Trump's AI executive order glosses over data privacy, funding

Trump released his AI executive order two weeks ago, and researchers are continuing to analyze its potential impact on the industry. One thing is certain: AI is about to change.

The American AI Initiative, signed on Feb. 11, 2019, emerged after a stretch of federal silence on AI funding and regulation. The AI executive order laid out a plan to introduce workforce and business deployments and reduce barriers to technological innovation, but it lacked information about perhaps the most sensitive and important aspects of the use of AI, according to experts.

Tech innovators and business leaders broadly support federal involvement in AI, but they've raised concerns over the executive order's (EO's) lack of focus on funding, data privacy and specific deliverables -- concerns that only grow when the executive order is compared to actions taken by international entities.

The order requires federal agencies to allocate some of their budgets to AI research and development, provide access to data and models to the private sector, and train federal workers in AI skills.

Will budget reallocation be enough?

The AI executive order has dozens of implications to examine, but what stands out immediately is the lack of federal funding.

In the initiative, President Donald Trump calls for a reallocation of internal federal agencies' budgets to prioritize R&D and create proofs of annual planning and trajectory. Experts question the lack of federal funding at a time when China is projected to spend $15 billion over the next 10 years.

"The EO's call upon all federal agencies to prioritize and budget from existing funds can only be an interim step, and most agencies may fall short in bolstering AI without increased funding," said Sagar Shah, principal consultant at Fractal Analytics. "Sustaining American AI leadership at such scale demands dedicated federal funding and continuous follow-through like we have seen in China."

Experts question the long-term benefits of prioritizing technology without monetary support. Some feel the strategy will be effective for deploying existing AI within federal agencies, but it may not be enough to sustain broad growth in the enterprise.

"AI technology can often be effectively used to increase inflows and reduce costs, so it is possible that deployment of mature AI approaches could be deployed in a cost-neutral way or even achieve cost savings," said Dustin Hillard, CTO of eSentire, a MDR service provider. "This could result in AI being used in scenarios that have already been proven, but [it] will limit the newer applications that require investment to develop new technology to yield gains."

Differences in industries could also lead to uneven adoption. If one sector can allocate more funding, they stand to become leaders, while others with less budgetary flexibility will be left playing catch-up.

"The degree to which AI improvements diffuse throughout the economy will depend on how well these initiatives are prioritized by the agencies and implemented," Shah said. "Defense, medicine, education, agriculture [and] energy are some sectors which will get significantly impacted and will start seeing slow but steady changes in their workforce DNA to incorporate the new way of the future."

Public vs. private tech businesses

The AI executive order seeks to create a relationship between private and public enterprises through shared technology and universal standards of AI development -- a subset of the order that is a welcome sign of growth to some in the industry.

"The executive order clearly demonstrates the critical importance of AI in the ability of any industrial nation to gain and maintain competitive advantage in a slew of fields -- civic, military and industrial," said Geoff Webb, VP of solutions at PROS.

While private innovation in AI has soared, slow federal innovation and implementation have created a gap between those who use AI in their business processes and those who don't. Experts hope that the potential partnerships between public and private enterprises that were laid out in the EO will close the gap and lead to broader AI implementation.

"Rapid prototyping in public AI ventures will allow initiatives of national interest and importance to get agile -- find opportunities for AI, build creative solutions, and accelerate deployment or pivot as needed," Shah said. "The country can leverage the government's role as a consumer to promote new AI technologies that have public benefits."

Big data

Looking beyond the potential funding issue, another cause for alarm is the subset of the AI executive order that calls for sharing government "data, models and computing resources" with researchers and the private sector. With AI regulation a future possibility and European data sources locked down by GDPR privacy rules, there may be implications to sharing sensitive government data with private companies.

"Businesses are shifting to increasingly digital models -- models in which competitive edge is defined by the ability to gather, process and extract insight from immense quantities of data," Webb said. "And that capacity is best delivered by AI technology. Any support from the government to enable businesses to develop, adopt and utilize AI to gain and maintain a competitive edge is not only welcome, but critical to our economic security."

"The speed and scale needed to train and implement reliable, robust systems with AI need big data," Shah said. "Having better access to government data promises to strengthen the training of AI algorithms and help software overcome the inherent biases of incomplete or misleading information."

But it's a double-edged sword. Experts agree that AI needs vast troves of data to train, model and perfect algorithms. The government has that data, so it seems natural to share it. However, with private tech's long-standing history of data breaches and misappropriation of data, how can federal entities be sure of its safety?

"Privacy is an American value and is a key area where the government should create legislation to ensure AI advances -- but not at the expense of people's privacy," said Eli Finkelshteyn, CEO of Constructor.IO a search-as-a-service software company. "As a tech founder, I believe in improving user experiences using AI without collecting users' personally identifiable information."

"Technology can be an important part of the solution in protecting our data," Hillard said.

In the nation's quest to support R&D, public and private use of data must be strictly monitored and accountable to the citizens who supply the data, experts said.

"Techniques for anonymizing data to protect against divulging personal information are starting to show promise, although there is significant risk that, over time, algorithms will have the ability to de-anonymize data that was initially thought to be safe to disclose," Hillard said.

Though a welcome boost to the universal implementation of AI, the AI executive order may not be detailed or funded enough to ensure the massive development required to compete with international AI giants.

Dig Deeper on AI technologies

Business Analytics
Data Management