Democratization, ethics and data poisoning -- AI and ML predictions for 2021
In the second of our series of pieces looking at technology predictions for 2021 we look at the field of AI and machine learning.
This has the potential to impact on many areas of commerce, science and digital transformation, but what do industry experts think?
Manoj Choudhary CTO of Jitterbit believes that, "In 2021 we will see AI, machine learning and IoT define and shape our lives and behaviors, a phenomenon that will continue for many years to come. These advancements impact how we work, how we buy, how we spend, how we do every little thing in our lives. But I think the real star that companies will turn to will be the enabling technologies such as cloud and edge computing, which will continue to dominate due to their ability to process and manage all the necessary data that fuels AI, ML, and IoT, as well as enabling technologies like iPaaS, APIM and RPA. These technologies will continue to lead the digital transformation charge for businesses as they move from manual or paper-driven business to digital businesses that can finally tap the power of AI and IoT."
Joanna Lowry-Duda, machine learning research scientist at Luminoso, says, "I think that a trend of democratization of AI will be ever more prevalent in 2021. There will be more toolkits, pretrained models, and datasets available for general consumption. I hope that this will result in more understanding, specifically from business users, of what problems ML can solve (and how well) but also what its limitations are."
James Johnston, regional VP at Cloudera believes, "We will see ethical AI top the list for business priorities in the next 12 to 24 months. While progress has so far been slow, as we enter a new decade, businesses will wake up to the business benefits of ethical AI and how it can give them a competitive edge. Enterprises will start to anonymize data for the good of society, and, furthermore, ensure they have strong data governance that monitors how this data is used."
Dipti Borkar, co-founder and chief product officer at Ahana thinks, "We'll see more data-driven companies leverage open source for analytics and AI in 2021. Open source analytics technologies like Presto and Apache Spark power AI platforms and are much more flexible and cost effective than their traditional enterprise data warehouse counterparts that rely on consolidating data in one place–a time-consuming and costly endeavor that usually requires vendor lock-in. Next year will see a rise in usage of analytic engines like Presto for AI applications because of its open nature -- open source license, open format, open interfaces, and open cloud."
Keith Neilson, technical evangelist for CloudSphere thinks AI will gain momentum in cloud security and governance, "In 2021, AI will go far beyond simply detecting anomalies and thereby flagging potential threats to security teams. Cloud governance is an increasingly complex task and is quickly reaching a point where it's impossible for humans to manage alone. AI will increasingly be relied on in the coming year to maintain cloud hygiene by streamlining workflows, managing changes and archiving. Once proper cloud hygiene is established and maintained with AI, it will also be used as a strategic predictive knowledge tool. By predicting and addressing threats and vulnerabilities, AI will help enterprises create the best possible outcome for their cloud environments. Leveraging AI as a strategic asset will empower CIOs to make informed decisions about their cloud environments, such as evaluating costs and compliance risks."
On the other side of the coin Ben Goodman, CISSP and SVP at ForgeRock believes attackers will start to see AI as a target:
In 2021, we will see an increased number of 'data poisoning' attacks occurring as more organizations are deploying AI platforms across their systems. In previous years, malicious hackers had already discovered that they can attack AI and machine learning software by feeding the AI illegitimate data to cause it to produce negative and/or inaccurate results. This will become a more prominent issue in 2021 and the following years. Bad actors can feed the AI software an image with another image inside that does the opposite of what the AI is supposed to do so it will poison the AI algorithm.
For example, when AI is used for detecting fraud, fraudsters can submit bad data that makes the software unable to detect the fraudulent activity. Many security platforms use AI and machine learning data to detect cyberattacks by identifying anomalies in existing data, so this is a considerable threat that could potentially throw off their detection methods. In 2021, it may be necessary to use separate AI to do integrity and security checks on data collected by the initial AI software.
Daniel Kivatinos, COO and co-founder of DrChrono, see the potential for ML in the healthcare field, "Machine learning will become an even bigger part of healthcare in 2021. For example, machine learning combined with telemedicine will give more insight to a physician in a remote setting reading voice intonation and facial expressions for a patient’s mood, pain, or depressional flags and the like. ML will be able to indicate and provide mood data to the physician in real-time and will allow the doctor to take any immediate action. ML-powered medical copilot scribing can help write a draft chart as the patient and doctor chat via telehealth. A lot of information is exchanged quickly, and having a real-time scribe will save time for the medical community."
Sastry Malladi, CTO of FogHorn, believes AI can will help older industries to clean up their act, "Industries like manufacturing and oil and gas have been slow to implement decarbonization efforts as they struggle to maintain productivity and profitability while doing so. However, climate change, as well as regulatory pressure and market volatility, are pushing these industries to adjust. To avoid the worst climate impacts, global greenhouse gas (GHG) emissions will not only need to drop by half by 2030, but then reach net-zero around mid-century and oil and gas and industrial manufacturing organizations are already feeling the pinch of regulators, that want them to significantly reduce CO2 emissions within the next few years. Technology-enabled initiatives were vital to boosting decarbonizing efforts in sectors like transportation and buildings -- and heavy industries will follow a similar approach. As a result of increasing digital transformation, carbon-heavy sectors will be able to utilize advanced technologies, like AI and machine learning, using real-time, high-fidelity data from billions of connected devices to efficiently and proactively reduce harmful emissions and decrease carbon footprints."