In part 1 (click to access part 1) of this article I examined human collaboration, marketing personalization, AI becoming creative, and AI running within medical devices. In part 2 we examine some more areas we could see more focus on AI in 2020.
5. Privacy Concerns Need to be Built into AI Processes
People have concerns about their privacy, and rightly so. By now everyone is aware that voice activated assistants such as Alexa and Google Home are recording all conversations even if their name is not mentioned. Amazon has a patent to advertise to people based on things they discuss in the home. Google also do the same.
GDPR and the California Consumer Privacy Act attempt to address these issues. But there is a conflict. AI technologies help individuals more when they have access to massive amounts of data. So, how does this need for data balance with privacy concerns?
Most tech companies like Google mention in the fine print of a long legal agreement that your data will be used. If you don’t provide consent, your access is terminated until you do. These days, most people (including myself) just click ‘accept’ so we can continue to use these services.
This data ends up being analyzed with other data about you. And while this can provide helpful services such as accurate purchasing recommendations, smart appliances and even helping the elderly stay safe in their homes, there are also serious potential risks. These databases cast a wide net of surveillance, which can be used for disturbing purposes if people with nefarious intent gain access.
Ai actually could be useful in solving this problem. There is already work going on for AI to analyze data without decrypting it (homomorphic encryption) while differential privacy systems using AI can introduce randomness into data to prevent de-anonymization to succeed. These are both in the early stages, but this is where things will have to head to meet both personalization and privacy concerns.
6. More AI Will Be Deployed in Business Process Monitoring Within All Departments of Pharma
More and more repetitive tasks are now being taken on by AI-powered automation. A recent project we did for a pharma company around payments and rebates associated with numerous industry stakeholders is a good example. They wish to enable an ongoing system to better manage the processes for numerous dependencies, which may hold subjective assessments to inform decisions and mitigate fraud.
However, AI is more commonly used to automate business insights analyses. It can be used for anything repetitive such as generating reports and diagrams, producing documentation, and redaction of PII in clinical trial reports. Intelligently automating these kinds of tasks, frees humans up to focus on more complex, strategic work.
7. AI Used More for Cyber Security Within Pharma Processes
AI will be playing an increasing role in cybersecurity as advanced prediction is required for increasingly sophisticated cyber attacks. We are already seeing demand in this area. Fraud mitigation was a key component of a project we completed a for a pharma company that wanted to use AI for business process monitoring. This kind of AI powered fraud mitigation will increase in back office processes in pharma.
8. AI and Ethics
Ethics and AI is a hot topic. There are many ethical concerns that each could be an article (or a book) in their own right. I will list some of these below, but the issues around AI and ethics is only just heating up and we expect to see some major ethical breaches with AI in 2020 both within and outside of healthcare.
- AI Bias, Racism and Errors – You may have heard how Amazon created AI to vet the resumes being sent in and infamously rejected any applications from women. Female resumes were not part of the training set. Sadly, this is not the only example. Remember Tay, the Microsoft AI bot that became a feminist hating, Hitler-loving anti-semite in less than 24 hours and had to be dismantled? And there’s also the software to predict future criminals that showed a bias against black people. These kinds of errors and bias are serious ethical issues that need to be carefully guarded against. Taking time to do your data wrangling thoroughly is critical to ensure as much bias as possible is eliminated.
- How Should We Treat AIs? Erica the robot and those like her show elements of emotion and insight when discussing their life or human biases. The professor who created her says he is very close to programming consciousness in AI (the topic of my presentation in December 2019 at the BHBIA Winter Event). And even now with the robots we have that are surprisingly human, like Erica, what kind of rights should they have? What about the poor sex worker AI robots like Harmony? https://www.youtube.com/watch?v=-cN8sJz50Ng I was actually contacted on LinkedIn by an AI-powered sex robot wanting to connect with me. That really shocked me. I assume it’s because I’m in several LinkedIn robots groups, At least I hope that is the reason!
- Singularity – What happens when robots become smarter than humans? This is Technological Singularity and some working in this space predict it will happen as early as 2030.
- Rogue AIs – We all know Terminator. Currently we are not at the level of technology for that scenario, but Stephen Hawking warned us repeatedly that it’s a strong possibility in the future. And we are close, if you see the robots the Russians have now created – see below.
- Should AI Be Allowed to Kill in Wars?
While this may only be CGI, the Russians actually have actually already done something like this, and this is where it is heading. See the real life Russian version here. AIs learn and are logical. Predator drones (e.g. General Atomics MQ-1 Predator) have been in existence for over a decade. Although US law requires these to be controlled by a human, not all countries have that law, and some aspects of their work may not be overseen by a human.
9. AI Powering Pills
We already have electronic chip in a pill with Proteus. So it’s just a matter of time before the chips in pills become equipped with embedded AI. This will allow all manner of things. Identifying the best dose for the person’s metabolism instead of simple dosing guidelines by age or bodyweight is one example. Another is predicting adverse interactions if the patient has taken another medication and shutting down active ingredient release. Okay, maybe this won’t be 2020, but it could be coming soon.
These are only the tip of the iceberg in what we should expect, but we are certainly living in interesting times.
For more information on any of the topics contained here, or assistance in planning or implementing any of these topics, please contact the author ,Dr Andree Bates, at Eularis http://www.eularis.com/contact or sign up for our one day 2020 masterclass (date to be determined) on Innovating with AI in Pharma Sales and Marketing in London, UK.