Machine Learning is Everywhere
In order to understand AI's current uses, it's important to distinguish between generative AI and machine learning, according to Jai Jaisimha, co-founder of Seattle-based non-profit The Transparency Coalition, which advocates for legislation requiring transparency around the data used to train AI models.
Predictive models take a set of existing data and use it to find patterns and predict future results.
"The way those work is that they look at information about, say, 'is this a fraudulent credit card transaction or not?' That's an example of a machine learning model,"Jaisimha said. "So they look at every transaction that was previously fraudulent. And then they look at a sample of them that are not. And then they use software to understand what distinguishes a fraudulent transaction from a valid transaction."
"When you apply for a credit card, there's a score called a FICO score. That credit rating score is computed by a machine learning model," Jaisimha said. "When you log on to a website, they ask us questions about, 'which of these addresses did you live at?' Those models are used to verify your identity, those are machine learning models. When people are applying for loans, there are machine learning models involved to make a decision."
Government decisions are even being aided by predictive models in some areas, according to Assefaw Gebremedhin, who leads Washington State University's AI research working group.
"You also have times where you use AI for public health — in finding out how spread of diseases happens, for example. Or how to equitably do services for health in areas where there is not much access," Gebremedhin said.
What About Generative AI?
For all the attention that it's gotten recently, generative AI, which takes existing information to create, say, an email, from scratch, isn't being used in day-to-day life as much as its older counterpart, according to Jaisimha.
"In the case of generative AI, the data doesn't have labels. So what generative AI tries to do is, at the heart of it all, it's basically what's called a next word predictor," Jaisimha said. "Let's say you're typing a sentence, stop mid-sentence. What is the most likely next word that would complete the sentence or would take that sentence one step further. So all forms of generative AI generally are some version of that."
Jaisimha compared generative AI to a parrot, which can learn to mimic humans without understanding what the words it's saying mean.
"Can a parrot think? No, but it can say things that sound human."
Much of the discussion surrounding generative AI has been about its use in education, both good and bad. While Jaisimha, who also serves as an affiliate professor at the University of Washington, said he's seen it used to help teachers craft lesson plans or personalize each student's assignments, students can also use it to write assignments.
"I get notes from students who are writing to me, and you can just see the way it is written," Jaisimha said. "I haven't had a class assignment turned in look like it was generated using artificial intelligence."
So why has generative AI gotten so much attention? Jaisimha said it's largely because now the public has access to the technology that was previously hidden behind software.
"Their entire Facebook feed, right, what's showing up? That's AI in action. Instagram — AI. TikTok. I think what's changing now is that we will have the ability to interact with these algorithms, as opposed to them being buried in the back of some other user experience," Jaisimha said.
Uses Specific to Washington
While AI is used in similar ways across the country, Washington is at the forefront of its integration in medicine, according to WSU's Gebremedhin.
"Healthcare is one big area where research in AI happens at WSU and I would say Washington State at large," Gebremedhin said. "You can have homes where elderly people reside in and you monitor their activities and assist them with their day to day activities. There is a lot of room for using AI in places like that."
Algorithms are being developed that can help monitor blood pressure or disease or detect signs of cancer.
"A smart home or a digital house is the big area... There's also wearable technologies like a smartwatch or some other things where you can wear them on your body and they could sense different things — how the heartbeat is doing, or level of glucose in the blood," Gebremedhin said.
Washington is also at the forefront of using AI to monitor natural resources, especially with the prominence of the fishing industry in the state. A professor at the University of Washington, Jenq-Neng Hwang has worked with the National Oceanic and Atmospheric Administration to develop models to track fish populations.
"The Pacific Northwest, we have a lot of fisheries. NOAA has several research centers in this area," Hwang said in an interview. "We try to put all different kinds of cameras and AI technology to electronically monitor all the fishing activity, as well as understanding more about the underwater fish abundance."
The technology can be used for both research and for commercial purposes.
"We also use cameras and AI to monitor commercial fishing boats. Seattle has a lot of fishing industry, so they want to put a lot of cameras on the fishing boats... Eventually, if you put a fishing boat and it's equipped with this kind of government technology, what they are trying to do is [see] how many fish you are catching on this boat. What are [their] sizes? What are [the] different species? Are you catching something which is illegal because they are too small, they are endangered species?" Hwang said.
According to Hwang, the technology allows fisheries to operate far more efficiently than they had in the past.
"So you throw [the fish] into a chute. We have a camera, we have [an] AI algorithm, we can automatically tell you what kind of fish, how big are they, and eventually, how many are you catching on this boat," Hwang said. "And that was previously done by human effort."
According to Gebremedhin, much of WSU's AI research is focused on integrating the technology into the state's agricultural industries. The school partners with National Science Foundation-supported AI research group AgAID to study potential uses of AI in farming.
"There's a very big emphasis at WSU on using AI in the space of agriculture — using robotics for picking farm products, it could be about having sensors to do better timing in how to increase productivity, or studying how manufacturing can be more automated," Gebremedhin said.
Washington Is A Leader in AI Infrastructure
While Washington ultimately uses AI in similar ways as other states, Washington is unique in the AI landscape in terms of how much infrastructure it has to support AI. Much of the country relies on Amazon Web Services or Microsoft's Azure cloud computing services, with both companies being based in Washington.
"Both Amazon and Microsoft have large cloud computing infrastructure, and that's what you really need. You need just a very large amount of computational power to build these models," Jaisimha said
The state government is also among the first to address AI. In early March, the Washington legislature approved a measure to establish an Artificial Intelligence Task Force, becoming the sixth state to do so. The 19-member committee will meet at least twice a year, before submitting a report to the governor by mid-2026.
(c)2024 The Bellingham Herald (Bellingham, Wash.) Distributed by Tribune Content Agency, LLC.
