Artificial intelligence used as chatbots made headlines in 2016 with 34,000 bots added to Facebook Messenger alone. Facebook’s F8 conference put artificial intelligence in center stage, as chatbots are now smart enough to fool students at a Georgia Tech CompSci course who didn’t realize their TA “Jill Watson” was only virtual until finals.
But Facebook is Facebook and development organizations are a long way from their technical capacity, available budget, and purpose. Still, the potential to automate conversations is tempting, especially considering these many digital feedback loops we’ve created and now, perhaps, need to scale without many more funds and, in most cases, limited local technical resources. So how feasible is this beyond prototypes?
Our July 2017 Technology Salon on AI, “To Bot or Not: Can Development Benefit from Artificial Intelligence?”, including comments from:
- Justin Herman, Leading Inter-Agency Federal Programs in Artificial Intelligence for Citizen Services, Virtual Reality and Social Tech, GSA
- Megan Vorland, Director of Partnerships, Dcode42
- Hedeer El-Showk, Co-founder & CEO, ai
We are now seeing chatbots in development, with even in smarter data processing, priming a development innovation debate. What are the ethics in artificial intelligence-directed program decisions? Where can organizations get access to affordable AI resources? How can machine learning really improve efficiency?
So what were our top takeaways from the conversation?
AI for development is in already in practice.
Nigeria, according to a UN figure, the country has set goals to dramatically increase the number of available doctors by 2030, but measuring current financial support and educational levels this is seemingly impossible to achieve. AI can help leapfrog and help achieve address this major issue.
Looking to US government agencies, some federal agencies such as DoD, FBI, State, DHS are preparing for an AI future. They are putting together a government wide programs and roadmaps both in AI and blockchain.
But expect additional investment costs.
A natural disaster in Louisiana and had about 30,000 people asking for help. At that scale it easily becomes difficult to manually figure out where to distribute the appropriate assistance. In this context, we have to make our data actionable and presentable, and to make our services compliant with the Americans with Disabilities Act (ADA) and multilingual.
Often times the people who need the services most are left behind. These gaps are not talked about when systems are inefficient. Bottom line is do not underestimate the deployment and know that whatever the estimated gains are there will be pitfalls on the way there.
Your organization can start getting involved without poaching people from Google.
AI products are now commoditized. Experts in an organizations do not necessarily need to be experts in AI. AI tools should help SMEs in organizations do their work better without the need to know much about AI. AI is about automating specific tasks, day to day interactions.
For example analyzing thousands and millions of documents to provide faster, more accurate and relevant research and analysis of the documents. AI is about hiring less and getting more. The human analyst will always be there in the loop – they are the experts and the ones who are doing the real analytics. AI learns from the experts and AI is no longer about rigid rules but fluid training and learning. A good AI platform should have a usable UI to access the services and not cumbersome to use.
Start small, then stay small.
If you are approaching a problem without the use of AI today, then you will have to redo in two years. There are lots of possibilities and people are excited about it. Start with automation of standard operating procedures (SOP), learn from it and then use the same technology to do it smarter.
Projects run for about two months, review data and then build automation to become smarter. Iterating means your teams can better understand the problems, better train the AI, and monitor for discrepancies such as bias. These smaller investments can compound to larger returns rather than seeking big funds to put Watson toward one project or task and not having anything to carry over to the next project or fiscal cycle.