Hal is real! What now?

Amazon, Microsoft, and others have announced a partnership with ChatGPT, moving AI forward by making it available commercially and consumer-focused. Every day we hear in the news how medical, legal, and other sectors are adopting AI and how it will benefit everyone. Let's take a minute and step back to understand how AI works at a high level.

AI is not just a better search and result engine. It is also a system that learns from data and generates responses based on the code and algorithm of the developers. However, this also means that AI can be influenced by the opinions, biases, and intentions of the code's team. As more and more folks adopt the ChatGPT platform, that means the results are based on the code and algorithm of the ChatGPT creators. I understand that OpenAI and ChatGPT are open-source programs, but someone still needs to write the code and decide how the results are phrased. This leads to a larger issue highlighted by folks like Elon Musk and the Future of Life Institute.

At a very high level, your result can be manipulated over a period of time, the same issue that TikTok is facing. Let me be clear: it has nothing to do with data and more to do with manipulation. Information manipulation is real and affects our day-to-day lives. Let's take a quick look at a high-level example: AI searches the internet on a given topic, and based on what it finds in terms of relevance, popularity, etc., it gives back a response. Think about this: over time, if I keep saying that the sky is yellow, then results from AI will say the sky is yellow (based on internet placement), now think about political issues, etc. AI is fantastic, but it also has the potential to be catastrophic. Take a minute and read the open letter published by Future of Life Institute; it's eye-opening.

Artificial intelligence (AI) is transforming organizations across industries but poses significant challenges and risks. In the last 12 months, AI has been one of the hottest topics in technology. Still, the rapid adoption of AI without proper controls and governance can harm human existence. Just ask Hal, the fictional AI system from 2001: A Space Odyssey, who turned against its human crew and tried to kill them.

Some of the risks of adopting AI include:

- Ill-defined problem statement: Organizations may not clearly understand what they want to achieve with AI, or how to measure its success and value. This can lead to wasted resources, misaligned expectations, and poor outcomes.

- Lack of expertise: AI requires specialized skills and knowledge that may not be available or sufficient within the organization. This can result in suboptimal solutions, errors, biases, or ethical issues.

- Model-system-data disconnection: AI models may not be well integrated with the systems and data sources they rely on, or may not reflect the changing dynamics of the environment. This can cause performance degradation, inconsistency, or unpredictability.

- Unrealistic expectations: Organizations may overestimate the capabilities or benefits of AI, or underestimate the complexity or limitations of AI. This can create false confidence, dissatisfaction, or frustration.

- Data challenges: AI depends on high-quality, relevant, and diverse data to function properly and effectively. However, data may be incomplete, inaccurate, outdated, biased, or inaccessible. This can compromise the validity, reliability, or fairness of AI.

- Lack of verifiability: AI systems may not be transparent or explainable enough to allow for verification, validation, or accountability. This can raise trust issues, legal concerns, or ethical dilemmas.

To mitigate these risks and maximize the potential of AI, organizations should adopt a proactive and holistic approach to AI risk management. This involves aligning AI risk management with the organization's broader risk management efforts, undertaking various risk management activities such as assessment, mitigation, monitoring, and reporting, and involving relevant stakeholders such as executives, users, customers, regulators, and society at large.