Business Insider – Meet the Princeton professor and his CEO son trying to reinvent the logistics industry with Optimal Dynamics, an AI startup that helps predict the futureEstimated Reading Time: just 4 min

As reported on Business Insider:

Princeton University professor Warren Powell spent decades developing technology that would help logistics firms optimize their routes. His son, Daniel Powell, is now trying to commercialize the technology and use it to overhaul the industry as CEO of Optimal Dynamics. The tech, which the company refers to as “high-dimensional AI,” allows logistics firms to make informed decisions on matters like which driver should pick up what load much farther into the future. Optimal Dynamics recently announced $4 million in seed funding. Visit Business Insider’s homepage for more stories.

The year was 1982, and Warren Powell was at a dinner with an executive from Schneider National, one of the nation’s largest logistics companies.
The exec asked Powell, a long-time professor and researcher at Princeton University, about whether it was possible to determine as far out as four days what certain routes would look like, and how they could use that information to optimize operations, as well as to work better with both shippers and drivers.
The technology was not available at the time. But the question took Powell on a journey to try to pioneer the algorithms that would help improve supply chain and logistics systems.
Now, he’s getting a chance to turn that vision into a reality, by way of Optimal Dynamics — a startup that’s something of a family business, with the elder Powell helping run the company alongside his son, CEO Daniel Powell. The company is still young, but recently raised $4 million in seed funding.

“Getting this technology out to industry, it’s just fabulously exciting,” Warren Powell, a professor at Princeton University, told Business Insider.
The company uses proprietary technology to optimize shipping routes based upon historical data like the loads moved over the past three years and the current driver network, along with external variables like traffic. The information is then used to train its forecasting models so executives can make more strategic decisions much farther into the future than both Warren and Daniel Powell say is possible with existing AI applications.
The Powells say that Optimal Dynamics currently has three customers, including large logistics providers, and is running pilots with several other major firms. One client was able to cut costs by 13% and cut the number of full-time planners by 90%, says Daniel Powell.
The goal is to top $2 million a year in annual recurring revenue over the next two years. As the company looks to scale its technology, it is bringing on heavy-hitters who understand the industry and can accelerate business quickly. Earlier this year, for example, Optimal Dynamics hired former Uber Freight executive Chris Torrence.
“There’s a lot of noise in the industry. We will start to have competitors as we expand on our growth goals,” said Daniel Powell. “This is very greenfield, but we’ll inevitably start bumping heads.” 

‘High-dimensional AI’ In many ways, the journey of Optimal Dynamics is a reflection of the history of artificial intelligence itself.
While it may be easy to assume all the technology underlying the plethora of AI-backed startups that are garnering billions of dollars in investment today is new, many are built upon discoveries in algorithmic science made as far back as the 1950s.
A common limiting factor in the field of AI has been the availability of computing power, which is required in amounts proportionate to the complexity and scale of the algorithm they run. Now, with the advent of cloud computing platforms like Amazon Web Services and a variety of other tools, those problems are easier to solve.
“The truth is a lot of AI isn’t new,” Jonnie Penn, a researcher at the University of Cambridge, told Business Insider. “The moment today is a combination of past techniques with computing power.”
Then, in the early 1990s, there was a surge in interest in “deterministic optimization,” or algorithms that could effectively analyze the future to provide actionable insight to guide decision-making.
Those efforts largely failed because the models simply couldn’t withstand the high level of uncertainty, in Warren Powell’s telling.  A decade later, he worked with Schneider to develop the technology that would eventually become the basis for Optimal Dynamics.
At its heart, the proprietary technology the company refers to as “high-dimensional AI” helps firms like Schneider optimize their driving networks. But it goes a step further than that, the Powells say.
“We have the system that can move that needle past just supporting a human to automating those decisions,” said Daniel Powell. “We see our applications spreading beyond trucking and more of a decision layer through all logistics and supply chain.”

‘Too many variables for a neural net to learn’ The tech is different from some of the other, more commonly known AI applications in existence now that rely on neural networks — or algorithms that are trained repeatedly to recognize patterns. That approach is used to underpin the AI at services as varied as Google Translate or self-driving cars.
But neural networks can’t be used as effectively to optimize processes like logistics because of all the uncertainty, Warren Powell says.
Unlike airlines, for example, which can plan out routes a year in advance, shippers work on a much more condensed timeline — often plotting out just days or even hours into the future. That creates a challenge for using technology to determine what load a driver should pick up next.
“The structure and the complexities of a logistics problem have simply too many variables for a neural net to learn,”  said Daniel Powell.” If you took all of the data in the entire US trucking market, you wouldn’t have enough data to fit a neural net. And so you need to create other techniques for systems to not just be pattern recognition, but to make robust decisions in this space.”
So say a driver is taking a shipment from Dallas to Chicago. It’s unlikely he or she will know before leaving Texas where their next destination is. The system can get complex extremely fast, with operators having to account for how long thousands of drivers have been on the road and when each will need to return home.
Optimal Dynamics can automate those choices around load pick-up to ensure maximum profits and efficiency, according to Daniel Powell.
“We can blend the whole planning process of operating these very complex truck networks together,” he said. “Actually planning for uncertainty is the real hard piece here.”

Source: Business Insider
Author: Joe Williams
Date: 2020 08 09

PS To People Who Want To Find Must-Read Content but Can’t Get Started:

Get Rid of That Information Overload!

We read everything and we keep only Pro Content about:
– Business #Strategy
– #Startups Strategy
– #Growth Hacking
– Artificial Intelligence #ai

Why do we do it? Because #SharingIsCaring

Sharing is Caring!
search previous next tag category expand menu location phone mail time cart zoom edit close