One of the most popular uses of artificial intelligence to date is to use it to predict things, using algorithms trained with historical data to determine a future outcome. But popularity doesn’t always mean success: predictive AI leaves out much of the nuance, context, and cause-and-effect reasoning that goes into an outcome; and as some have pointed out (and as we’ve seen), this means that predictive AI’s “logical” responses can sometimes be disastrous. A startup called causaLens has developed causal inference technology — presented as a no-code tool that doesn’t require a data scientist to introduce more nuance, reasoning, and cause-and-effect sensitivity into an AI-based system — which it believes can solve this problem.
CausaLens’ goal, CEO and co-founder Darko Matovski said, is for AI to “understand the world the way humans understand it.”
Today, the startup is announcing $45 million in funding after some early success with its approach, growing 500% in revenue since it came out of stealth a year ago. This is described as a “first close” of the round, meaning it is still open and may grow in size.
Dorilton Ventures and Molten Ventures (the VC renamed Draper Esprit) led the way, with previous backers Generation Ventures and IQ Capital, and new backer GP Bullhound also taking part. Sources tell us the London-based causaLens is valued at around $250 million.
CausaLens’ customers currently include healthcare, financial services, and government organizations, as well as a number of other industries, where the technology is being used not only for AI-based decision-making, but also to add more cause-and-effect nuance. in achieving results. Critical, the
An illustrative example of how this works can be found at the Mayo Clinic, one of the startup’s clients, which uses causaLens to identify cancer biomarkers.
“Human bodies are complex systems, and so if you apply basic AI paradigms, you can find any pattern you want, correlations of any kind, and you won’t get anywhere,” said Darko Matovski, the CEO and founder of the startup, in an interview. “But if you apply cause-and-effect techniques to understand the mechanics of how different bodies work, you can understand more of the true nature, of how one part affects another.”
Given all the variables that could be involved, this is the kind of big data problem that’s nearly impossible for a human, or even a team of humans, to compute, but it’s a table game for a computer to work through. While not a cure for cancer, this kind of work is an important step toward considering different treatments tailored to the many permutations involved.
CausaLens’ technology has also been applied in a less clinical way in healthcare. A public health agency from one of the world’s largest economies (causaLens can’t publicly disclose which) has used its causal AI engine to determine why certain adults have opted out of getting Covid-19 vaccinations so the agency could better strategies thinking about getting them on board (plural “strategies” is the operative detail here: the whole point is that it’s a complex issue with a number of reasons depending on the individuals in question).
Other clients in areas such as financial services have used causaLens to inform automated decision-making algorithms in areas such as loan evaluations, where previous AI systems introduced bias in their decisions when only using historical data. Hedge funds, meanwhile, use causaLens to better understand how a market trend might develop to inform their investment strategies.
And interestingly, a new wave of customers may be emerging in the world of autonomous transportation. This is one area where the lack of human reasoning has hindered progress in the field.
“No matter how much data is fed into autonomous systems, it’s still just historical correlations,” Matovski said of the challenge. He said causaLens is now in talks with two major auto companies, with “lots of use cases” for its technology, but especially on autonomous driving “to help the systems understand how the world works. It’s not just correlated pixels that relate “But also what the effect will be if that car slows down at a red light. We bring reasoning into the AI. Causal AI is the only hope for autonomous driving.”
It seems like a no-brainer that those who use AI in their work want the system to be as accurate as possible, which begs the question of why the brilliant improvement in causal AI isn’t built into AI algorithms and machine learning in the first place.
It’s not that reasoning and answering ‘why’ weren’t priorities in the beginning, Matovski explained – ‘People have been investigating cause-effect relationships in science for a long time. You could even argue that Newton’s equations are causal. It’s super fundamental in science,” he said, but it’s that AI specialists couldn’t understand how to teach machines to do this. “It was just too hard,” he said. “The algorithms and technology didn’t exist.”
That started to change around 2017, he said, as academics began publishing the first approximations, exploring how to represent “reasoning” and cause and effect in AI based on finding cues that contributed to existing results ( rather than using historical data to determine results), and build models based on that. Interestingly, it’s an approach that Matovski says doesn’t require huge amounts of training data to work. The CausaLens team is very busy with PhDs (you could say that the startup really ate its dog food here: it considered 50,000 resumes while assembling its team). And this team took that baton and started working on it. “Since then, it’s been an exponential growth curve” in terms of discovery, he said. (You can read more about it here.)
As you might expect, causaLens isn’t the only player exploring how advances in causal inference can be harnessed in larger projects that rely on AI. Microsoft, Facebook, Amazon, Google and other big tech players with substantial AI investments are also working on the field. Among startups, there’s also Causalis that focuses specifically on the possibility of using causal AI in medicine and healthcare, and Oogway looks set to build a causal AI platform aimed at consumers, a “personalized AI decision assistant” like the describes itself. All this speaks to the possibility of developing more and quite a huge market for the technology, covering both specific commercial and more general use cases.
“AI needs to take the next step toward causal reasoning to unleash its potential in the real world. causaLens is the first to use Causal AI to model interventions and enable machine-driven introspection,” Dorilton Ventures’ Daniel Freeman said in a statement. “This world-class team has built software with the sophistication to win over serious data scientists and the usability to empower business leaders. Dorilton Ventures is very excited to support causaLens in the next phase of its journey.”
“Every company will use AI not just because it can, but because it has to,” added Christoph Hornung, investment director at Molten Ventures. “We at Molten believe that causality is the key ingredient needed to unlock the potential of AI. causaLens is the world’s first causal AI platform with a proven ability to turn data into optimal business decisions.”