AI Predictions for 2024

2023 was a whirlwind year for AI news and development. As we look ahead, I'd like to offer six non-exhaustive predictions on AI trends I believe we'll see emerge in 2024. In no particular order:

1. The AI chat prompt will establish itself as a “universal interface”
2. AI tooling will promote a disintermediation of data access
3. We can expect to see ever-eroding trust in reality
4. The deliberate investment in healthcare AI will begin to bear real fruit
5. Alliances will be formed between companies whose names don’t start with “Micro*”
6. AI Alignment, regulation, and bias will dominate the philosophical discussion space

Let's break it down.

The AI chat prompt will establish itself as a “universal interface”, to both spectacular success and resounding failure. Just as the Google search box forever changed information retrieval, and the touch screen has displaced tactile controls like knobs and buttons, so too will the humble text input replace many advanced UI features on websites. This will spur an increase in the accessibility and availability of information, as well as the speed of retrieval. However, as companies are wont to do with a good thing, this trend will go too far. Like the touchscreen, the prompt is a simpler and more cost-effective element to design (despite, in the case of the touchscreen, at least in an automotive context, near-universal complaints on complexity and safety concerns). I also expect that this trend will come full circle in the years ahead, with pushback from power users who find prompts ungainly and inefficient. Eventually, actual interfaces to support AI workflows (with buttons, links, and inputs) will return, but they will do so in more customized, bespoke, applications. That is, in high-end interfaces by companies who can afford to invest in good design. Much like luxury vehicle manufacturers who strive to carve out distinct high-end niches, we’ll see a similar arc with AI interfaces.

The simplified interface will in turn promote a disintermediation of data access, as everyone becomes an analyst. This generally won’t be a good thing. Very quickly, many people will discover that they don’t want to be analysts – they just want good information and actionable answers to difficult questions. There will be lots of discussion about how this enables C-suite leaders to bypass analytics teams and go directly to the data. I expect that the more likely outcome will be an empowerment of analytics teams to create better and more meaningful metrics, faster. Most leaders don’t need real-time access to data; they need timely and relevant insights. In the business world, as in the LLM world, context is king. Knowing how to interrogate data, and having the experience to ask the right questions becomes even more valuable, as the tools themselves bear more of the heavy lifting of the analysis itself. Undoubtedly, AI will transform the daily workflows of analytics teams - alongside much of the knowledge industry. But the need for organizations to train, coach, and retain top-tier analytic talent will be more relevant than ever, as the analytics function becomes an even greater agent of scale.

Further down the dystopian spectrum, we can expect to see an ever-eroding trust in reality, as AI-generated advertisements, images, and “grassroots” armies of AI bots spread disinformation and misinformation across the Internet. We will see this starkly in the run-up to the 2024 presidential election, particularly if the race is closely contested. It will only be marginally apparent that this information is AI-generated. That will not matter, as a subset of the population, regardless of party, will steadfastly refuse to apply any semblance of reasoning when presented with AI content. Quantity, not quality, is of the essence, and the flood of content will quickly create new problems across our political system. And that's to say nothing about the sinister lurking presence of ever improving deepfake technology. Thankfully, smart people are working hard to combat this issue, though I fear that any solution will come with its own challenges.

In more positive news, the deliberate investment in healthcare AI will begin to bear real fruit and kick off a virtuous cycle of health advances. Major EHR vendors will begin to widely deploy their embedded AI technologies, directly assisting advances in image (X-Ray, MRI, etc.) processing, transcription, translation, and care delivery. There will also be pushback, as these systems struggle to gain traction in some areas, or are held back over concerns about quality and ethics. Surely, these are issues that must be carefully tread and solved for, but the arrow points up and to the right for healthcare progress. Advances in cancer care and individually targeted treatments will benefit from the accelerated R&D cycles that AI will enable. This gets most interesting as we look out 25 years, and begin to wrestle with the (seemingly good?) impact of 150 year lifespans. The healthcare industry continues to see robust AI investments, perhaps counter-intuitively given the segment's historical reticence toward new technology adoption. This will be an exciting space to watch.

New alliances will be formed between companies whose names don’t start with “Micro*”, and we'll see a convergence of capabilities toward a mean. This one is a freebie. Don't expect the entrenched contenders to give in easily; Google, Meta, Amazon, Salesforce, and Oracle will continue their investments in AI research and product development at breakneck speed. Most generative AI capabilities will converge toward the same "generally good" set of capabilities, with contenders jockeying to spike in lucrative domains or feature sets (see Google's strong push into healthcare). At the same time, advances in open source models will increasingly commodify this "generally good" feature set, which will in turn further invigorate open source development teams.

AI Alignment, regulation, and bias will dominate the philosophical discussion space. With advances happening so quickly, true believers understand the need to think five years out - but they also understand that five years from now, we will likely be in an exponentially different place, not a linearly different one. Discussion on AI alignment, already a contentious topic within firms such as OpenAI, will break into the open, as humanity grapples with the broad impact of AI. Conversations around regulation will only accelerate as AI claims jobs and attention spans (particularly as the EU continues to take a leading role), and we begin to ask more loudly: who is serving whom, and to what end?

It’s going to be an exciting year. Buckle up. 🏍️