Neuromorphic Computing: Discover the Future of AI

Well, that's because it kind of is. While we are not debating the many advantages of using AI for many of our daily activities (especially in a business context), the reality is that AI is not as sophisticated as we like to think. The main problem lies in the underlying approach of current AI algorithms, which depend almost entirely on the training phases they have obtained.

I'm not just talking about the complicated balance that AI developers need to strike to avoid their algorithms falling victim to overfitting or underfitting (at least, that's not all I'm suggesting). I'm also talking about the seemingly impossible autonomy these algorithms can aspire to (something extremely well exemplified by the continued failure of self-driving cars).

In other words, these algorithms may appear to be learning, but in reality, they are adapting what they “know” from their training sessions to the new contexts they encounter. And that's the limiting aspect of it all. Why? Because there is no way for a development team to train their AI algorithms with every possible situation they may encounter.

Does this mean that AI is not the technology of the future, as many of us have said in the past? No, it doesn't really mean that. But to truly embrace this position, AI engineers need to radically change how they build algorithms. Fortunately, they are doing this in the form of neuromorphic computing.

What is neuromorphic computing?

Neuromorphic computing (also known as neuromorphic engineering ) aims to replicate the way the brain works through a series of interconnected chips. Each chip “behaves” like a neuron, organizing itself, communicating and interacting with other chips. What researchers working on this are trying to achieve is harnessing insights from neuroscience to build an artificial human brain.

I know it sounds crazy, incredible, fake and even a little dangerous. But this is the path that researchers believe will take us forward in our AI ambitions. What is most important is that we are already following this path. Leading the way is Intel with its Loihi Research Chip and its recently released open source framework, Lava.

This doesn't mean we'll be close to having neuromorphic computers anytime soon. The new Loihi 2 Chip is the most powerful chip of its kind and has “only” 1 million “neurons”. To put this into perspective, the human brain has about 100 billion neurons. Intel hopes to improve this architecture, but understands that this is extremely difficult, especially when it comes to developing software for it. That's why they launched Lava – to attract engineers to create applications for Loihi.

Even with this reality check, neuromorphic computing is a truly exciting premise. In fact, experts argue that this is the only way we can truly achieve the AI ​​goals we have set for ourselves.

I'm talking about those goals that go beyond the mere analysis of large data sets, especially those related to autonomous robots that can think and learn for themselves. This is because the neuromorphic architecture abandons the synchronous, structured processing of CPUs and GPUs in favor of asynchronous, event-based bursts.

This allows neuromorphic chips to process information much faster and in a less data-intensive manner, a key to dealing with ambiguous information in real time. In fact, neuromorphic systems will be crucial to the next generation of AI, as they will allow algorithms to become more adept at dealing with probabilistic computing, which implies noisy and uncertain data. Neuromorphic computing could also theoretically help with causality and non-linear thinking, but this is nothing more than an engineer's dream in the current scenario.

What are the challenges of neuromorphic computing?

If you haven't heard about neuromorphic computing, you're not alone. While it's not a particularly new concept, it's only recently that researchers have been able to start working on hardware that could actually bring the concept to life. This is not everything. Because neuromorphic systems work in such a complex way, understanding them is a challenge, let alone putting them to work.

This means that the first challenge for neuromorphic computing is to gain more visibility. Engineers working in the field of AI might have heard about it, but most of them still work with the traditional approach of AI algorithms. If neuromorphic computing is to gain critical mass, it will need as many creative minds pushing for it as it possibly can.

Unfortunately, this is far from the only challenge. As you can probably imagine, developing a replica of the human brain is a difficult task. We still haven't fully figured out how the brain works, so trying to build an artificial brain from the missing puzzle pieces could be tricky. Although we have more understanding than ever about our brains, neurobiologists and neuroscience as a whole still have many mysteries to solve .

Fortunately, researchers building neuromorphic chips can mean a mutually beneficial relationship between them and neurobiologists. As developers delve deeper into building their “artificial brain,” neuroscientists can begin to verify hypotheses and formulate new ones. Likewise, neurobiologists can inform researchers about new developments so they can use new approaches for their neuromorphic chips.

Another major challenge of neuromorphic computing is the dramatic changes it will bring with it and radically reshape the way we understand computing norms. Instead of following the von Neumann model (which separates memory and processing), neuromorphic computing will introduce its own standards.

A great example of this is how modern computers handle visual information. Following von Neumann's model, computers today see an image as a series of individual units or frames. Neuromorphic computing would throw out this notion in favor of encoding information as changes in the visual field over time. It is a radical departure from our current standards, which will force engineers to learn to think under this approach.

As if that weren't enough, neuromorphic computing will need new programming languages ​​and frameworks, as well as more powerful memory, storage, and sensory devices that make the most of the new architecture. As the relationship between memory and processing will change, so will the integration between the devices that are part of these processes. As you can see, this is a paradigm shift and we are in the early stages.

A new path to follow

Neuromorphic computing is starting to receive some attention at a time when another promising avenue has already been touted as AI's next big thing: quantum computing. But unlike quantum computing, the requirements for neuromorphic computing are not as demanding. Where quantum computers require temperatures close to absolute zero and insane power demands, neuromorphic computers can easily function under normal conditions.

This naturally tips the scales towards neuromorphic computing, mainly due to the practicality and potential for integration of this architecture into all types of devices. We shouldn't get ahead of ourselves, however. Both quantum computing and neuromorphic computing are still far from commercial applications, so we will have to make do with the AI ​​we have today.

However, it's understandable if you're excited about the prospect of neuromorphic computing. Truly intelligent devices and robots have the potential to completely change the way we live, and although they may not be available in the near future, neuromorphic computing offers us a new way forward. In fact, neuromorphic systems seem like the true future of AI, as they promise to finally make all our AI-related dreams come true. Only time will tell if this is the case.

If you liked this, be sure to check out our other articles on AI.

• Not adopting AI? You could be hurting your productivity
• Unmatched Accuracy: Optimizing Business Analytics with AI and BI
• Three common pitfalls you need to avoid in your AI implementation strategy
• 5 problems with AI that remain unsolved
• Psychometrics and AI: How to Empower Talent Acquisition