Horns or Halos: The true nature of AI

In Disney’s Fantasia, a sorcerer returns to his workshop to find that his apprentice has unleashed forces that have quickly spiraled out of control, threatening to destroy everything in sight; the magician takes charge to restore order. Who on earth today will fill the sorcerer’s role and catch the AI lightning in a bottle? Can AI indeed be regulated? 

Regulating AI is like nailing a cheesecake to the wall. Perched on the threshold of a technological revolution – a Cambrian explosion, some liken it to – it is essential to dispel misconceptions, establish clear boundaries, and understand the potential of this powerful tool. Just as the Internet evolved from a medium of mistrust to a productivity enhancer across industries, AI is poised to undergo a similar transformation. However, like any potent tool, AI’s impact will be defined by how we choose to wield it.

The journey of the Internet offers a valuable lesson. Initially, it was conceived as a free medium for information exchange, prioritizing democratization over commercialization. However, the culture of ‘free’ demanded a business model – advertising. This led to the rise of targeted advertising, incentivizing platforms to grow their user base and keeping us glued to the internet. Inadvertently, this also fueled the spread of misinformation, hyperpartisanship, and new forms of bullying and bad behavior.

Today, AI stands where the Internet once did, misunderstood and mistrusted. As the fog in the air clears, the roadblocks stand exposed.

Bias in AI

Over 20,000 cases falsely flagged for unemployment benefit fraud by Michigan’s MIDAS algorithm. False arrest records (‘hallucinations’, they call it!) by automated tenant screening tools pushing low-income populations into no-housing situations. Potential revocation of college admissions, wrongful arrests due to incorrect facial recognition tools, harsh sentences from judges due to skewed predictions on whether a given criminal will reoffend… the list abounds of the ills of bias in AI. Teething troubles aren’t new to deployments of new technologies. But the good news is the first adopters have the opportunity and responsibility to set the norms for its use. 

Regulators across continents and countries are already actively building guardrails to make AI safer and productive for everybody. The AI Act, spanning 100 pages, has received recent approval from the European Parliament. This statute aims to proactively prohibit applications judged to possess “unacceptable” risk levels. It mandates pre-approval and licenses for certain applications within the EU and imposes significant fines on developers for various infractions. Chinese regulators are taking swift steps towards establishing clear guidelines on the permissible and impermissible operations of AI technologies. In the United States, multiple actors are jostling to lead the regulation of AI. Aside from the AI Bill of Rights from the Biden Administration itself, Senate Majority Leader Chuck Schumer in the Congress, the  National Telecommunications and Information Administration (NTIA) at the Department of Commerce, the feisty Lina Kahn at the Federal Trade Commission and provincial efforts in some 17 states are all moving in parallel to build the guardrails.

At Vue.ai’s AI Transformation Retreat REBUILD ’23 Jamaica, Prof. Amit Arora (Columbia Univ) explained the then newly instituted Bias Audit Law instituted in the state of NY, USA.

At Vue.ai’s AI Transformation Retreat earlier this year (REBUILD ’23 Jamaica), Prof. Amit Arora (Columbia Univ) contextualized the Human-In-the-Loop intervention, being implemented by the State of New York. Regulating AI isn’t a quest to tame rogue tech; it’s a reality check for our wild human imaginations. Picture AI as a vast freeway. You can mold it to guide collective behaviors, but, like any highway, you need mechanisms for speed demons (think abuse and bias) and surprise potholes (hello, accidents). AI developers must tap into their inner mad scientists, predicting how the system might play tricks and baking in transparency and accountability. Enter the stage: regulatory frameworks with audits and AI police – because even tech needs a traffic cop.

Won’t AI take my job away?

Fears of high-tech robots replacing front-line workers may not have totally subsided yet. But AI isn’t a people replacer; it’s a professional enhancer. Most companies have found human employees to be indispensable, their efforts being enhanced – not replaced – by machines. The near-full employment in the US is proof that despite the advancement of ChatGPT and other next-generation bots assuming some work processes, AI remains dependent on humans. Even when deployed to fill certain tasks, generative AI chatbots’ abilities remain limited to doing one discrete information-oriented task, basically, at a time, before they need a human to review what they’ve done and then move to the next thing. “We know from many areas that have rapidly automated that they don’t deliver the types of returns that they promised,” observes Daron Acemoglu, an economics professor at MIT. Infact, going beyond supplementing and supporting the work of trained radiologists in analyzing X-ray images to diagnose disorders, AI can in some respects actually increase demand for radiology, “because AI helps us get more information out of images than we could do before,” said Bradley Erickson, M.D., a neuroradiologist who runs the Mayo AI Lab. AI-native enterprises will displace only AI-naive businesses.

AI, what’s your ROI?

Lots of AI ventures are actually moonshots that never quite landed. Of the 62 AI models that were programmed to diagnose COVID from chest X-rays or CT scans, none of the models was fit for clinical use due to various flaws and biases. Despite predictions of the specialty’s demise, AI body scan readers have failed to replace human radiologists. Safe, dependable self-driving cars too have so far proved elusive. And a 2023 Altair Engineering survey reveals that 36 percent of AI projects flopped. Numbers like these necessitate a very thorough vetting of AI’s applicability and feasibility by business leaders before piloting projects. 

So long

The rise of AI isn’t a cause for alarm but a call to action. The path ahead requires not passive observation but active participation. As we stand on the brink of this transformative era, seize the opportunity to wield AI as a strategic tool, ushering in an age where innovation knows no bounds. Welcome it, harness its power, and watch it redefine professional successes.