AI in the hype cycle again; will it emerge this time?
- By Ross Wilkers
- Sep 23, 2019
Artificial intelligence, machine learning, analytics and other technologies under the broad umbrella of automation have become a buzzword throughout the government technology arena in the same way cybersecurity was not too long ago.
But AI and those other tools mentioned above are not all that new. AI specifically was founded as an academic and research discipline in 1956 at Dartmouth College and has since gone through waves of optimism but then disappointment, followed by an “AI Winter” of low interest and dry funding. That cycle has repeated many times since.
So why is there another hype cycle in AI these days, and will that hype and promise eventually become material for many agencies that are hungry for such technologies?
I recently spoke with Steve Escavarage, Booz Allen Hamilton senior vice president and a leader in its digital and analytics practice, to see why there is so much renewed optimism in AI and its related technologies.
One key factor Escavarage pointed to in this new or renewed AI hype cycle: the amount of data that exists in the world.
Now also think about the vast amounts of data agencies have to store and by law cannot dispose of, not just the sets created in commercial sectors.
“At the same time, you have these big organizations coming up with computational architectures that are specifically good at doing the type of computation associated with machine learning and big data analytics,” Escavarage said.
Some of that advancement could also be attributed to Moore’s Law: the golden rule of technology price and performance that was originally coined for cost per transistors on chips, but should also allegory to broader trends like cloud computing that can dramatically shrink the cost of storage.
There also might be an argument that even despite the hype, AI is actually undervalued today.
“We have a hard time imagining what the capabilities are going to be like in a few years,” Escavarage said, and hence “probably don’t have a good handle on how good some of these capabilities are going to be.”
It only seems natural then that agencies see those trends and want to take advantage of them. But any progress is going to come in a different form, according to Escavarage.
“Because of the hype cycle and because of all the noise about artificial intelligence, we keep looking for the next thing that’s going to blow us away: the next demo, the next paper that we read that blows us away,” Escavarage said.
“Maybe a different strategy would be thinking of an intermediate outcome of just saying: ‘I think I’ve seen enough evidence to suggest this stuff is improving and I want to create the conditions for myself, for my organization, certainly for my clients,’” he added.
That approach is before that moment Escavarage says “technology takes the next step that we can’t imagine today, I’ve done everything to put myself, my organization, my customers in the position where we can benefit from it.”
Escavarage said Booz Allen has at least 60 ongoing projects across the firm that are using advanced methods such as machine learning to achieve AI, which he views as an outcome of using machines to perform tasks normally carried out by a human.
Then there is the link between AI and off-premise cloud computing infrastructures that can house data and perform automated functions at scale.
Escavarage believes that migration to commercial cloud environments will continue but pointed out another related trend he sees worth watching.
“We’ve seen a nice re-imaging of compute capability we need on premise. There’s a lot we do in machine learning and artificial intelligence using absolutely huge data payloads that you don’t want to have to move across a wide area network to get into a cloud,” he said. “Where you want to have the compute capability co-located where the data is being generated, so you can run your algorithms right there, and this applies to edge locations.”
Booz Allen is also constantly looking outside to see what companies are out there with emerging automation technologies or tools that might apply in a certain agency. The firm in essence acts as a tech scout and adviser to both federal agency customers and industry partners on how to integrate those offerings.
One example of that searching and guiding approach is seen in Booz Allen’s partnership with graphic processing unit maker NVIDIA to train federal employees on what deep learning and artificial intelligence is.
A second partnership Booz Allen announced in August with Hypergiant Industries centers around health care information and fraud detection, plus defending against malware attacks.
“We’re always looking for solutions, products, even just smart people who are building, inventing and integrating capabilities that would benefit our customers,” Escavarage said. “Our goal is to try to be the translator, be able to look through all those capabilities and identify something that is a widget that fits a problem specifically to our clients.”
Ross Wilkers is a senior staff writer for Washington Technology. He can be reached at firstname.lastname@example.org. Follow him on Twitter: @rosswilkers. Also connect with him on LinkedIn.