While artificial intelligence applications in business and industry remain limited to narrow machine learning tasks, we are seeing progressive improvements in the convergence of algorithms and hardware that will have significant implications for how well and how quickly we can implement AI. Researchers can now train neural networks within a few hours or days, which opens up an amazing range of possibilities, products, and things to learn — as well as challenges — that we could not have even considered before.
For example, Google’s AI group, DeepMind, is hard at work unraveling the mysteries of how proteins fold themselves, a discovery that could have far-reaching implications for health care. It is also very much involved with the research community in working through the ethical issues of AI.
As I see it, 2018 will be the year AI will meet a crossroads — when companies are better able to skim the hype from the reality, and when we can apply AI for both the good and the bane of humanity. Here’s how it can happen:
1. Hackers reverse-engineer and defeat ML-based security systems
Recent, widespread security attacks are strong evidence that hackers are becoming more perverse and clever. With the use of artificial intelligence, computers can actually corrupt themselves and hackers can achieve their ends far more quickly and surreptitiously. In 2018, there is a strong likelihood of a high-profile data breach in which hackers reverse-engineer or decompile, then defeat, machine learning (ML) security systems via an insider assault, malware, ransomware, or machine-based attack.
2. AI resolves the coming backlash over ownership and control of data
The coming year may see a new and powerful backlash set off by a data breach or in response to the upcoming enforcement of General Data Protection Regulation (GDPR) in the E.U. or rescinding of net neutrality in the U.S. During this backlash, individuals could demand that their personal actions on the web, stored as data, be legally recognized as their owned IP. If this happens, industry giants including Facebook and Google, which possess a growing monopoly over this data, will have to answer fundamental questions over who actually owns it. This means users and tech companies alike will have to decide who decides how data is used, profited from, and shared — and AI can provide the answers.
3. Bad actors turn AI chatbots into a new menace
This will be the year hackers, fraudsters, and other organizations operating in the dark web can emerge from the shadows in a new and frightening way, learning how to influence AI chatbots. These interactive agents can already update your bank account balance or serve as your hotel concierge, and bad actors may turn their ability to self-initiate tasks and engage in quasi-conversations to nefarious activities such as crashing utilities, stealing money, and manipulating human actions, opinions, and decisions. The silver lining is that we can also deploy AI to detect the ever more complex ways that charlatans interfere in our lives.
4. AI combines with blockchain to power deep learning on steroids
In the year ahead, we can begin to see AI combined with blockchain to spawn a remarkable new level of deep learning that is smarter and learns faster than previously conceived. It’s only the beginning of a wave as the burgeoning prevalence and immutable nature of the data stored on the blockchain could yield more accurate AI predictions. Cutting-edge companies could, in turn, surprise us with futuristic ways to find, extract, and analyze data in the blockchain to solve old problems.
5. Breakthroughs result from using NLG and NLU to auto-teach AI learning systems
As the year progresses, we may learn about AI breakthroughs by computer scientists who use natural language generation (NLG) and natural language understanding (NLU) to auto-teach AI learning systems. While companies already use unsupervised machine learning algorithms such as generative adversarial networks (GANs) to perform more simple tasks, computer scientists are now on the road to making progress that allows for “one-shot” learning with contextual data like contracts, speech, and video. We’re likely to see a single example that uses NLU and NLG to automatically generate new but valid items to learn from, and thus auto-improve the models.
6. AI goes mainstream with lawyers, shifting away from external legal services
This year we may see a massive and seemingly overnight shift in the legal industry away from external legal services as AI and other advanced technologies go mainstream with global corporations. Exacerbated by growing security concerns, and accelerated by the fact that AI has become relatively easy to deploy and use, procurement will be a driving factor for legal professionals to use AI-driven systems to create savings. This change could be the catalyst that moves the influence of chief performance officers, chief legal officers, and legal ops from the backroom to the boardroom, with support coming from all departments as clients no longer accept high billing rates for something AI systems can accomplish easily.
7. Technology and standards merge to enable a new intelligent contracts framework
I expect that a merging of technology and standards will begin to occur in 2018, with the core functionality for intelligent contracts (IC) becoming available at the protocol level. This is already starting to materialize as smart contracts, which keep sensitive data encrypted at all times, have become available on the blockchain. End-to-end encryption and security can enable further expansion into secure contracting between many parties, and this key component of both IC and AI-enabled secure learning could see applications built on a new IC framework.
The building blocks are already in place for a new era in AI, when AI systems are capable of learning complex tasks with little human intervention. While dramatic claims about the future of AI ring alarms, I’m convinced the year ahead will also highlight our capability to innovate in substantially positive ways. AI can bring a high quality of change to our lives — safer roads, cleaner oceans, and predictive health care, to name a few areas. We are still, after all, the masters of AI’s destiny.
Kevin Gidney is cofounder of Seal Software, a company that provides contract discovery and analytics.