JPMorgan Chase wants to make sure you don’t quit on your financial goals. 

The bank filed a patent application for a system to use artificial intelligence for “personal financial planning.” JPMorgan’s system is akin to using a personal trainer rather than going on a crash diet: It gathers information about your spending habits and goals, and builds financial plans you can stick to. 

“Alternatives to speaking to an advisor include personal finance assessment tools and questionnaires which offer semipersonalized advice to users based on their input,” JPMorgan noted. “However, these tools fail to recommend actionable points of advice on a more personal and detailed level.” 

JPMorgan’s system takes in information on a user’s current financial state – such as income, expenses, investments, and savings – and monitors their spending habits to find out where their money is going. Next, it processes information about their financial goals, such as 

saving up a certain amount of money or paying down their debts. 

Once the system has all of this information, it defines a set of actions that the user could take to reach this goal, and assigns each of them a probability indicating how likely the user is to take that action. For example, it may recommend decreasing certain discretionary spending habits over others, such as cutting out online shopping versus cutting out weekly coffee runs, based on the likelihood that the user can stick to that budget. 

Finally, JPMorgan’s AI algorithm balances this information and designs a plan with the maximum “likelihood of execution.” It gives the user a step-by-step plan to achieve those goals and calculates the feasibility that all steps can be successfully completed. 

If its patent activity shows us anything, it’s that JPMorgan is working hard to embed AI functions throughout its operations. Its patent history includes an investor-company matchmaking tool, “data science as a service,” no-code machine learning systems, and an AI tool that does due diligence for you.   

The company has been public about its commitment to the tech, too. It launched a cash flow management AI tool that’s being used by around 2,500 clients, according to Bloomberg, and it claims it has cut human-oriented work by 90%. CEO Jamie Dimon said in the annual letter to shareholders last year that the company has more than 300 use cases for AI in practice. And Dimon said earlier this week at the Australian Financial Review Business Summit that the company is making AI “part of the management conversation.” 

“Finance offers plenty of room for innovation,” said Tejas Dessai, research analyst at Global X ETFs. “Right from front-office tasks of customer relationship to due diligence, primary research, collateral creation and modeling, there’s plenty of knowledge work that will eventually end up automated or outsourced to these intelligent AI agents.”  

AI tends to be really good at performing the main function of financial processes: analyzing numbers and making predictions based on them. But the risk rises as AI becomes more entrenched throughout Wall Street. AI tends to hallucinate and make mistakes when asked questions outside of its purview, and it can adopt biases if trained wrong and left unchecked. 

In an interview with 60 Minutes in December, Federal Reserve Chairman Jerome Powell said the organization is researching AI’s impact on employment, distribution of wealth, and productivity, noting that the tech “needs to be appropriately regulated.” The Financial Stability Oversight Council classified AI as an “emerging vulnerability,” citing cybersecurity issues, privacy concerns, compliance, and accuracy problems. 

Dessai notes that these problems can often be resolved with “time, data, and specific training,” he said. “But there’s also a human training aspect associated with these systems. People have to learn to use and prompt design processes with these models effectively, and that’s going to be a long learning process.” 

Plus, as every industry rushes full speed ahead to implement AI into their processes, no one wants to be left behind. “The risk of not making the most of the technology is much higher,” Dessai said.

Adobe doesn’t want its AI models to know the facts – just where to get them. 

The company is seeking to patent a large language model that works with an “external knowledge base.” Adobe’s system uses an AI model as a middleman: Rather than storing facts themselves, it simply pulls them from a large database to make its predictions. 

“Retrieving relevant information from the knowledge base and allowing the (natural language processing) model to use the information when making a prediction reduces the need for the NLP model to store facts,” Adobe said in the filing. 

Adobe’s system first searches for and extracts the different “entities,” or subjects such as names, dates, or quantities, mentioned with a natural language sentence. The system then creates a modified version of the sentence through a process called “masking,” which essentially blurs out one of the entities named in the query to prevent the model from overfitting to the data. 

