On Thursday, May 25, over 250 public service professionals joined our InnovateUS workshop intended to demystify AI, explain generative AI tools like ChatGPT and Bard, and demonstrate how they can be relevant in government. A follow-up question and answer session will be conducted on Thursday, June 1 (registration link here).
The workshop was led by Alexis Bonnell, one of the founding members of the Internet’s original Trade Association, which helped companies understand the impact that the Internet, innovation, and technology would have on business, customers, and society. Formerly the “Emerging Technology Evangelist for Public Sector and Strategic Business Executive” at Google, Bonnell dedicated her time to helping public servants catalyze their missions with technology, solving the world’s toughest challenges, including working on digital transformation in healthcare, education, COVID response, natural disaster, defense, benefits, and service institutions. While at USAID, Bonnell led transformation and knowledge management in the Management Bureau, was the first Telework Executive, and served as the Chief of Engagement for Education. She co-founded the U.S Global Development Lab of USAID, one of the premier Innovation Labs in government, and she served as the Chief of Applied Innovation and Acceleration tackling the critical challenge of innovation adoption and institutionalization.
Change is happening faster
At the beginning of the workshop, Bonnell explained that change, specifically in the technology and government sphere, is happening much quicker than it used to. Therefore, she argued that government workers should give themselves grace as they learn to adapt to new policies and technologies.
“The rate of change in government used to be an average of five years to fifteen years,” she said. “You would usually have five to 15 years where it would stay somewhat the same or evolve slowly. What has happened in the last three to four years is that the changing paradigm went down from five to 15 years to six months to 1.2 years. If you think change is happening more quickly, you’re right. As public servants, this job has gotten harder.”
Bonnell presented some staggering facts about the rate of change in technological and informational advancement. For example, we’re navigating hundreds of thousands of times more information than we had four to five years ago. In terms of the data we are presented with, 90% of it was only generated within the last two years, indicating how rapidly our knowledge base is expanding. Bonnell notes that it is not within the human capacity to keep up with and process this information at once. That is where artificial intelligence, or AI, can help come into play.
“AI is really our ally in information and attention overload. It serves as my super processor so I can make more evidence-driven decisions while still maintaining my discretion,” she said.
Bonnell then made the distinction between structured and unstructured information. Structured information is human-generated, highly organized, and formatted to be searchable and accessible. In contrast, unstructured information, which makes up about 80% of information, can be found in places like emails, chats, and human-generated PDFs. This information does not have a predefined format or system of organization. When looking at AI, it is important to conceptualize how this technology could potentially change that.
AI vs machine learning
We as humans typically assign human characteristics to non-human things, like AI, Bonnell said. Despite this tendency, she stressed the importance of taking ownership of the relationship we as humans have with knowledge, as opposed to assuming human intent with AI.
“The AI does what we train it with the things we give it to do it. We need to own our relationship with knowledge and make sure we don’t give away that power to technology,” she said.
With this in mind, Bonnell also differentiated between AI and machine learning. AI is able to perform tasks that normally require human intelligence, such as visual perception, speech recognition, and decision-making. Meanwhile, machine learning is the science of getting computers to do something without being programmed with rules. AI and machine learning is already incorporated into our daily lives whether it be through fraud detection, weather forecasts, and rideshare apps.
AI and machine learning has four main functions: classification, prediction, generation, and language understanding. These tools have become essential in all kinds of sectors, most specifically in finance and healthcare. Government can especially take advantage of these tools. For example, flood prediction and weather pattern technology can help government agencies adequately prepare for disaster while also giving officials the capability ton alert impacted citizens.
Stages of information constraint and context
Bonnell identified the stages of information constraint and context for AI and machine learning by comparing the individual stages to supervision and protection of a baby:
- Normal report/rules-based system (a swaddled baby): There is a predetermined output. We limit and direct what we want from a system.
- RPA (baby in a crib): Process automation, constrained automation, or form automation
- Machine learning (baby in a play pen): Modeling within constraints (for example: auto-translation). Give the machine boundaries to work with.
- Artificial intelligence (supervising children at a playground): Geo-based or contextual-based recommendation. Active supervision of the technology.
When discussing AI specifically, Bonnell pointed out that this technology is not something you set and forget.
“AI needs a lot of direction,” she said. “It’s incredibly powerful, but it is not constrained in the same way of our normal reporting, swaddled relationship with information. Therefore, we need to have a dynamic engagement with it.
When should humans make the call?
The “human in the loop” is a model requiring human interaction. In terms of AI, this falls in the middle of the AI intentionality line, right between low human safety impact and high human safety impact. AI is able to help us in so many ways, but there are certain circumstances in which humans need to step in and use their expertise to make the final decision.
Bonnell compares AI to a good intern. It can complete 60 to 80% of the task assigned to them and make progress, but you ultimately review the assignment and make edits before using it in the form returned to you. The point of review and decision is often called “human in the loop.”
Tool demonstration
Bonnell modeled to attendees how to use tools such as ChatGPT and Teachable Machine. In ChatGPT, Bonnell prompted the AI to put a complicated body of text into simple and easy-to-understand language. Within seconds, by taking other kinds of existing data, she was able to get a contextualized body of text that fit her query. In Teachable Machine, Bonnell created labeled image data for different classifications. The system learned patterns from labeled images and tried to classify new, unlabeled images.
Q&A Session
Attendees had several questions about these technologies, and due to high demand, InnovateUS will be hosting an additional Q&A session with Bonnell on Thursday, June 1 to discuss these topics further. Common questions revolved around topics such as security and privacy, the future of work, bias, and the accessibility of AI moving forward.
“If we are worried about [this technology] being a fairness issue or safety issue or things like that we need to be more intentional [with AI usage],” she said.
In her responses to audience questions, Bonnell encouraged attendees to embrace AI in their work rather than fear it.
“I encourage people to use it, to find the boundaries and to find how your unique experience can be manifested positively within it,” Bonnell said.
You can register for the additional Q&A session on June 1 here.
Participant feedback
Of the attendees surveyed, 92% would recommend this training to a friend or a colleague and 90% felt they were likely to use what they learned in their work. Here is what some of our participants had to say about the workshop!
“It was an excellent down-to-earth explanation of AI and how GPT and other such programs work. I'd been told about it by friends but had no real grasp of what it can do. I expect that AI, as discussed in the webinar, will help me do that extra bit of research and delve deeper into topics enabling me to provide more valuable input and make more grounded decisions on my projects.” -New Jersey, Advanced Career Professional
“Super timely and fantastic information about AI from an industry expert.” -California, Advanced Career Professional
“This was so very informative and cool. Alexis was very knowledgeable and comfortable with the material. Great use of tools and demonstration.” -Massachusetts, Advanced Career Professional
To watch the workshop, check out the recording here! Make sure to sign up for our upcoming workshops here!
Want to be a part of our community of innovators?
We’d love to keep in touch!
This work is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License.