AI vs OI: What's The Difference?

We’ve seen an incredible emergence of AI, deep learning, and transformers in the past few years. But alongside AI, a new type of intelligence has emerged, and it’s called organoid intelligence, or OI for short. So, AI vs OI: what’s the difference?

OI relies on neural stem cells instead of computers to work. But the key difference between AI and OI is in how they learn. While AI learns by studying patterns in vast amounts of data, progressively getting better, OI learns on its own through feedback, allowing it to learn and adapt more quickly.

If this is the first time you hear about OI, you’re in for a ride. Join me as I explore this new technology dubbed DishBrain. I’ll cover everything you need to know about OI and how it stacks up against traditional computer-based AI.

What Is Organoid Intelligence (OI)?

Organoid intelligence (OI) is a type of biocomputer that was recently developed by researchers at Johns Hopkins University and learned how to play Pong. It’s essentially neural stem cells arranged on a microelectrode array (MEA), which allows the cells to interact with electricity.

I know; I threw a lot at you at once. So, let’s explain how the researchers did it using plain English.

Essentially, DishBrain consists of a sensory region and two motor regions.

The regions consist of electrodes and neural stem cells. The electrodes are just for communicating with the outside world.

The “brain” part is the neural cells that allow the rest of the nervous system to develop in normal conditions. But this is an in vitro (outside the body) study, so the cells are grown from human skin or blood cells and then reprogrammed.

Now, back to the two regions. The sensory region receives information directly from the Pong game through electrical signals. In this case, there are 8 electrodes. You can think of it as any other sensory organ like the eyes or ears, but it’s much more primitive.

The two motor regions receive the information from the sensory region and move the paddle appropriately. One region moves it up, the other moves it down.

The scientists exploited an interesting characteristic of the human brain — it hates unpredictable outcomes.

So, they conditioned the brain by punishing it with random stimulation from all 8 electrodes for 4 seconds when it missed the ball.

Conversely, hitting the ball would result in all electrodes being fired at once. That’s a predictable outcome; our neurons are strongly biased toward predictability.

And that’s essentially how DishBrain learned how to play Pong.

Do you know what the best part is?

OI had no previous knowledge of the rules of Pong. DishBrain learned strictly through negative feedback.

It has arguably achieved sentience as it became responsive to its environment. Note that it’s far from being conscious, though.

To see a visual representation of DishBrain and how it works, check out this 1-minute video by New Scientist:

What Makes OI Different From AI?

OI is completely different from AI because it doesn’t rely on computing power to solve problems. AI relies on a combination of data and iterative processing and algorithms to learn. Conversely, OI learns similarly to the human brain, which is through stimulation, feedback, and experience.

One of the most impressive stats regarding OI is that it learned significantly faster than AI from doing the same task. It took OI 10–15 rallies of Pong to perform at the same level as AI did after 5,000 rallies.

Moreover, AI needs tremendous computing power and, consequently, electricity to learn. It also needs swaths of data as it slowly and iteratively learns to achieve better results.

But OI works in a completely different way. Since we’re talking about brain cells, they learn by firing electrical impulses between neurons and strengthening the synapses.

It remains to be seen how powerful OI can really be, though. OI is a young technology, so it’ll take some time to see everything it can do.

Theoretically, OI will excel and outperform AI in the following situations:

Rapidly changing environments
When data is scarce
In low-power, low-cost environments
Unique problems without a precedent
When a quick solution is needed

OI is naturally intuitive and knows how to find solutions to problems extremely quickly. In a way, OI is the exact opposite of AI. So, it’s likely that once we have developed OI, it’ll be used in all situations where AI is struggling.

On the other hand, it has some drawbacks compared to AI. We already know that when we compare ourselves to AI — the ability to adapt quickly also means we forget previous solutions rapidly.

And AI is already great at natural language processing, voice recognition, autonomous driving, etc. AI’s capabilities will only improve as computing power increases, so it’s winning the race to AGI (artificial general intelligence).

It also faces significantly worse ethical challenges than AI.

While AI is a computer-based lifeform (or isn’t really a lifeform at all), OI bears a lot of resemblance to human intelligence. Researchers have to figure out whether OI can feel and think to prevent human suffering, which is a tough nut to crack.

We’ll also have to figure out how to avoid biases and discrimination with OI. That’ll require more than just changing a few parameters like with AI.

Why Was OI Developed?

OI was developed to enable researchers to study the human brain outside the human body. By understanding biological learning, we will be able to use that to our advantage. OI is more energy and data-efficient compared to AI, which means that it could provide solutions where AI falls short.

As I explained above, OI needs very little electricity to function. Right now, it just needs some tiny electrodes to “whip it” into learning solutions. Compare that to a modern gaming computer that can consume upwards of 800 watts, and you’ll quickly understand OI’s benefits.

Modern AI models could easily use up the world’s most powerful supercomputers, and it wouldn’t be enough. Supercomputers can cost millions of dollars to design and build, and I’m pretty sure that DishBrain didn’t cost nearly as much. Not to mention that computer hardware is constantly increasing in power, so there’s always something better and more expensive to buy.

Theoretically, a super-powerful OI could adapt to solve any problem in minutes, if not seconds.

Also, AI needs a lot of data to learn, which is a problem both in terms of resources and ethicality.

Data takes up a lot of space, which can get expensive in terms of hardware. The human brain has an unknown storage capacity — it’s speculated that it’s over 2.5 petabytes.

In the Pong example above, OI doesn’t need data; stimulation is all it takes to learn. Brains are natural problem-solvers.

But, perhaps most importantly, by studying human brain cells in vitro, we may learn something about our own brains.

It has some major implications in the field of medicine, such as:

Regenerative medicine
Drug research and testing
How diseases occur

Will OI Surpass and Outperform AI?

Forbes’ headline says that OI might power our computers in the future.

However, I don’t think that’s going to be the case. Computers excel at performing complex calculations and executing algorithms rapidly, tasks the human brain isn’t naturally equipped to handle.

OI is unlikely to surpass and outperform AI. OI and AI are good at solving problems but have completely different approaches. AI is great for going through vast amounts of data quickly, task automation, natural language processing, and much more. OI is good at adapting to changing environments.

Remember that much of this is speculation based on what we know.

It would be interesting to compare OI and AI regarding autonomous driving. AI is already decent at driving cars, but you’ve probably seen clips of Teslas making major errors and crashing.

Driving a car requires quick adaptations to ever-changing environments. This will remain true until we figure out a way to have all cars communicate and drive. So, OI could excel at that task if we develop the technology quickly enough. OI is good at adapting and learning quickly, so it should be a breeze to teach it how to drive compared to AI.

Final Thoughts

Organoid intelligence is an interesting new technology. It blurs the lines between AI and the human brain.

OI learns much faster than AI while using fewer resources. This has significant implications for the near future.

AI vs OI: What’s The Difference?

What Is Organoid Intelligence (OI)?

What Makes OI Different From AI?

Why Was OI Developed?

Will OI Surpass and Outperform AI?

Final Thoughts

Sources

Recent Posts