When it Comes to AI, it’s Time for Companies to Think Bigger

845142_Bitvore images_4_092820-2

Right now, much of the corporate and governmental investment in artificial intelligence is iterative. We're fond of citing that 50% of AI products are concentrated in just two product categories—chatbots and fraud detection. You can see that these two fields exert a sort of gravity well when it comes to innovation and new product introduction. "Startup X got a good Series A round because it makes a good chatbot, so I'm going to found Startup Y and make a better chatbot so I'll get even more investment."

Companies that do this may one day be acquired by Microsoft, but they won't ever become the next Microsoft. The only way to do this is to take risks, find AI applications that no one else is thinking of, and put their money where your mouth is. In other words, it's time for companies to invest in AI moonshots.


Why Are Moonshots the Way to Go?


From 1989 until 1998, Greg Bolcer—the Chief Data Officer at Bitvore—worked as part of a DARPA-funded research lab at UC Irvine. DARPA (the Defense Advanced Research Projects Agency) has a storied history of turning big ideas into big realities. The internet you're reading this article on grew out of a DARPA initiative known as ARPANET. If you took a car ride using GPS, then your navigation was assisted by a DARPA brainchild. Lastly, if you've been impressed (or terrified) by a Boston Dynamics robot, DARPA enabled your entertainment.


"When I worked for DARPA," says Greg, "the biggest thing is they tell you not to do stuff that companies could think of. They tell you to look at 15 years out and figure out what's really going to change the world in 15 years."


Suppose you're an investor, an innovator, or an established company with a big pocketbook. In that case, moonshots are your chance to create a product that changes lives on the scale of the internet, the iPhone, or the automobile—and AI is a field that's ripe for exploration. We've only scratched the surface of what artificial intelligence can do. While AI products have achieved widespread adoption, there's no single product that one can point to as a groundbreaking AI implementation instead of an iteration on something that has gone before. In short, we've underestimated what AI can do. It's time to go forward, fail boldly, and succeed triumphantly.


How Does an AI Moonshot Work?


The thing about DARPA is that it's largely state-funded, and many of its biggest innovations come from working in concert with university research departments. In a similar vein, a recent report from the AI Council—a British academic group—also promotes the idea that the UK government should fund artificial intelligence moonshots using public-sector funding, research fellowships, and Ph.D. programs. The implication is that moonshots are something that governments do.


Here's the thing: moonshots related to AI aren't—and probably should not—be the sole purview of governments and large corporations. This is because AI, unlike building a continent-spanning internet or a global array of navigation satellites, is relatively democratic. This is to say that a small company with good developers can make as much progress working on AI as an academic research program. Training data and cloud infrastructure are relatively inexpensive, so there's an opportunity for very small companies to create very big strides.


So, how can AI companies learn from DARPA and other innovators to successfully pull off a moonshot?


The first lesson is to fail fast, pivot early, and try multiple pathways. Although DARPA is known for many world-beating innovations, its hit rate is surprisingly low. For every one project that changed the world, dozens exploded on the runway. This is how it should be—if a project fails to achieve initial success, pivot and try something new before you experience sunk costs.


Second, incrementally capitalize on what works. ARPANET started with four nodes. It took a year to grow to nine nodes and a decade to pass the two hundred mark. Alternatively, consider how polished the first iPhone was at its release and how easily Apple could have simply put its iTunes software on a flip phone and called it a day. Creating a moonshot product also mean going beyond the minimum viable product approach.


Finally, watch this space. Bitvore is constantly working on new approaches to machine learning, artificial intelligence, and natural language programming.



Subscribe to Bitvore News Blog Weekly Email

Recent Posts


See all