• Dr Scott Zoldi, Chief Analytics Officer at FICO


• 18.12.2023 12:00 pm
#AI #GenAI #predictions

Like Sam Altman himself, I was quite surprised by the recent whirlwind speed of OpenAI’s revolving door: He’s out! He’s trying to get in! He’s at Microsoft! He’s back at OpenAI! This collective corporate behavior can only be described as dysfunctional and destabilizing, qualities that OpenAI’s Generative AI (GenAI) tool, ChatGPT, also earned a reputation as being. It’s pretty impossible to depend on anything that acts erratically, but that is exactly what has happened in 2023, with the world (including Microsoft) quickly becoming dependent on ChatGPT. Therefore, my artificial intelligence (AI) predictions for 2024 are also prescriptive: we––the tech industry, business world, cheating college students and the world in general––need to take another piece of advice from TIME Magazine’s Person of the Year, Taylor Swift. Instead of acting on Generative AI with truly blind faith, we need to calm down from the AI hoopla of 2023 and retrain our focus on practical artificial intelligence. Here are four ways I believe that will happen:

1. Auditable AI will make accountability cool

The last few weeks of 2023 showed us how organizational dysfunction at AI firms can translate into a disturbing opacity about how AI tools work. In 2024 I expect to see a backlash against opacity, and a resulting surge in market demand for AI model transparency. Auditable AI will become mandatory for artificial intelligence models; with so much riding on AI decisioning and Generative AI outputs, understanding how models operate will no longer be a nice-to-have. Auditability requirements will catalyze AI governance structures so much that, in 2024, accountability will be cool.

Blockchain technology makes Auditable AI readily attainable. Blockchain has matured and been applied in extremely useful ways beyond cryptocurrency, such as for AI model management governance. Blockchain will be a key component of Auditable AI strategies, through immutable chains of AI model requirements met, tested and accepted. This will help to reduce the risk of negative AI outcomes.

In 2024, I anticipate immutable blockchain technology to become more ingrained in the data science ecosystem. Auditability and transparency will be part of the conversation much earlier as part of demonstrable Responsible AI practices. As more organizations adopt auditability into their workstreams, there will be an appreciable improvement in model auditability and adherence to model development standards.

If 2023 has taught the world anything, it’s that rushing AI models and products into production is flat-out dangerous. Organizations that want their AI systems to be trusted will need to adopt Auditable AI.

2. Small will be beautiful

Large AI models, including large language models (LLMs) like ChatGPT, Google Bard AI, et. al., are incredibly unwieldy, difficult and expensive to build and operate. Furthermore, they are not easily retrained without cannibalistic entropy.

I have long held an unorthodox belief in “explainability first, predictive power second,” a core tenet of Responsible AI. In 2024, small, highly explainable AI models will be accepted and embraced as being more effective than outsized large models, as organizations focus on practical, reliable, net-positive business outcomes.

3. Humans will reassert themselves

As 2023 saw humans giddily outsourcing much thinking to Gen AI. In 2024 we will realize that we have a responsibility to be the overlords of this new technology, not the other way around. Why? Because in outsourcing one’s critical thinking to a chatbot, with every prompt we cede a bit of our soul and creativity. Critical thinking atrophies.

I believe that in 2024 humans will move past Gen AI’s ‘wow’ factor and focus on getting practical. People need to critically understand, trust and validate the AI tools they leverage, and in the coming year will prioritize finding new ways to work together using artificial intelligence as a tool instead of a magic wand. Doing so will produce far better, ethical and more responsible outcomes.

4. Gen AI will become less dramatic, more pragmatic

It’s hard to imagine Generative AI ever falling far from top-of-mind. But that is what happens with all technologies as they move, with almost clockwork reliability, through their hype cycle.

While its impact has been unprecedented, in 2024 Generative AI will follow in the hype cycle footsteps of other breakthrough technologies like blockchain. At their outset, both technologies appeared to be powerful novelties with great but unknown potential. As blockchain has matured and been applied in extremely useful ways beyond cryptocurrency – such as for model management governance – Gen AI will find similar tributary applications that will be less dramatic, but far more pragmatic.

Where will we land in 2024? Will we enter into a new, calmer era of AI? I sure hope so.