Data is the first hurdle of AI implementation
Many government agencies interested in implementing AI lack the data governance, lineage, and master data management necessary to address redundancy and build business unit trust in their data. AI models are useless without quality data on which to train them.
The reality is that many government agencies have a mix of critical and non-critical data in silos, and will continue to do so for some time to come. One solution is to put all your data in one place on the cloud. The other is to abstract these data wamps onto the orchestration plane for data scientists to manipulate.
Agencies should also ensure that data is not biased in ways that could cause AI models to make false predictions, and should consider anonymizing certain data used for training.
Dig deeper: How AI and IoT improve efficiency while reducing costs.
AI requires a cultural shift and serious investment by agencies
Government agencies that need AI should be prepared to invest in research and development and the talent needed to operate and maintain the technology.
Government agencies must honestly assess whether they have or lack the essential skills for AI adoption. Programmers and data scientists need specialized skills because computing and data storage devices are connected differently in an AI environment.
Internal training of employees on AI comes with its own set of challenges. That talent could be poached by industries that can offer higher salaries. Given that it will be another decade before agencies can replace the workforce lost to AI, if agencies are serious about this technology, they must pay AI talent differently starting today. it won’t work. It is expected that AI personnel will earn more than managers, although it will not reach the general salary level.
We also need to invest in back-end infrastructure that can parse petabytes of data, and our $25 million annual budget isn’t enough. It costs tens of millions of dollars a year to keep ChatGPT running. The agency should be prepared to spend his billion dollars on one project, instead of spreading millions of dollars across various pilots that would never be viable at that level of funding.
read more: Check out the “blueprint” of the Federal AI Bill of Rights.
Machine learning is the gateway to AI
Some of the best use cases for early AI include mundane tasks that governments criticize as slow to handle, such as backlog of paperwork that large language models can help reduce. A simple self-service feature like an FAQ chatbot can greatly reduce the workload of agency staff responsible for interacting with the public. In the long term, governments may use AI to evaluate vendor performance on contracts.
Federal education is needed to understand the options that industry can address. CDW·G holds on-site workshops with agencies to identify such use cases and the data, hardware and computing resources that should be part of the AI roadmap.
In some cases, ML is easier to implement and can serve as a gateway to AI while solving pressing problems for government agencies. Microsoft Azure offers several automated ML solutions for easy short-term classification.
At ChatGPT, we have agencies looking for AI grand slams when they really just need a base hit to kick off their program.
This article is part of fedtechof CapITal blog series.
