The AI businesses know isn’t really the AI that excites researchers
BUSINESS leaders might argue that artificial intelligence (AI) is here and that organizations must start adopting the technology to improve their business, and they’re right — but there’s another side to the story.
The intelligent AI solutions we see today are only tools, “narrow or brittle AI systems” as researchers call them, and are only able to perform one task to perfection.
These systems are phenomenally valuable in the world of business because they can help maximize efficiency and minimize human error while simultaneously also reducing costs and downtime.
Speech recognition, character recognition, data analysis, and translating between languages are some easy examples.
Although narrow or brittle AI systems are exciting to businesses because of the value they can get out of implementations immediately, researchers — in universities or within companies — concern themselves with general AI.
“General AI is computer-based intelligence that mimics human intelligence. Brittle AI is unable to do anything outside the area or domain it has been trained on. It can’t adapt like humans. The idea is that general AI will be able to adapt and learn more quickly,” Portland State University Professor Melanie Mitchell told Tech Wire Asia.
In her book Artificial Intelligence: A Guide for Thinking Humans due for release in October this year, Mitchell talks about the differences in vivid detail and answers many questions about the world of AI and future scenarios that the technology enables.
Today, general AI doesn’t have any business use cases because the technology is still in the laboratory and researchers are still grappling with the basics.
However, once ready, it could really transform the world of business — especially when it is supplemented by quantum computers, real-time data from the internet of things (IoT), and reliable high-speed wireless connectivity via 5G networks.
According to one expert, with general AI in play, shortages in programming talent, management of complicated compliance requirements, and unavailability of strategic consultants could be a thing of the past.
Like brittle AI, general AI won’t make existing jobs redundant — they’ll only make it easier for computers to pick up the slack when it comes to managing the businesses.
Gaining a better understanding of brittle AI
Simply put, general AI is what we see in science fiction movies — and building that system is a dream most researchers interested in the field are pursuing.
Brittle AI’s biggest flaw is that it needs to be fed a lot of data about something before it can learn from it. It’s why driverless cars are so complicated to build. General AI is expected to solve that problem.
“Everything must be tagged. Millions of examples must be tagged before the computer can begin to recognize a category with high confidence.”
The problem Mitchell touches upon briefly has been making headlines for the past year or so with near-ready Level 5 autonomous or driverless cars failing to recognize “all” living things — causing companies to stall their commercial debuts.
“We find that our (human) intelligence is more complicated than we know. There’s a lot of nuance to what human beings do and how our intelligence works below the surface, and computers today just don’t learn like that.
“Take a child and show it what a living thing looks like, and the child learns immediately. We call it active, unsupervised learning. The child can then go on to identify living things by itself.
“Brittle AI, on the other hand, needs a lot of strictly supervised learning before it can be tasked with a job.”
Why general AI needs common sense
Given the fact that universities, governments, and companies are already pouring in billions into developing general AI, it’s important to ask why progress is so slow.
“Well, the reality is that there’s a lot that we don’t understand about the way human intelligence works. In fact, one of the most important things in the AI research arena right now is the understanding of common sense and understanding of the world in the context of AI.
“Human beings have a lot of knowledge about the world around us and we have common sense. Children are born with a lot of that intelligence. We see shards of glass on the road or a flock of birds and we know that the first won’t fly away but the second will — and so it’s safe to drive through in the second case but not the first.
Till we can figure out a way to do that with computers, general AI will continue to be a challenging problem.
Moving from brittle AI to general AI
Talking to Tech Wire Asia, Mitchell reveals that there’s a lot that researchers need to do to accelerate the journey from brittle AI to general AI.
However, the reality is that businesses, as things stand, don’t play a big role in that.
“Businesses tend to be focused on immediate results. Hence, their research efforts are focused on short terms goals that will yield useful solutions immediately.
“Except for big industry leaders such as Google and Microsoft, researchers in company labs don’t really have the bandwidth or power to work on long-term projects — which is why they’re not fuelling the move to general AI.”
According to Mitchell, businesses might find that working with researchers in universities is a great idea and can really help them get a leg up on the competition in the long-run when bigger projects — such as general AI — actually pan out and begin to shape the world around us.
While we don’t really know what AI can help us do tomorrow, speaking with Mitchell proves that there’s definitely interesting opportunities coming up that have the potential to entirely transform or even revolutionize business and customer applications.
- Rockwell Automation is striving in SEA, with huge potential in Vietnam, Malaysia
- Data protection is vital: 85% of Singaporeans concerned about how companies use their data
- HPE delivers the world’s fastest, energy-efficient supercomputers at SC22
- Game on: iion launches ‘immersiion’
- Making regulations work for the morphing world of Artificial Intelligence