Were you unable to attend Transform 2022? Check out all of the summit sessions in our on-demand library now! Watch here.

Everyone and their mother is racing toward AI these days. Spending on the technology was already at a hefty $640 million in 2016, and analysts predict it’ll reach $37 billion by 2025. While we’re all ultimately working together, there’s still fragmentation in the industry as competing schools of thought grapple for supremacy. How do you build a thinking machine?

Thousands of developers have worked for decades to discover the right methodology to create machine learning and the singularity. Some believe neural networks are the key to unlocking AI’s potential, while others are exploring probabilistic programming. Many more schools of thought exist, from fuzzy systems to evolutionary computation, chaos theory, Gaussian processes, computational creativity, and more — all with their own supporters and detractors.

Each group is certain its discipline holds the key to unlocking the mystery. But it’s likely a combination of these technologies that will solve the puzzle. Not all human brains work the same, so why should every AI?

Combining the competition

With a neural network, an AI system — such as Google’s Assistant, Apple’s Siri, or Amazon’s Alexa — creates a platform other third parties can tie into. The memory of the neural network increases as more simple systems are connected, allowing it to detect more patterns and appear smarter. When you ask your Google Pixel to play ’80s music on Spotify, you’re tapping into Google’s neural network.


MetaBeat 2022

MetaBeat will bring together thought leaders to give guidance on how metaverse technology will transform the way all industries communicate and do business on October 4 in San Francisco, CA.

Register Here

Probabilistic programming creates intelligence by doing what its name suggests: letting reasoning, coupled with the probability of various outcomes, determine a solution.

Both neural networks and probabilistic programming represent how the human mind works. Our minds have systems of information we can access to come up with conclusions. We analyze data, predict potential outcomes, and make decisions. A single individual learns, remembers, and draws conclusions in dozens of ways depending on which methodology is best suited for the decision at hand.

Researchers are realizing it’s the combination of all these disciplines that allows for a genuine thinking machine. Each has yielded amazing results independently, and by combining them, one technique’s strengths augment another’s weaknesses. A connectionist system, for example, requires huge amounts of properly tagged data to recognize what it sees, but probabilistic programming mitigates this concern.

This approach solves one of the biggest problems facing AI: anticipating uncertainty.

Joining the fray

Even with AI getting so much buzz, jumping on board isn’t as easy as you’d think. Unless your goal is to build a company that utilizes fuzzy systems or Gaussian processes, there’s little way your specific website or application can create a viable input-output contribution to the discipline. There’s still hope to join the fray, however, by following three simple steps:

  1. Integrate into a neural network. Because they require large amounts of data to connect to, neural networks offer the best opportunity for entrepreneurs to get involved outside of founding another AI think tank. You may not develop and deploy the next Facebook M or Samsung Bixby, but you can certainly integrate into them. Whether it’s Uber connecting to Google Home or Spotify integrating into Amazon Alexa, these networks get smarter, faster, and stronger as more companies expose their proprietary technologies to the broader system. Each advancement builds on the past achievements, causing each competitor to advance AI faster.
  1. Expose as much as possible. These networks have APIs and reams of documentation to help development teams take advantage of what they offer. And because no one knows which ecosystem will win in the end, integrating into as many as possible is preferred. Ordering a car through your Echo is nice, but how much better is it to get regular updates on its arrival through all your devices? Further, this two-way stream between the AI system and specific providers increases user adoption, reduces frustration, and improves quality of life. Spotify’s 50 million monthly subscribers are there because of the company’s AI efforts like Discover Weekly, which personalizes playlists according to your tastes and the listening habits of other users.
  1. Develop AI-specific features. Don’t just expose product features to Alexa or Siri; make your product appear magical by tying into their massive learning power to become predictive. Adapting existing functionality is a start, but the world really starts to open up when facets of a system can be enjoyed only when tied to an AI system. Imagine an integration with OpenTable that knew your favorite foods, wines, and cuisines. That system could link with your AI of choice to make recommendations on the fly, sync reservations to your calendar, and suggest friends to join you.

What’s currently missing from the AI arms race is the combination of all these features. Saying “OK Google, set up a date night for me and my wife” will one day cover everything. The AI assistant will schedule an Uber, reserve a table at your favorite restaurant, find a movie you’ll both enjoy, and even find a frozen yogurt place for dessert.

No taps, swipes, or iFuss — just a combination of every AI methodology integrated into one seamless experience. That’s the next evolution of AI.

Q Manning is CEO of Rocksauce Studios, which crafts custom mobile apps for all platforms.

VentureBeat's mission is to be a digital town square for technical decision-makers to gain knowledge about transformative enterprise technology and transact. Learn more about membership.