The Dark Side of AI

Over the last five thousand years, humanity has witnessed many groundbreaking technologies that transformed our lives. The groundbreaking technology ranges from a wheel in the 4th millennium BCE to a plow, the printing press, the steam engine, and the internet in the 2000s AD. AI is the next in order.

Though AI has existed for over two decades, it is still nascent. We and corporations benefit from AI daily, but AI is still applied in a limited capacity. We have seen many early benefits and share a few concerns. Now AI is reaching an inflection point where the capabilities are used not just in a specialized situation (narrow AI) but can be applied in a generic case (general AI). With the popularity of ChatGPT recently, there is a renewed sense of purpose and excitement around AI among us and corporations. If AI technology risks are unaddressed in the next few years, then in the long run, the risks outweigh the benefits, and it might be too late for humanity to change course.

To put AI excitement in context: In Q1 2023, VCs invested more than $1.6B in generative AI startups alone. You can’t scroll Twitter, YouTube videos, or new feeds without an article on generative AI. Half of US companies have been using ChatGPT in the last three months. Open AI, the ChatGPT owner, is already forecasting $1B in revenue within a year of product launch. This is incredible.

We are at an inflection point where technology has become pervasive, with most corporations trying to benefit from AI. But let’s not ignore the risks. It is more critical now than ever to have an active conversation on reducing the negative AI implications on humanity.

Here are a few areas we need to address risks:

Job Loss: This week Goldman Sachs reported that AI systems like ChatGPT could impact 300 million jobs worldwide. Jobs that involve repetitive or routine tasks like customer service, professions that require a high degree of text-based communication like writers, and professions that require data analysis like financial analysts and accountants are a higher chance of getting impacted.

In just three months, half of the US companies started using ChatGPT and have already replaced workers. Just give global companies another year; they will identify areas to apply regenerative AI technology to increase productivity and cut costs impacting millions of people globally.

Corporations will become more productive and profitable. But millions of people will lose jobs eventually.

Ethics: AI has three major ethical concerns. 1) enabling surveillance, 2) pollution of public disclosure by social media bots, and 3) algorithmic bias. Algorithmic bias is the first concern experienced by many users of ChatGPT.

Algorithmic bias has been a well-documented problem in AI for the last three decades. We heard sad stories where algorithms recommended less medical care to African Americans than white counterparts in hospitals, whites males are perceived as better intellectuals than others, amazon recruiting tool discriminates female applicants, Meta had to shut down ChatGPT equivalent in just three days because it spewed false and racist information.

The algorithmic bias is induced due to data set limitations (data is uncurated, scrapped from the internet), limited diversity in algorithm creators (primarily white males), and the urge to publish unsafe models to monetize quickly. These limitations are easy to overcome if there is an oversight on new AI models similar to FDA for drugs.

Wealth inequality: Chat GPT and other AI technologies will cause job loss to millions of people. Our (USA) wealth inequality is the highest among all developed countries and has been at an all-time high since world war II. The biggest beneficiaries of the AI boom will be venture capitalists, workers at AI companies, and large shareholders of corporations saving money with AI, which is a small percentage of our population.

Millions of office and white-collar workers will be affected by Chat GPT. These workers will lose their jobs or face downward mobility, and they need to learn new skills and find new jobs. The remaining workers may not benefit from AI either. Corporations making more money from applying artificial intelligence will reinvest for future profits or return profits to their investors. This unsustainable trend will further deteriorate wealth inequality.

Control: Intelligence equates to the ability to act rationally. AI meets this criterion, but a human defines it, and humans often find it challenging to specify clear objectives for many real work problems. A simple recommendation engine could recommend a “stool” when you purchase a “rope,” as the correlation between rope and stool exists. Though this is an algorithmic mistake, this problem is profound.

To ensure AI doesn’t harm humans or the environment, experts recommend built-in safety rules, making the algorithm transparent, aligning the AI system with ethical and moral principles, ensuring personal data captured is secure and private, and making the AI systems and creators accountable for the decisions of the AI system.

AI technology implications go beyond economic, legal, political, and geopolitics. I hope all the AI creators, consumers, and lawmakers globally share the collective responsibility and urgently deal with matters that make the technology safe for everyone.