The system then takes this information to scour a “knowledge base,” which can be any database that stores a larger corpus of information than an AI model would be able to (for example, Adobe noted, Wikipedia). The system then searches for relevant “triples,” which include three descriptors of the entity: subject, relationship, and object, to give the model more context. Adobe’s tech then ranks these triples based on similarity to the original sentence, selecting the most relevant ones from the bunch.

All of that information is taken back to the model, which is trained to make a prediction or response based on a user’s specific query. 

Adobe’s patent highlights that the future of AI models may be smaller than we think. While Big Tech continues to concern itself with building out massive, do-it-all language models, niche and domain-specific models may be better equipped for specific jobs, said Bob Rogers, PhD, co-founder of BeeKeeperAI and CEO of Oii.ai. 

Rather than focusing on a model that can act as a jack-of-all-trades, creating a small model that gets its information from a database of topic-specific information could lead to more accurate answers. “It’s an interesting effort to constrain a large language model to respect facts,” said Rogers. “This idea of making AI models that are more closely tuned to a knowledge base, that sticks to a certain set of facts, could be very appealing.” 

These “pocket-sized models” suck up way less power than a large language model, said Rogers. For reference, OpenAI’s ChatGPT reportedly eats up 17,000 times more energy than the average US home consumes in a day. 

Smaller, locally run models are also more secure, Rogers noted, as the entity that’s training the model knows exactly where the data is going. In the case of the Adobe patent, the person running the model may have control over the so-called knowledge base, rather than feeding data into the Big Tech AI machine. “If you can realistically run local models where you’re controlling where the data goes, that’s instantly a solution,” he said.

For Adobe’s purposes, this tech could add to its business of making AI easier to access for marketing and advertising, Rogers noted. The company’s patent activity is littered with AI innovations that fit this bill. The company also offers Firefly, its generative AI image model that it claims is commercially safe. 

But this time, rather than tackling images, Adobe’s patent helps make AI-generated copy less bland and more specific, he noted. “I think that’s probably very valuable to Adobe in terms of selling tools that create useful advertising content,” Rogers said.

Snap wants to augment the world around you – including your hands. 

The company is seeking to patent a system for “selecting AR buttons on a hand.” Snap’s patent essentially aims to create more intuitive overlap and interaction between augmented reality content and the real world, specifically between its AR creations and your hands. 

With a set of smart glasses, “The interaction with the embedded sensor to perform various modifications of the virtual content is not very intuitive and has a very steep learning curve,” Snap said in the filing. 

Snap’s system aims to improve the “efficiency, appeal, and utility” of these smart glasses by displaying selectable virtual content on a “first real-world object,” a user’s hand, and then tracking movement of a “second real-world object,” or a user’s other hand or finger, to figure out which content the user selected. 

The system takes into account “spatial relationship factors” between the user’s hands and the displayed AR objects to dynamically adjust the virtual world in response to user movements. If there is an overlap between the user’s second hand and the AR object — meaning, if a user taps a certain selection of the content displayed on their other hand — it will activate whatever content they selected. 

“This improves the accuracy of detection of the correct AR object the user intends to select,” Snap said. 

Snap’s patent activity is filled with AR glasses-related inventions as it seemingly tries to snag as much IP as possible in the growing space. The company has sought to patent ways to bring books to life with AR glasses, gaze-tracking techniques, high-tech prescription lenses, and even AR-enabled contacts. The company also received a $20 million grant from the State of California to expand manufacturing of these devices in December. 

Snap has offered AR Spectacles since 2016, dropping three generations of the glasses. Despite its efforts, the company has struggled to scale adoption of these devices, as most of the people interacting with its AR innovations do so through its flagship app.

The road to creating long-lasting and scalable artificial reality glasses faces a number of obstacles. The devices lose power quickly due to high processing power on a small amount of hardware, they overheat, and the field of view is still quite small. These factors are hard to overcome from an engineering perspective, limiting broader adoption by consumers. 

Snap also has quite a bit of competition as far as AR wearables go. Both Meta and Apple are plugging away at lightweight and comfortable mixed reality headwear, all seemingly to make the experience usable for more than a short burst. Plus, if the diagrams are any indication, the form factor of this invention in particular has a similar look and feel to the Humane AI pin palm projector, potentially presenting another competitor if glasses with this function make it onto people’s faces.