track and what new ones may be emerging.
There’s one thing that we know for sure when it comes to the tech marketplace – the focus on AI has been growing and there’s no sign of it letting up. From 2018 to 2019 alone, the use of AI in corporations has tripled according to Gartner. IDC expects investment in AI solutions to increase 44 percent globally this year over last, with retail and banking leading the way.
We hate to say we told you so, but it’s true. We predicted that AI would become more pervasive in all types of organizations in 2019 and that trend is sticking, with new use cases, technologies and players entering the fray almost daily.
Yet, as we mentioned in our January predictions, while AI use is growing, the task is still rather difficult for many companies for a variety of reasons. Consider the following:
- Data is needed to drive AI adoption – Having an AI solution without good data is like having a car with no gasoline – you’re not going to get very far. This has been and continues to be a critical challenge to organizations of all types and sizes. Data is the fuel that runs AI programs and the more and better data that organizations have, the more accurate the AI program will be. Organizations want to set the world on fire with AI, but they’re realizing that it’s much more difficult than it sounds. Because of the data problem and the demand for AI programs to address critical business needs, data scientists, as well as data engineers, are taking on the added responsibility of making sure that the data is clean and in good shape before it is fed into AI algorithms. There has also been a growth in data labeling and data cleansing services to accomplish that.
- The shortage of data scientist continues – While more courses, certificates and university degrees in data science are becoming available than even a few months ago, the need for data scientists still outstrips the demand and the gap continues to grow. Some data scientists are dealing with data issues, as mentioned above, instead of spending their quality time experimenting and training and fine-tuning algorithms. And, the need doesn’t end with data scientists, but extends to data engineers as well. Many companies are turning to nearshorers and other outsourcers to address their AI needs in the midst of the shortage of these experts.
- Heavy-duty computing power is required – Some companies are using laptops as a learning tool for AI development, but when it comes to developing robust AI programs at scale, there is still no substitute for using advanced GPU processors which are capable of crunching volumes of data and handling simultaneous calculations quickly. Given the cost of these GPUs, and the shortage of in-house data scientist and data engineering expertise, the trend of outsourcing AI projects continues to grow; it’s a lot cheaper than investing in the infrastructure for many companies in the long run. One interesting new offshoot, however, is that I suspect the cost of GPU computing may begin to come down as new GPU providers enter the mainstream.
Earlier this year, we also looked ahead at specific areas of growth, and predicted:
- Data will move to the edge – With the increased focus on video surveillance and security applications, the need to process and analyze the data at the edge, where it is captured, is continuing to increase and AI is stepping up to the plate. A first key step is to ensure that the data captured at the edge is clean, so companies can not only avoid the latency, cost and bandwidth issues of sending it to the cloud first, but even more importantly, so it can be accessed quickly for real-time insight. Keeping the data local also helps to ensure greater security and privacy. According to VentureBeat, looking ahead, there will not only be a focus on data at the edge, but AI at the edge too, as 5G enables greater connectivity with very fast speeds and extremely low latency.
- 2019 will be the year of computer vision – According to Forrester, computer vision is really taking off, with lots of investment and hundreds of startups. It’s being used to monitor people’s behaviors for security or for diagnosing certain conditions in medical imaging, among other applications. Forrester notes that most companies do not have the in-house expertise to undertake these programs themselves.
It’s hard to believe it’s almost here, but as we sprint to the finish line in 2019, we can expect more AI developments and use cases, as well as solutions to the challenges of AI adoption. What lies beyond that is anyone’s guess, but based on our track record to date, AI should remain the focus for quite some time.