Apple’s Trinity AI code without code handles complex spatial data sets

That Transform Technology Summits launch on October 13 with Low-Code / No Code: Enabling Enterprise Agility. Register now!

Apple has slowly but surely made a name for itself in the low-code / no-code movement. In July, the Cupertino-based company announced the launch of Trinity AI, a no-code platform for complex spatial datasets. Trinity enables machine learning researchers and non-AI developers to tailor complex spatiotemporal datasets to fit deep learning models.

Back in 2019, Apple unveiled the SwiftUI, a programming language that required much less coding than the Swift language. With the release of Trinity, Apple is doubling its efforts to significantly lower the threshold for non-devs and non-ML devs.

Fusemachine CEO Sameer Maskey, who also teaches AI as an adjunct associate professor at Columbia University, sees Trinity as a great way for developers to use machine learning in their apps. “Initially, I see Trinity being used by devs who already create apps for iOS, but who do not know machine learning, so they can incorporate spatial datasets into their work,” Maskey told VentureBeat.

We asked Maskey to give VentureBeat its bid on Apple’s platform and what it means for the future of AI and the low-code / no-code industry. This is a literal transcript of the interview.

VentureBeat: What sets Trinity apart from other AI platforms without code?

Sameer Maskey: It’s not that groundbreaking, really. By creating a similar system, the difference is that it is more focused on geospatial data, such as maps and moving objects. Many people are trying to build apps with geospatial data for a phone. If you do not know machine learning, but if you have a background building app, you can now do it with Trinity.

Let’s say you’re trying to build an app that recommends the best eateries in an area. Let’s also say you have access to how many people are going to that specific place. Before, you had to collect all the data and stream the collected data and build it on a server or any system you used. With neural networks, you experiment with many different models. For example, you will find a model that predicts what are the best places to eat; you should know all the different dev-ops behind it. All of this gets easier with Trinity because you dump the data and set goals for what you want it to do and do all the training; it does everything for you behind the curtains.

VentureBeat: What are Apple’s goals with this platform?

Maskey: I would not say that it is so, so, so groundbreaking in the sense that they are creating a similar system as other systems out there that have tried to do something similar. I guess the difference for Trinity is that it’s more focused on geospatial data, especially things related to maps and moving objects in maps. Especially with the phone, there are many people who tried to build all kinds of applications using geospatial data. And if they try to build an app on top of the iPhone first, for some of them, it may be easier to use Trinity than other platforms because it’s probably very closely integrated. Even if you do not know machine learning, but you have a framework for building apps, you can quickly tap the Trinity platform to build models for different ML work.

VentureBeat: Can you give us an example of how Trinity would work with geospatial apps?

Maskey: Of course. Let’s say you’re trying to build an app that automatically recommends the hottest eateries to go to. Let’s say it’s in a small part of town. And you somehow have access to the data from people in that place, e.g. How many people there are, how many people are going to that place and so on. You basically get to predict what the hot joints are and what hot joints you would like based on your preferences.

And let’s say you’re taking all this streaming data, all this location data. You would build it on your computer or on a server or any system you use – many people write code in Jupiter notebooks – you try many different machine learning algorithms. You try, let’s say, even with neural networks, many different types and sizes of neural networks. You keep experimenting with many, many different models and then say, OK, this is the model that makes the best prediction of what the next popular food bird will be. Then you have to produce it. And let’s say that your products, AWS or GCP, you should know all the development facilities behind them in order to take it to production. And then create an API. All of this gets easier in Trinity because Trinity allows you to just dump the data and set goals for what it will do. And it will figure out what machine learning algorithm to use, what kind of neural network architecture you should choose to perform all the training, and come up with all the production.

VentureBeat: Can Trinity Really Be Used In A Professional Setting? Can we rely on its prediction models, or does it need fine-tuning?

Maskey: Trinity and other similar platforms are professional systems and for some issues they work really well. They are good enough even for production quality systems. But in many cases, they are not in the sense that they may provide 5% less accuracy than an engineer who would adjust at the very low level to how the machine learning system is built. And they are able to squeeze out an additional 5% accuracy, which can be a difference in the competitive world where you charge money for the APIs.

VentureBeat: Where do you see the future of platforms like this? Low-code / no-code AI?

Maskey: AI is overhyped right now. I believe that more and more of these platforms will become more and more comprehensive in terms of being able to support more than one other kind of machine learning systems within it and more and more information about the algorithms we need within it. Hopefully the accuracy will be better on different tasks. They will probably become more specialized at some point. Trinity is already a more specialized version of this kind of systems, which is more focused on geospatial data, but my guess is that they will also later expand beyond geospatial data.

I generally think that more platforms will be launched and they will become more and more specialized. And if they get accuracy to a level where they are pretty much on par with what developers are capable of now, then it really becomes a transforming tool. Because at that point, there won’t be a need for many machine learning engineers for many of the current AI bids.


VentureBeat’s mission is to be a digital urban space for technical decision makers to gain knowledge about transformative technology and transactions. Our site provides important information about data technologies and strategies to guide you as you lead your organizations. We invite you to join our community to access:

  • updated information on topics that interest you
  • our newsletters
  • gated thought-leader content and discount access to our valued events, such as Transform 2021: Learn more
  • networking features and more

sign up

Leave a Comment